Publications

2026

1. Zero-Shot Test-Time Canonicalization Using Out-of-Distribution Scoring

   Dominik Lindner, Johann Schmidt, Tom Siegl, and 2 more authors

   In Machine Learning and Knowledge Discovery in Databases. Research Track, Sep 2026

   Abs DOI

   Pretrained vision models often misclassify inputs that are rotated, scaled, or sheared, even though these affine transformations leave the object class unchanged. Robustness is usually restored either by building equivariance into the architecture or by retraining with augmentation, both of which require changing or retraining the model. Test-time canonicalization instead leaves the classifier untouched. It undoes the transformation of each input, mapping it to a canonical form near the training distribution before classification. Existing canonicalizers, however, rely on a narrow set of logit-based energy scores and bespoke search procedures, leaving the design space of scoring functions and optimizers unexplored. We reframe canonicalization as out-of-distribution (OOD) detection, which lets any OOD score serve as the energy minimized over transformations. Across benchmarks ranging from handwritten characters and sketches to natural images and 3D point clouds, we systematically evaluate around twenty OOD scores and nine search algorithms, finding that distance-based scores paired with random search and local refinement perform best overall. Because canonicalizing an already-aligned input can hurt accuracy, we add a gated mechanism that transforms an input only when its OOD score indicates this is needed, preserving most in-distribution accuracy while retaining the robustness gains on transformed inputs. Code is available at https://github.com/johschm/its.

2. Geometrically Constrained and Token-Based Probabilistic Spatial Transformers

   Johann Schmidt, Tom Siegl, Martin Becker, and 1 more author

   In 2nd International Workshop on Biomedical Image and Signal Computing for Unbiasedness, Interpretability, and Trustworthiness, Sep 2026

   Abs HTML

   Spatial transformations such as rotation and scale obscure the morphological cues needed for accurate image classification. Careful consideration is required for reliable use in high stakes settings. A model should stay robust under such transformations, expose why a correction was applied, and signal when its input is ambiguous. While geometrically equivariant architectures provide a mathematically grounded solution, they often limit model flexibility through strict symmetry constraints and incur significant computational overhead. Spatial Transformer Networks (STNs) offer a data-driven, flexible alternative for learning pseudo-equivariances to affine transformations. However, STNs have historically been restricted to convolutional architectures and suffer from training instability. To address this, we introduce a novel STN framework. It leverages the global modeling capabilities of transformers to regress the affine transformation acting on the input. For this, we decompose affine transformations into interpretable primitives, regressed under adaptable geometric constraints, thereby preventing the training instability typically caused by degenerate transformations. By sharing weights between the localization network and the classification backbone, the framework requires minimal computational overhead. Extensive experiments on challenging insect biodiversity and medical imaging benchmarks demonstrate that our approach achieves superior predictive performance under diverse spatial transformations while maintaining high efficiency. Code is available at https://github.com/johSchm/TokenSTN.

3. Development of an automated mechanical testbed for efficient experiments on smart composites

   Emanuel von Cramon-Taubadel, Michelle Hedvard, Martin Becker, and 1 more author

   tm - Technisches Messen, Sep 2026

   Abs DOI

   Impact detection for carbon-fiber reinforced plastics plays an important role in structural health monitoring systems. Such systems can extend the service life of these valueable engineered materials and thus contribute to improving their viability for demanding applications such as aerospace. The proliferation of machine learning (ML) has enabled the use of low-cost micro-electromechanical systems sensors in impact monitoring. ML models trained on impact data require consistent datasets containing a large number of individual impact samples. To advance research in this area, we developed an automated testbed for collecting low-energy impact data on composite structures. Since the testbed is fully computer-controlled, the generated impacts are precise, repeatable, and inexpensive. Additionally, we developed a complete data processing pipeline that further reduces the effort required for large-scale impact experiments. Emphasizing runtime stability allows the testbed to support error recovery and collect over 1,750 impacts per hour. Consequently, we can to collect two large datasets daily, each containing over 17,500 impact samples.

4. Maternal overweight and obesity during pregnancy are associated with proinflammatory changes in the serum proteome and lipidome.

   Marie Albrecht, Kutalmış Coşkun, Hanno Steen, and 12 more authors

   Sep 2026

   Poster, Adipositas Kongress 2026, Deutsche Adipositas Gesellschaft (DAG), Potsdam

   DOI
5. Dynamic trajectories of neonatal laboratory measurements in the first 26 weeks of life

   Seyedeh Neelufar Payrovnaziri, Sayane Shome, Jean Costello, and 22 more authors

   npj Digital Medicine, Aug 2026

   Abs DOI

   Laboratory measurements are vital for obtaining quantitative updates on patient’s health status. This study focused on analyzing 1173 distinct types of laboratory measurements collected from 34,605 infants, including ~12% born preterm (\textless37 weeks gestation) and ~88% born term admitted to the Stanford Children’s Hospitals utilizing electronic health records spanning from birth to 26 weeks of age. Our objective was to uncover temporal patterns and interdependencies among the measurements to characterize longitudinal laboratory trajectories during early infancy and identify potential biomarkers during this critical window of neonatal care. Correlation networks were constructed, revealing clustering patterns within the measurements. Notably, a distinct trend was observed in neutrophil counts, exhibiting a decreasing trend until around 12 weeks, followed by a subsequent increase until 26 weeks. This inflection point may reflect a transition phase in early immune system maturation beyond the immediate neonatal period, a stage that remains relatively under characterized in clinical studies. This significant observation was validated with the de-identified laboratory measurements recorded for 10,765 infants admitted in UCSF Benioff’s Hospital. Further, we explored the underlying potential clinical factors associated with the observed trend and found a significant association between observed white blood cell counts and the Phecode category, which includes congenital heart diseases (CHD). This data-driven analysis provides insights into immune trajectories during early infancy [https://neonatesmeasures2025.streamlit.app/] highlighting the potential for applying data analytics for prognosis and early medical interventions.

6. CytoBERT: A Foundation Model for Cytometry Data

   Syed Abdul Haseeb Qadri, Bjarne C. Hiller, Felix Blanke, and 7 more authors

   Aug 2026

   arXiv:2608.14414 [cs.LG]

   Abs DOI

   Cytometry measures the complex characteristics of single cells (e.g., counts and protein expression of immune cells) and is widely used across immunological research and clinical settings. However, cytometry data is highly heterogeneous and unstandardized due to experimental protocols and the choice of measured features. While machine learning methods hold the potential to gain deeper insights into cell biology, these challenges make them difficult to apply and transfer across studies. Recent advances in foundation models can alleviate these issues, but corresponding approaches are still scarce in this field. To address this, we provide CytoBERT, a publicly available, open-source, open-weight foundation model for single-cell cytometry data with variable marker panels. CytoBERT is pretrained in a self-supervised manner on a large-scale cytometry corpus (15 human datasets with heterogeneous marker panels and more than 50 million cells) curated through marker standardization, enabling it to learn transferable inter-marker relationships within cells. Fine-tuning CytoBERT for sample-level classification demonstrates that transfer learning across heterogeneous cytometry datasets is feasible, providing a starting point for scalable, generalizable cytometry analysis. Code is available on GitHub

7. Übergewicht und Adipositas in der Schwangerschaft sind mit proinflammatorischen Veränderungen der Fett- und Proteinzusammensetzung im Serum assoziiert (ID 12)

   Marie Albrecht, Kutalmış Coşkun, Hanno Steen, and 12 more authors

   In Adipositas - Ursachen, Folgeerkrankungen, Therapie, Aug 2026

   DOI
8. Network-based disruption analysis of biological coordination

   Martin Becker, Dorien Feyaerts, Ina Annelies Stelzer, and 23 more authors

   Patterns, Jul 2026

   DOI
9. PregMedNet: Multifaceted maternal medication impacts on neonatal complications

   Yeasul Kim, Ivana Marić, Chloe M. Kashiwagi, and 28 more authors

   Nature Communications, Jul 2026

   Abs DOI

   While medication use is common among pregnant women, medication safety remains insufficiently characterized because studies in pregnant women are challenging due to safety concerns. The recent digitization of healthcare databases and advances in computational methods have created new opportunities for large-scale, retrospective drug safety evaluations. Here, we present PregMedNet, a platform that characterizes multifaceted maternal medication associations on neonatal outcomes during pregnancy, covering more than 27,000 drug-disease pairs across 1,152 medications and 24 outcomes. These results encompass known and additional odds ratios (ORs), adjusted ORs, and drug-drug interactions, systematically analyzed using nationwide claims data and an advanced machine learning pipeline. Notably, one of the associations identified in this study is supported by in vivo experiments, increasing confidence in PregMedNet’s findings and highlighting the utility of claims data and machine learning for perinatal medication safety studies. Additionally, potential biological mechanisms underlying the associations are explored using a graph learning method, providing candidate pathways for future mechanistic investigations. We expect that PregMedNet will contribute to advancing maternal medication safety and improving neonatal outcomes by providing extensive, multifaceted drug safety information on this previously underrepresented population.

10. Explainable AI for impact detection in smart composites: insights into transfer learning and model robustness

    Emanuel von Cramon-Taubadel, Martin Becker, and Philipp Weißgraeber

    Smart Materials and Structures, Jul 2026

    Abs DOI

    Smart composite structures have substantial utility in high-performance applications such as aerospace and wind energy, where real-time impact detection is essential for structural health monitoring (SHM). machine learning (ML) enables the use of low-cost micro-electromechanical systems (MEMS) accelerometers for impact localization; however, existing approaches often lack robustness outside controlled laboratory conditions. In this study, we propose a full-stack framework for systematic evaluation of ML methods for impact localization on smart composites. The contributions include: (i) a customizable experimental setup for efficient data acquisition, (ii) a software framework for training and benchmarking ML and deep learning models under varying conditions, (iii) a regression-based architecture with improved accuracy, robustness, and transferability, (iv) transfer learning for efficient adaptation across structural configurations, and (v) explainable artificial intelligence (XAI) components for physical interpretation. Our results show that the proposed regression approach is robust to small deviations in the experimental setup ( accuracy loss), though larger variations, such as structural modifications, significantly degrade the models’ performance. Regression models consistently outperform classification methods, with gains of 1%–5% under minor variations and up to 33% under severe changes. Transfer learning reduces required training data by 88% while maintaining accuracy within 2% of a fully trained model. Integrated gradients further provide preliminary insight into the underlying localization mechanisms. These results demonstrate improved robustness, data efficiency, and interpretability, supporting the deployment of data-driven SHM in real-world smart composite structures.

11. Machine learning–guided multimodal profiling defines perturbed immune states at the time of cancer diagnosis

    Peggy Berlin, Amin Mirzaei, Felix Steinbeck, and 22 more authors

    Briefings in Bioinformatics, May 2026

    Abs DOI

    Altered immune states at the time of cancer diagnosis remain insufficiently characterized. Although circulating immune biomarkers offer a promising, non-invasive way of analysing systemic tumour–host interactions, their potential remains poorly defined. Here, we present an integrated multi-omics analysis of peripheral blood mononuclear cells from treatment-naïve cancer patients, minimizing confounding by therapy-induced immune changes, combining immune phenotyping (flow cytometry, FC), multiplex cytokine profiling, and single-cell RNA sequencing (scRNA-seq). Compared with healthy donors, patients exhibited widespread immune dysregulation, including expansion of FOXP3+ regulatory T cells, depletion of CD16+CD11b+ monocytes and CD56^dim^ Natural killer (NK) cells, and elevated plasma IL-6 and IL-4 levels. scRNA-seq identified cancer-associated immune signatures, notably consistent upregulation of THBS1 and CH25H, indicative of systemic imprinting by tumour-derived cues. We further developed machine learning-guided models integrating single-cell multi-omics data (sc-FC and scRNA-seq) to characterize cancer-associated immune patterning and cancer type–related signal structure, while providing biologically interpretable feature attribution across modalities. The models achieved robust classification performance within the cohort and revealed modality-spanning features linked to immune state alterations.Together, these findings establish a framework for immune-based, multi-omics profiling of peripheral blood and provide a resource for discovering circulating cancer-associated immune signatures. This supports future development of immune-based diagnostics and disease monitoring approaches.

12. A machine learning model for frailty based on wearable device measurements

    Anthony Culos, Asier Manas, Kie Shidara, and 16 more authors

    Communications Medicine, Feb 2026

    Abs DOI

    Frailty is an important factor in human aging associated with a broad range of adverse outcomes. Frailty metrics are time intensive to collect making them difficult for larger scale application.

13. Quantitative assessment of neonatal health using dried blood spot metabolite profiles and deep learning

    Alan L. Chang, Jonathan D. Reiss, Anthony Culos, and 26 more authors

    Science Translational Medicine, Jan 2026

    Abs DOI

    Neonatal prematurity leads to considerable morbidity and mortality, partly because of acquired conditions such as bronchopulmonary dysplasia (BPD), intraventricular hemorrhage (IVH), necrotizing enterocolitis (NEC), and retinopathy of prematurity (ROP). Standard gestational age and birthweight-based classifications of prematurity inadequately capture the variation in newborns’ health outcomes, creating an urgent need to develop risk stratification tools for vulnerable newborn infants to initiate the most appropriate care pathways as early as possible. We hypothesized that the metabolic profiles of newborn infants capture additional risk information beyond current measures. A total of 13,536 newborn screening (NBS) blood spot tests from preterm infants in California with linked clinical outcomes of prematurity were used to develop an NBS-based metabolic health index to stratify preterm infants at risk for BPD, IVH, NEC, and ROP (12,096 cases with one or more conditions and 1440 controls) through a deep learning model that provides a single index score in tandem with subgroup discovery to identify individuals with the strongest metabolite biomarker signals for adverse outcomes of prematurity. This metabolic health index captured risk signals that were distinct from gestational age and birthweight and outperformed other machine learning algorithms and clinical risk variable-based models in stratifying at-risk individuals for adverse outcomes of prematurity. The metabolic health index was externally validated in an independent retrospective cohort of 3299 very premature newborns from Ontario, Canada (2117 cases and 1182 controls), which recapitulated common metabolic risk subgroups. In summary, combining widespread metabolite screening with deep learning established a generalizable biological risk metric of prematurity.

14. When Does Causal Background Knowledge Help Few-Shot Tabular Learning?

    Kutalmış Coşkun, Felix Blanke, Vanja Sophie Cangalovic, and 5 more authors

    2026

    In preparation
15. Maternal lipidome is shaped by maternal overweight and obesity during pregnancy

    Marie Albrecht, Kutalmış Coşkun, Anna Worthmann, and 5 more authors

    2026

    European Lipoprotein Club
16. Informed, but Not Always Improved: Challenging the Benefit of Background Knowledge in GNNs

    Kutalmış Coşkun, Felix Blanke, Vanja Sophie Cangalovic, and 5 more authors

    2026

    Under review

    Abs

    Biological background knowledge (BK) graphs, such as protein-protein interactions (PPIs), are widely used to guide graph neural networks (GNNs) toward biologically informed predictions, especially in low-sample, high-dimensional settings. Yet a basic question remains largely untested: do these models actually use the provided biological graph, and what happens when that graph is incomplete, noisy, or biased? In this work, we systematically probe BK utilization through random and targeted perturbations across two cancer-related tasks: graphlevel cancer subtype classification and node-level cancer gene detection. Contrary to expectations, in cancer subtype classification, uninformed baselines outperform BK-informed GNNs and graph perturbations have negligible effect, suggesting that the provided biological networks are barely used at all. In cancer gene detection, we find stronger but highly model-dependent use of the PPI graph, ranging from topology-dominated behavior to more balanced use of graph structure and node features. We complement our real-world findings with controlled synthetic graph- and node-classification experiments to provide a positive control: when graph informativeness is ensured by construction, informed GNNs outperform uninformed baselines and degrade predictably under perturbation. Overall, our results challenge the assumption that adding biological BK automatically improves GNN performance, and show that its contribution must be explicitly tested through perturbations, a shift we argue should become standard practice when reporting BK-informed GNNs. Our code is publicly available at bckrlab.org/p/bk-buster.

17. Evolving Markov Chains: Online Mode Discovery and Recognition from Data Streams

    Kutalmış Coşkun, Borahan Tümer, Bjarne C. Hiller, and 1 more author

    IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026

    Abs DOI

    Markov chains are simple yet powerful mathematical structures to model temporally dependent processes. They generally assume stationary data, i.e., fixed transition probabilities between observations/states. However, live, real-world processes, like in the context of activity tracking, biological time series, or industrial monitoring, often switch behavior over time. Such behavior switches can be modeled as transitions between higher-level modes (e.g., running, walking, etc.). Yet all modes are usually not previously known, often exhibit vastly differing transition probabilities, and can switch unpredictably. Thus, to track behavior changes of live, real-world processes, this study proposes an online and efficient method to construct Evolving Markov chains (EMCs). EMCs adaptively track transition probabilities, automatically discover modes, and detect mode switches in an online manner. In contrast to previous work, EMCs are of arbitrary order, the proposed update scheme does not rely on tracking windows, only updates the relevant region of the probability tensor, and enjoys geometric convergence of the expected estimates. Our evaluation of synthetic data and real-world applications on human activity recognition, electric motor condition monitoring, and eye-state recognition from electroencephalography (EEG) measurements illustrates the versatility of the approach and points to the potential of EMCs to efficiently track, model, and understand live, real-world processes.

2025

1. SPACEc: a streamlined, interactive Python workflow for multiplexed image processing and analysis

   Yuqi Tan, Tim N. Kempchen, Martin Becker, and 11 more authors

   Nature Communications, Nov 2025

   Abs DOI

   Multiplexed imaging has transformed our ability to study tissue organization by capturing thousands of cells and molecules in their native context. However, these datasets are enormous, often comprising tens of gigabytes per image, and require complex workflows that limit their broader use. Current tools are often fragmented, inefficient, and difficult to adopt across disciplines. Here we show that SPACEc, a scalable Python platform, streamlines spatial imaging analysis from start to finish. The platform integrates image processing, cell segmentation, and data preprocessing into a single workflow, while improving computational performance through parallelization and GPU acceleration. We introduce innovative methods, including patch proximity analysis, to more accurately map local cellular neighborhoods and interactions. SPACEc also simplifies advanced approaches such as deep-learning annotation, making them accessible through an intuitive interface. By combining efficiency, accuracy, and usability, this platform enables researchers from diverse backgrounds to gain deeper insights into tissue architecture and cellular microenvironments.

2. An unsupervised XAI framework for dementia detection with context enrichment

   Devesh Singh, Yusuf Brima, Fedor Levin, and 44 more authors

   Scientific Reports, Nov 2025

   Abs DOI

   Explainable Artificial Intelligence (XAI) methods enhance the diagnostic efficiency of clinical decision support systems by making the predictions of a convolutional neural network’s (CNN) on brain imaging more transparent and trustworthy. However, their clinical adoption is limited due to limited validation of the explanation quality. Our study introduces a framework that evaluates XAI methods by integrating neuroanatomical morphological features with CNN-generated relevance maps for disease classification. We trained a CNN using brain MRI scans from six cohorts: ADNI, AIBL, DELCODE, DESCRIBE, EDSD, and NIFD (N = 3253), including participants that were cognitively normal, with amnestic mild cognitive impairment, dementia due to Alzheimer’s disease and frontotemporal dementia. Clustering analysis benchmarked different explanation space configurations by using morphological features as proxy-ground truth. We implemented three post-hoc explanations methods: (i) by simplifying model decisions, (ii) explanation-by-example, and (iii) textual explanations. A qualitative evaluation by clinicians (N = 6) was performed to assess their clinical validity. Clustering performance improved in morphology enriched explanation spaces, improving both homogeneity and completeness of the clusters. Post hoc explanations by model simplification largely delineated converters and stable participants, while explanation-by-example presented possible cognition trajectories. Textual explanations gave rule-based summarization of pathological findings. Clinicians’ qualitative evaluation highlighted challenges and opportunities of XAI for different clinical applications. Our study refines XAI explanation spaces and applies various approaches for generating explanations. Within the context of AI-based decision support system in dementia research we found the explanations methods to be promising towards enhancing diagnostic efficiency, backed up by the clinical assessments.

3. Multi-omics and machine learning integration for early cancer detection: Identifying immune dysregulation and novel biomarkers in treatment-naïve patients

   C. Maletzki, P. Berlin, M. Mirzai, and 18 more authors

   Annals of Oncology, Sep 2025

   DOI
4. Informed, but Not Always Improved: Challenging the Benefit of Background Knowledge in GNNs

   Kutalmış Coşkun, Ivo Kavisanczki, Amin Mirzaei, and 4 more authors

   In 24th International Workshop on Data Mining in Bioinformatics (BIOKDD 2025), Aug 2025

   arXiv:2505.11023 [cs.LG]

   Abs DOI

   In complex and low-data domains such as biomedical research, incorporating background knowledge (BK) graphs, such as protein-protein interaction (PPI) networks, into graph-based machine learning pipelines is a promising research direction. However, while BK is often assumed to improve model performance, its actual contribution and the impact of imperfect knowledge remain poorly understood. In this work, we investigate the role of BK in an important real-world task: cancer subtype classification. Surprisingly, we find that (i) state-of-the-art GNNs using BK perform no better than uninformed models like linear regression, and (ii) their performance remains largely unchanged even when the BK graph is heavily perturbed. To understand these unexpected results, we introduce an evaluation framework, which employs (i) a synthetic setting where the BK is clearly informative and (ii) a set of perturbations that simulate various imperfections in BK graphs. With this, we test the robustness of BK-aware models in both synthetic and real-world biomedical settings. Our findings reveal that careful alignment of GNN architectures and BK characteristics is necessary but holds the potential for significant performance improvements.

5. Electroencephalogram features support the retrogenesis hypothesis of Alzheimer’s disease: Exploratory comparison of brain changes in aging and childhood

   João Areias Saraiva, Martin Becker, Martin Dyrba, and 8 more authors

   Journal of Alzheimer’s Disease, Aug 2025

   Abs DOI

   BackgroundThe retrogenesis hypothesis (RH) suggests that the functional and cognitive decline observed in Alzheimer’s disease dementia mirrors in reverse order the brain development during childhood and adolescence.ObjectiveEquivalent electroencephalogram (EEG) patterns between older adults across different cognitive decline stages and children across different brain maturation stages were directly compared.MethodsTo capture the complex patterns that allow for such a comparison, a regression model was trained on EEG data from N = 510 older adults, at different stages of cognitive reserve, to identify EEG markers predictive of global cognitive status. The model was then applied on the same EEG markers of N = 696 children across different ages.ResultsThe model predicted MMSE scores with an average error of 2.53 and R2 of 0.80. When applied to children, predictions correlated positively with age (r = 0.73). Key predictors of cognitive function concordant in both populations were theta coherence (right frontal-left temporal/parietal), temporal Hjorth complexity, and beta edge frequency, supporting the RH.ConclusionsThese EEG features were inversely associated between older adults and children, supporting a functional underpinning of the retrogenesis model of dementia. Clinical validation of these biomarkers could favor their use in the continuous monitoring of cognitive function.

6. Pregnancy-acquired memory CD4+ regulatory T cells improve pregnancy outcome in mice

   Kristin Thiele, Christopher Urbschat, Julia Isabel Amambay Riquelme, and 12 more authors

   Nature Communications, Jul 2025

   Abs DOI

   Subsequent pregnancies are generally less prone to obstetric complications. A successful pregnancy outcome requires pivotal immunological adaptation to ensure immune tolerance towards the foetus. Thus, the lower risk for pregnancy complication during subsequent pregnancies may be attributable to immune memory mounted during first pregnancies. Here we identify higher frequencies of fetal-antigen-specific CD4+ regulatory T (Treg) cells both postpartum and in subsequent pregnancies in mice which are partly originating from trans-differentiated Th17 cells. Our functional experiments demonstrate that these CD4+ Treg cells have memory functions (CD4+ mTreg) and account for an improved fetal development and pregnancy outcome, also during adverse conditions, such as gestational sound stress. Using a high-throughput single-cell quantification method, we identify candidate markers for the detection of CD4+ mTreg cells, which include CXCR4 and CD274. Our findings thus contribute to the improved understanding of pregnancy-induced immune memory and foster the identification of immune targets aiming to reduce the risk for immune-mediated pregnancy complications.

7. Infusion of young donor plasma components in older patients modifies the immune and inflammatory response to surgical tissue injury: a randomized clinical trial

   Brice Gaudilliere, Lei Xue, Amy S. Tsai, and 22 more authors

   Journal of Translational Medicine, Feb 2025

   Abs DOI

   Preclinical evidence suggests that young plasma has beneficial effects on multiple organ systems in aged mice. Whether young plasma exerts beneficial effects in an aging human population remains highly controversial. Despite lacking data, young donor plasma infusions have been promoted for age-related conditions. Given the preclinical evidence that young plasma exerts beneficial effects by attenuating inflammation, this study examined whether administering a young plasma protein fraction to an elderly population would exert anti-inflammatory and immune modulating effects in humans, using surgery as a tissue injury model.

8. SubROC: AUC-Based Discovery of Exceptional Subgroup Performance for Binary Classifiers

   Tom Siegl, Kutalmış Coşkun, Bjarne C. Hiller, and 3 more authors

   In ECAI 2025, 2025

   Abs DOI

   Machine learning (ML) is increasingly employed in real-world applications like medicine or economics, thus, potentially affecting large populations. However, ML models often do not perform homogeneously, leading to underperformance or, conversely, unusually high performance in certain subgroups (e.g., sex=female {}wedge marital}_status=married). Identifying such subgroups can support practical decisions on which subpopulation a model is safe to deploy or where more training data is required. However, an efficient and coherent framework for effective search is missing. Consequently, we introduce }emph{SubROC}, an open-source, easy-to-use framework based on Exceptional Model Mining for reliably and efficiently finding strengths and weaknesses of classification models in the form of interpretable population subgroups. SubROC incorporates common evaluation measures (ROC and PR AUC), efficient search space pruning for fast exhaustive subgroup search, control for class imbalance, adjustment for redundant patterns, and significance testing. We illustrate the practical benefits of SubROC in case studies as well as in comparative analyses across multiple datasets.

9. Maternal IGHG locus duplications impair infants’ passive immunity

   Ioannis Belios, Dimitra E. Zazara, Ioannis Evangelakos, and 7 more authors

   medRxiv, 2025

   HTML
10. Evaluating the Fidelity of Explanations for Convolutional Neural Networks in Alzheimer’s Disease Detection

    Bjarne C. Hiller, Sebastian Bader, Devesh Singh, and 3 more authors

    In Bildverarbeitung für die Medizin 2025, 2025

    DOI
11. Cross-Sectional Associations Between the Electroencephalogram and Cognitive Status: Toward Scalable Monitoring Solutions

    João Areias Saraiva, Martin Dyrba, Martin Becker, and 4 more authors

    Alzheimer’s & Dementia, 2025

    Abs DOI

    Background Alzheimer’s disease (AD) strains healthcare systems in an aging population, emphasizing the need for continuous cognitive decline monitoring and its early detection. The Mini-Mental State Examination (MMSE) remains a widely used and cost-effective diagnostic tool, with efforts underway to adapt it for digital home-based assessments, enabling more frequent monitoring while minimizing patient burden and mobility. Similarly, electroencephalograms (EEG) have been investigated to monitor cognitive status in ambulatory settings. In this cross-sectional study, we identified key EEG features reflecting the cognitive decline process and assessed their feasibility to estimate cognitive status using machine learning (ML). Method An international and diverse cohort (France, Greece, Turkey, Argentina, Colombia) was gathered comprising N = 510 older adults (40-98 years, 46% male). At the time of stationary EEG recording, subjects exhibited MMSE scores ranging from 30 (cognitively normal) to 4 (severe dementia). A Gradient Boosting ML regressor was developed to estimate their cognitive status based on their EEG spectrum, complexity, and connectivity, focusing on identifying features strongly associated with MMSE scores. The model estimations were evaluated in a leave-one-out cross-validation procedure. Result Key EEG features significantly correlated with MMSE scores included Hjorth Complexity in the left temporal lobe (r=0.58), alpha coherence between the left and right temporal lobes (r=0.48), and beta occipital edge frequency (r=0.42). Eighty combined EEG features were identified as predictors of cognitive status. Using these features, the ML regressor estimated cognitive status with an average error of 2.53 points in the MMSE scale (95% CI±5.36). The model demonstrated strong predictive performance, achieving an R2 value of 0.80 between estimated and actual MMSE scores. Conclusion Specific EEG features, particularly those of temporal and occipital activity, can serve as reliable predictors of cognitive status. While cohort diversity enhanced the generalizability of these findings, more EEG recordings in the low MMSE range are needed to improve regression performance. Longitudinal studies are required to validate the tracking of intra-subject EEG activity changes associated with cognitive decline. In the future, ML could automate periodic monitoring assessments of cognitive health based on EEG in its wearable and low-resolution format, especially in regions with limited specialized staff and imaging technology.

2024

1. Mode of delivery predicts postpartum maternal leukocyte telomere length

   Danielle M. Panelli, Jonathan A. Mayo, Ronald J. Wong, and 11 more authors

   European Journal of Obstetrics & Gynecology and Reproductive Biology, Sep 2024

   Abs DOI

   Background Recent studies have suggested that pregnancy accelerates biologic aging, yet little is known about how biomarkers of aging are affected by events during the peripartum period. Given that immune shifts are known to occur following surgery, we explored the relation between mode of delivery and postpartum maternal leukocyte telomere length (LTL), a marker of biologic aging. Study design Postpartum maternal blood samples were obtained from a prospective cohort of term, singleton livebirths without hypertensive disorders or peripartum infections between 2012 and 2018. The primary outcome was postpartum LTLs from one blood sample drawn between postpartum week 1 and up to 6 months postpartum, measured from thawed frozen peripheral blood mononuclear cells using quantitative PCR in basepairs (bp). Multivariable linear regression models compared LTLs between vaginal versus cesarean births, adjusting for age, body mass index, and nulliparity as potential confounders. Analyses were conducted in two mutually exclusive groups: those with LTL measured postpartum week 1 and those measured up to 6 months postpartum. Secondarily, we compared multiomics by mode of delivery using machine-learning methods to evaluate whether other biologic changes occurred following cesarean. These included transcriptomics, metabolomics, microbiomics, immunomics, and proteomics (serum and plasma). Results Of 67 included people, 50 (74.6 %) had vaginal and 17 (25.4 %) had cesarean births. LTLs were significantly shorter after cesarean in postpartum week 1 (5755.2 bp cesarean versus 6267.8 bp vaginal, p = 0.01) as well as in the later draws (5586.6 versus 5945.6 bp, p = 0.04). After adjusting for confounders, these differences persisted in both week 1 (adjusted beta −496.1, 95 % confidence interval [CI] −891.1, −101.1, p = 0.01) and beyond (adjusted beta −396.8; 95 % CI −727.2, −66.4. p = 0.02). Among the 15 participants who also had complete postpartum multiomics data available, there were predictive signatures of vaginal versus cesarean births in transcriptomics (cell-free [cf]RNA), metabolomics, microbiomics, and proteomics that did not persist after false discovery correction. Conclusion Maternal LTLs in postpartum week 1 were nearly 500 bp shorter following cesarean. This difference persisted several weeks postpartum, even though other markers of inflammation had normalized. Mode of delivery should be considered in any analyses of postpartum LTLs and further investigation into this phenomenon is warranted.

2. Association of pregnancy complications and postpartum maternal leukocyte telomeres in two diverse cohorts: a nested case-control study

   Danielle M. Panelli, Xiaobin Wang, Jonathan Mayo, and 9 more authors

   BMC Pregnancy and Childbirth, Jul 2024

   Abs DOI

   Biologic strain such as oxidative stress has been associated with short leukocyte telomere length (LTL), as well as with preeclampsia and spontaneous preterm birth, yet little is known about their relationships with each other. We investigated associations of postpartum maternal LTL with preeclampsia and spontaneous preterm birth.

3. Rotor Thrust Estimation Using RNN in TensorFlow - a Preliminary for Wind Turbine Control

   Jeffrey Stegink, Andreas Klein, Julia Kersten, and 5 more authors

   In Proceedings of the seventh international conference on multibody system dynamics, Jun 2024
4. Computational Approaches for Connecting Maternal Stress to Preterm Birth

   Amin Mirzaei, Bjarne C. Hiller, Ina A. Stelzer, and 3 more authors

   Clinics in Perinatology, Jun 2024

   DOI
5. CompTrails: comparing hypotheses across behavioral networks

   Tobias Koopmann, Martin Becker, Florian Lemmerich, and 1 more author

   Data Mining and Knowledge Discovery, May 2024

   Abs DOI

   The term Behavioral Networks describes networks that contain relational information on human behavior. This ranges from social networks that contain friendships or cooperations between individuals, to navigational networks that contain geographical or web navigation, and many more. Understanding the forces driving behavior within these networks can be beneficial to improving the underlying network, for example, by generating new hyperlinks on websites, or by proposing new connections and friends on social networks. Previous approaches considered different hypotheses on a single network and evaluated which hypothesis fits best. These hypotheses can represent human intuition and expert opinions or be based on previous insights. In this work, we extend these approaches to enable the comparison of a single hypothesis between multiple networks. We unveil several issues of naive approaches that potentially impact comparisons and lead to undesired results. Based on these findings, we propose a framework with five flexible components that allow addressing specific analysis goals tailored to the application scenario. We show the benefits and limits of our approach by applying it to synthetic data and several real-world datasets, including web navigation, bibliometric navigation, and geographic navigation. Our work supports practitioners and researchers with the aim of understanding similarities and differences in human behavior between environments.

6. Multimodal machine learning for modeling infant head circumference, mothers’ milk composition, and their shared environment

   Martin Becker, Kelsey Fehr, Stephanie Goguen, and 15 more authors

   Scientific Reports, Feb 2024

   Abs DOI

   Links between human milk (HM) and infant development are poorly understood and often focus on individual HM components. Here we apply multi-modal predictive machine learning to study HM and head circumference (a proxy for brain development) among 1022 mother-infant dyads of the CHILD Cohort. We integrated HM data (19 oligosaccharides, 28 fatty acids, 3 hormones, 28 chemokines) with maternal and infant demographic, health, dietary and home environment data. Head circumference was significantly predictable at 3 and 12 months. Two of the most associated features were HM n3-polyunsaturated fatty acid C22:6n3 (docosahexaenoic acid, DHA; p = 9.6e−05) and maternal intake of fish (p = 4.1e−03), a key dietary source of DHA with established relationships to brain function. Thus, using a systems biology approach, we identified meaningful relationships between HM and brain development, which validates our statistical approach, gives credence to the novel associations we observed, and sets the foundation for further research with additional cohorts and HM analytes.

2023

1. Deep representation learning identifies associations between physical activity and sleep patterns during pregnancy and prematurity

   Neal G. Ravindra, Camilo Espinosa, Eloïse Berson, and 18 more authors

   npj Digital Medicine, Sep 2023

   Abs DOI

   Preterm birth (PTB) is the leading cause of infant mortality globally. Research has focused on developing predictive models for PTB without prioritizing cost-effective interventions. Physical activity and sleep present unique opportunities for interventions in low- and middle-income populations (LMICs). However, objective measurement of physical activity and sleep remains challenging and self-reported metrics suffer from low-resolution and accuracy. In this study, we use physical activity data collected using a wearable device comprising over 181,944 h of data across N = 1083 patients. Using a new state-of-the art deep learning time-series classification architecture, we develop a ‘clock’ of healthy dynamics during pregnancy by using gestational age (GA) as a surrogate for progression of pregnancy. We also develop novel interpretability algorithms that integrate unsupervised clustering, model error analysis, feature attribution, and automated actigraphy analysis, allowing for model interpretation with respect to sleep, activity, and clinical variables. Our model performs significantly better than 7 other machine learning and AI methods for modeling the progression of pregnancy. We found that deviations from a normal ‘clock’ of physical activity and sleep changes during pregnancy are strongly associated with pregnancy outcomes. When our model underestimates GA, there are 0.52 fewer preterm births than expected (P = 1.01e − 67, permutation test) and when our model overestimates GA, there are 1.44 times (P = 2.82e − 39, permutation test) more preterm births than expected. Model error is negatively correlated with interdaily stability (P = 0.043, Spearman’s), indicating that our model assigns a more advanced GA when an individual’s daily rhythms are less precise. Supporting this, our model attributes higher importance to sleep periods in predicting higher-than-actual GA, relative to lower-than-actual GA (P = 1.01e − 21, Mann-Whitney U). Combining prediction and interpretability allows us to signal when activity behaviors alter the likelihood of preterm birth and advocates for the development of clinical decision support through passive monitoring and exercise habit and sleep recommendations, which can be easily implemented in LMICs.

2. Pregnancy-induced memory CD4+ regulatory T cells: identification of specific immune memory markers using Infinity Flow

   K. Thiele, C. Urbschat, S. Schepanski, and 4 more authors

   Journal of Reproductive Immunology, Sep 2023

   Abs DOI

   Pregnancy success is critically dependent on specific maternal immune and endocrine adaptions to induce immune tolerance towards the allogenic fetus. Hereby, second pregnancies have been observed to be less prone to obstetric complications and the acquisition of immune memory has been proposed. We could recently show that CD4+ regulatory T (Treg) cells are crucially involved in ameliorating the course of subsequent pregnancies. However, the lack of specific immune markers to reliably detect memory CD4+ Treg cells still aggravates the investigation of immune memory during pregnancy. Therefore, we employed a sophisticated screening approach combining high throughput flow cytometry and machine learning, a method called Infinity Flow. Cells from the uterus-draining lymph nodes were harvested within 7 days postpartum from parous mice allogenically mated twice and virgin control mice. Confirming previous results using conventional flow cytometry, the proportion of the CD4+ Treg cell cluster from parous mice was visibly larger compared to virgin mice. Within the CD4+ Treg cluster, we further observed the highest fold change between virgin and parous samples for Integrin-b7 followed by CD73 and Ly6C. Among the Top10 marker, we could also detect the memory marker CD44 as well as the folate receptor (FR)4 and CD137. We could independently verify higher gene expression levels for FR4 and CD73 in CD4+ Treg cells sorted from parous mice compared to virgin mice. Our results indicate CD73 and FR4 as novel immune memory markers to allow the detection of pregnancy-generated CD4+ Treg cells. A thorough validation in an in-vivo setting remains necessary.

3. Differential cellular immune response in pregnant and non-pregnant women upon vaccination against SARS-CoV-2 infection

   Christopher Urbschat, Kristin Thiele, Ina Stelzer, and 10 more authors

   Journal of Reproductive Immunology, Sep 2023

   Abs DOI

   During pregnancy, a fine-tuned adaptation of the maternal immune system is necessary to promote immune tolerance towards the allogeneic fetus, while simultaneously protecting mother and child from pathogens. Since pregnant women are known to be at high risk for severe courses of infections such as Influenza and SARS-Cov2, vaccination during pregnancy is recommended. However, as a highly vulnerable subpopulation, pregnant women are often excluded from clinical trials. Consequently, information regarding vaccination response and efficiency during pregnancy is still lacking. We recruited pregnant (n=31, 2nd trimester) and non-pregnant (n=29) SARS-Cov2-naïve women who were vaccinated with Comirnaty (BioNTech) twice at an interval of six weeks. Two to six weeks post-vaccination, blood samples were taken and PBMCs isolated. To investigate the effect of pregnancy on T and B cells in response to vaccination, we used scRNA-sequencing to identify differentially expressed genes, represented by distinct transcriptomes, as well as T and B cell receptor sequencing from women in both groups, to map clonal diversity of T and B cells. In addition to sequencing, T cells from both groups were stimulated with the SARS-CoV-2 specific spike protein and analyzed by cytometry by time-of-flight to assess the cytokine response and proliferation capacity of T cells upon potential reinfection. In this study, we used state-of-the-art technologies to gain a deeper understanding of vaccination responses of pregnant women optimizing vaccination doses during pregnancy to maximize the safety of and beneficial effects for fetus and mother.

4. Cross-species comparative analysis of single presynapses

   Eloïse Berson, Chandresh R. Gajera, Thanaphong Phongpreecha, and 18 more authors

   Scientific Reports, Aug 2023

   Abs DOI

   Comparing brain structure across species and regions enables key functional insights. Leveraging publicly available data from a novel mass cytometry-based method, synaptometry by time of flight (SynTOF), we applied an unsupervised machine learning approach to conduct a comparative study of presynapse molecular abundance across three species and three brain regions. We used neural networks and their attractive properties to model complex relationships among high dimensional data to develop a unified, unsupervised framework for comparing the profile of more than 4.5 million single presynapses among normal human, macaque, and mouse samples. An extensive validation showed the feasibility of performing cross-species comparison using SynTOF profiling. Integrative analysis of the abundance of 20 presynaptic proteins revealed near-complete separation between primates and mice involving synaptic pruning, cellular energy, lipid metabolism, and neurotransmission. In addition, our analysis revealed a strong overlap between the presynaptic composition of human and macaque in the cerebral cortex and neostriatum. Our unique approach illuminates species- and region-specific variation in presynapse molecular composition.

5. Whole genome deconvolution unveils Alzheimer’s resilient epigenetic signature

   Eloise Berson, Anjali Sreenivas, Thanaphong Phongpreecha, and 16 more authors

   Nature Communications, Aug 2023

   Abs DOI

   Assay for Transposase Accessible Chromatin by sequencing (ATAC-seq) accurately depicts the chromatin regulatory state and altered mechanisms guiding gene expression in disease. However, bulk sequencing entangles information from different cell types and obscures cellular heterogeneity. To address this, we developed Cellformer, a deep learning method that deconvolutes bulk ATAC-seq into cell type-specific expression across the whole genome. Cellformer enables cost-effective cell type-specific open chromatin profiling in large cohorts. Applied to 191 bulk samples from 3 brain regions, Cellformer identifies cell type-specific gene regulatory mechanisms involved in resilience to Alzheimer’s disease, an uncommon group of cognitively healthy individuals that harbor a high pathological load of Alzheimer’s disease. Cell type-resolved chromatin profiling unveils cell type-specific pathways and nominates potential epigenetic mediators underlying resilience that may illuminate therapeutic opportunities to limit the cognitive impact of the disease. Cellformer is freely available to facilitate future investigations using high-throughput bulk ATAC-seq data.

6. Multiomic signals associated with maternal epidemiological factors contributing to preterm birth in low- and middle-income countries

   Camilo A. Espinosa, Waqasuddin Khan, Rasheda Khanam, and 54 more authors

   Science Advances, May 2023

   Abs DOI

   Preterm birth (PTB) is the leading cause of death in children under five, yet comprehensive studies are hindered by its multiple complex etiologies. Epidemiological associations between PTB and maternal characteristics have been previously described. This work used multiomic profiling and multivariate modeling to investigate the biological signatures of these characteristics. Maternal covariates were collected during pregnancy from 13,841 pregnant women across five sites. Plasma samples from 231 participants were analyzed to generate proteomic, metabolomic, and lipidomic datasets. Machine learning models showed robust performance for the prediction of PTB (AUROC = 0.70), time-to-delivery (r = 0.65), maternal age (r = 0.59), gravidity (r = 0.56), and BMI (r = 0.81). Time-to-delivery biological correlates included fetal-associated proteins (e.g., ALPP, AFP, and PGF) and immune proteins (e.g., PD-L1, CCL28, and LIFR). Maternal age negatively correlated with collagen COL9A1, gravidity with endothelial NOS and inflammatory chemokine CXCL13, and BMI with leptin and structural protein FABP4. These results provide an integrated view of epidemiological factors associated with PTB and identify biological signatures of clinical covariates affecting this disease.

7. Evidence for human milk as a biological system and recommendations for study design—a report from “Breastmilk Ecology: Genesis of Infant Nutrition (BEGIN)” Working Group 4

   Sharon M. Donovan, Nima Aghaeepour, Aline Andres, and 9 more authors

   The American Journal of Clinical Nutrition, May 2023

   Abs DOI

   Human milk contains all of the essential nutrients required by the infant within a complex matrix that enhances the bioavailability of many of those nutrients. In addition, human milk is a source of bioactive components, living cells and microbes that facilitate the transition to life outside the womb. Our ability to fully appreciate the importance of this matrix relies on the recognition of short- and long-term health benefits and, as highlighted in previous sections of this supplement, its ecology (i.e., interactions among the lactating parent and breastfed infant as well as within the context of the human milk matrix itself). Designing and interpreting studies to address this complexity depends on the availability of new tools and technologies that account for such complexity. Past efforts have often compared human milk to infant formula, which has provided some insight into the bioactivity of human milk, as a whole, or of individual milk components supplemented with formula. However, this experimental approach cannot capture the contributions of the individual components to the human milk ecology, the interaction between these components within the human milk matrix, or the significance of the matrix itself to enhance human milk bioactivity on outcomes of interest. This paper presents approaches to explore human milk as a biological system and the functional implications of that system and its components. Specifically, we discuss study design and data collection considerations and how emerging analytical technologies, bioinformatics, and systems biology approaches could be applied to advance our understanding of this critical aspect of human biology.

8. Large-scale correlation network construction for unraveling the coordination of complex biological systems

   Martin Becker, Huda Nassar, Camilo Espinosa, and 27 more authors

   Nature Computational Science, Apr 2023

   Abs DOI

   Advanced measurement and data storage technologies have enabled high-dimensional profiling of complex biological systems. For this, modern multiomics studies regularly produce datasets with hundreds of thousands of measurements per sample, enabling a new era of precision medicine. Correlation analysis is an important first step to gain deeper insights into the coordination and underlying processes of such complex systems. However, the construction of large correlation networks in modern high-dimensional datasets remains a major computational challenge owing to rapidly growing runtime and memory requirements. Here we address this challenge by introducing CorALS (Correlation Analysis of Large-scale (biological) Systems), an open-source framework for the construction and analysis of large-scale parametric as well as non-parametric correlation networks for high-dimensional biological data. It features off-the-shelf algorithms suitable for both personal and high-performance computers, enabling workflows and downstream analysis approaches. We illustrate the broad scope and potential of CorALS by exploring perspectives on complex biological processes in large-scale multiomics and single-cell studies.

9. Data-driven longitudinal characterization of neonatal health and morbidity

   Davide De Francesco, Jonathan D. Reiss, Jacquelyn Roger, and 30 more authors

   Science Translational Medicine, Feb 2023

   Abs DOI

   Although prematurity is the single largest cause of death in children under 5 years of age, the current definition of prematurity, based on gestational age, lacks the precision needed for guiding care decisions. Here, we propose a longitudinal risk assessment for adverse neonatal outcomes in newborns based on a deep learning model that uses electronic health records (EHRs) to predict a wide range of outcomes over a period starting shortly before conception and ending months after birth. By linking the EHRs of the Lucile Packard Children’s Hospital and the Stanford Healthcare Adult Hospital, we developed a cohort of 22,104 mother-newborn dyads delivered between 2014 and 2018. Maternal and newborn EHRs were extracted and used to train a multi-input multitask deep learning model, featuring a long short-term memory neural network, to predict 24 different neonatal outcomes. An additional cohort of 10,250 mother-newborn dyads delivered at the same Stanford Hospitals from 2019 to September 2020 was used to validate the model. Areas under the receiver operating characteristic curve at delivery exceeded 0.9 for 10 of the 24 neonatal outcomes considered and were between 0.8 and 0.9 for 7 additional outcomes. Moreover, comprehensive association analysis identified multiple known associations between various maternal and neonatal features and specific neonatal outcomes. This study used linked EHRs from more than 30,000 mother-newborn dyads and would serve as a resource for the investigation and prediction of neonatal outcomes. An interactive website is available for independent investigators to leverage this unique dataset: https://maternal-child-health-associations.shinyapps.io/shiny_app/.

10. Shorter maternal leukocyte telomere length following cesarean birth: Implications for future research

    Danielle M. Panelli, Jonathan A. Mayo, Ronald J. Wong, and 9 more authors

    American Journal of Obstetrics & Gynecology, Jan 2023

    DOI
11. Deleterious and Protective Psychosocial and Stress-Related Factors Predict Risk of Spontaneous Preterm Birth

    Martin Becker, Jonathan A. Mayo, Nisha K. Phogat, and 10 more authors

    American Journal of Perinatology, Jan 2023

    Abs DOI

    Objectives The aim of the study was to: (1) Identify (early in pregnancy) psychosocial and stress-related factors that predict risk of spontaneous preterm birth (PTB, gestational age Study Design This is a prospective cohort study. Relative risk (RR) analyses investigated association between individual factors and PTB. Machine learning-based interdependency analysis (IDPA) identified factor clusters, strength, and direction of association with PTB. A nonlinear model based on support vector machines was built for predicting PTB and identifying factors that most strongly predicted PTB. Results Higher levels of deleterious factors were associated with increased RR for PTB: General anxiety (RR = 8.9; 95% confidence interval [CI] = 2.0,39.6), pain (RR = 5.7; CI = 1.7,17.0); tiredness/fatigue (RR = 3.7; CI = 1.09,13.5); perceived risk of birth complications (RR = 4; CI = 1.6,10.01); self-rated health current (RR = 2.6; CI = 1.0,6.7) and previous 3 years (RR = 2.9; CI = 1.1,7.7); and divorce (RR = 2.9; CI = 1.1,7.8). Lower levels of protective factors were also associated with increased RR for PTB: low happiness (RR = 9.1; CI = 1.25,71.5); low support from parents/siblings (RR = 3.5; CI = 0.9,12.9), and father-of-baby (RR = 3; CI = 1.1,9.9). These factors were also components of the clusters identified by the IDPA: perceived risk of birth complications (p p Conclusion These findings represent an important step toward identifying key factors, which can be assessed rapidly before/after conception, to predict risk of PTB, and perhaps other adverse pregnancy outcomes. Quantifying these factors, before, or early in pregnancy, could identify women at risk of delivering preterm, pinpoint mechanisms/targets for intervention, and facilitate the development of interventions to prevent PTB. Key Points

12. Prediction of neuropathologic lesions from clinical data

    Thanaphong Phongpreecha, Brenna Cholerton, Syed Bukhari, and 21 more authors

    Alzheimer’s & Dementia, 2023

    Abs DOI

    INTRODUCTION Post-mortem analysis provides definitive diagnoses of neurodegenerative diseases; however, only a few can be diagnosed during life. METHODS This study employed statistical tools and machine learning to predict 17 neuropathologic lesions from a cohort of 6518 individuals using 381 clinical features (Table S1). The multisite data allowed validation of the model’s robustness by splitting train/test sets by clinical sites. A similar study was performed for predicting Alzheimer’s disease (AD) neuropathologic change without specific comorbidities. RESULTS Prediction results show high performance for certain lesions that match or exceed that of research annotation. Neurodegenerative comorbidities in addition to AD neuropathologic change resulted in compounded, but disproportionate, effects across cognitive domains as the comorbidity number increased. DISCUSSION Certain clinical features could be strongly associated with multiple neurodegenerative diseases, others were lesion-specific, and some were divergent between lesions. Our approach could benefit clinical research, and genetic and biomarker research by enriching cohorts for desired lesions.

13. Prediction of multiple neuropathologic changes from features available during life

    Thanaphong Phongpreecha, Davide De Francesco, Alan Chang, and 10 more authors

    Alzheimer’s & Dementia, 2023

    Abs DOI

    Background Neuropathologic changes are central for both understanding of the patients’ brains and making the definitive diagnosis of dementia-related diseases. However, many of them are only obtainable post-mortem. To make this information available while an individual is still alive, this study developed machine learning models that predict neuropathologic changes based on features obtainable during life. Method The multi-site post-mortem data (n∼5000) was obtained from National Alzheimer’s Coordinating Center. A multitask long-short term memory-based neural network architecture was developed with custom loss to predict the 13 neuropathologic changes from features during life measured longitudinally. The performance of the model was evaluated using entire unseen sites as test sets. Evaluation metrics include area under receiver’s operating curve (AUROC) and area under precision recall curve (AUPRC). Result The model was able to predict Alzheimer’s Disease neuropathologic changes (ADNC) and any Alzheimer’s-related pathologies, such as Braak score, with great sensitivity, specificity, and precision (for example AUROC = 0.85; AUPRC = 0.96 for ADNC). Apart from these, the model can also predict hippocampal sclerosis accurately (AUROC = 0.79; AUPRC = 0.80) and Lewy Body disease at higher precision than clinician’s diagnosis. Model interpretation shows patterns in neuropsychological tests that are predictive of pathologic changes. Additionally, model error analysis revealed factors, such as resilient case ratios, that explain variation in performance between sites. Conclusion Patterns of measurable features during life can be used by machine learning to predict ADNC, hippocampal sclerosis, and to lesser extent Lewy Body.

14. In-silico generation of high-dimensional immune response data in patients using a deep neural network

    Ramin Fallahzadeh, Neda H. Bidoki, Ina A. Stelzer, and 13 more authors

    Cytometry Part A, 2023

    Abs DOI

    Technologies for single-cell profiling of the immune system have enabled researchers to extract rich interconnected networks of cellular abundance, phenotypical and functional cellular parameters. These studies can power machine learning approaches to understand the role of the immune system in various diseases. However, the performance of these approaches and the generalizability of the findings have been hindered by limited cohort sizes in translational studies, partially due to logistical demands and costs associated with longitudinal data collection in sufficiently large patient cohorts. An evolving challenge is the requirement for ever-increasing cohort sizes as the dimensionality of datasets grows. We propose a deep learning model derived from a novel pipeline of optimal temporal cell matching and overcomplete autoencoders that uses data from a small subset of patients to learn to forecast an entire patient’s immune response in a high dimensional space from one timepoint to another. In our analysis of 1.08 million cells from patients pre- and post-surgical intervention, we demonstrate that the generated patient-specific data are qualitatively and quantitatively similar to real patient data by demonstrating fidelity, diversity, and usefulness.

2022

1. Revealing the impact of lifestyle stressors on the risk of adverse pregnancy outcomes with multitask machine learning

   Martin Becker, Jennifer Dai, Alan L. Chang, and 29 more authors

   Frontiers in Pediatrics, Dec 2022

   Abs DOI

   Psychosocial and stress-related factors (PSFs), defined as internal or external stimuli that induce biological changes, are potentially modifiable factors and accessible targets for interventions that are associated with adverse pregnancy outcomes (APOs). Although individual APOs haven been shown to be connected to PSFs, they are biologically interconnected, relatively infrequent, and therefore challenging to model. In this context, multi-task machine learning (MML) is an ideal tool for exploring the interconnectedness of APOs and at the same time building on joint combinatorial outcomes to increase predictive power. Additionally, by integrating single-cell immunological profiling of underlying biological processes, the effects of stress-based therapeutics may be measurable, facilitating the development of precision medicine approaches.The primary objectives were to jointly model multiple APOs and their connection to stress early in pregnancy, and to explore the underlying biology to guide development of accessible and measurable interventions.In a prospective cohort study, PSFs were assessed during the first trimester with an extensive self-filled questionnaire for 200 women. We used MML to simultaneously model and predict APOs (severe preeclampsia, superimposed preeclampsia, gestational diabetes and early gestational age) as well as several risk factors (BMI, diabetes, hypertension) for these patients based on PSFs. Strongly interrelated stressors were categorized to identify potential therapeutic targets. Furthermore, for a subset of 14 women, we modeled the connection between PSFs, the maternal immune system, and APOs by building ML models based on an extensive single cell immune dataset (mass cytometry time of flight, i.e., CyTOF).Jointly modeling APOs in a MML setting significantly increased modeling capabilities and yielded a highly predictive integrated model of APOs underscoring their interconnectedness. Most APOs were associated with mental health, life stress, and perceived health risks. Biologically, stressors were associated with immune characteristics revolving around CD4/CD8 T cells. Immune characteristics predicted from stress were in turn found to associate with APOs. Elucidating connections among stress, multiple APOs simultaneously, and immune characteristics has the potential to facilitate the implementation of ML-based, individualized, integrative models of pregnancy in clinical decision making. The modifiable nature of stressors may enable the development of accessible interventions, with success tracked through immune characteristics.

2. Early prediction and longitudinal modeling of preeclampsia from multiomics

   Ivana Marić, Kévin Contrepois, Mira N. Moufarrej, and 30 more authors

   Patterns, Dec 2022

   DOI
3. Leukocyte telomere dynamics across gestation in uncomplicated pregnancies and associations with stress

   Danielle M. Panelli, Stephanie A. Leonard, Ronald J. Wong, and 10 more authors

   BMC Pregnancy and Childbirth, May 2022

   Abs DOI

   Short leukocyte telomere length is a biomarker associated with stress and morbidity in non-pregnant adults. Little is known, however, about maternal telomere dynamics in pregnancy. To address this, we examined changes in maternal leukocyte telomere length (LTL) during uncomplicated pregnancies and explored correlations with perceived stress.

4. A data-driven health index for neonatal morbidities

   Davide De Francesco, Yair J. Blumenfeld, Ivana Marić, and 18 more authors

   iScience, Apr 2022

   DOI
5. Machine Learning Techniques

   Jeff Choi, Nima Aghaeepour, and Martin Becker

   In Statistics and Research Methods for Acute Care and General Surgeons, 2022

   Abs DOI

   Machine learning comprises algorithms that can perform tasks they were not explicitly programmed to perform. Explicitly programmed algorithms perform tasks according to a predefined sequence of instructions. Conversely, machine learning algorithms are programmed to learn to perform tasks using input data. In the era of abundant data, affordable data storage, and computational capabilities, understanding machine learning algorithms is critical to better explore and answer questions that can advance surgical science.

2021

1. Objective Activity Parameters Track Patient-specific Physical Recovery Trajectories After Surgery and Link With Individual Preoperative Immune States

   Ramin Fallahzadeh, Franck Verdonk, Ed Ganio, and 8 more authors

   Annals of Surgery, Oct 2021

   DOI
2. Single-synapse analyses of Alzheimer’s disease implicate pathologic tau, DJ1, CD47, and ApoE

   Thanaphong Phongpreecha, Chandresh R Gajera, Candace C Liu, and 8 more authors

   Science advances, 2021
3. Risk assessment analysis for maternal autoantibody-related autism (MAR-ASD): a subtype of autism

   Alexandra Ramirez-Celis, Martin Becker, Miriam Nuño, and 3 more authors

   Molecular Psychiatry, 2021
4. Redescription Model Mining

   Felix I. Stamm, Martin Becker, Markus Strohmaier, and 1 more author

   In Proceedings of the 27nd ACM SIGKDD international conference on knowledge discovery and data mining, 2021

   DOI
5. Integrated trajectories of the maternal metabolome, proteome, and immunome predict labor onset

   Ina A. Stelzer, Mohammad S. Ghaemi, Xiaoyuan Han, and 35 more authors

   Science Translational Medicine, 2021

   Abs DOI

   Current methods to predict spontaneous labor are fairly inaccurate. To provide better estimates and biomarkers of labor onset, the biological processes that lead up to labor need to be better understood. Stelzer et al. performed metabolome, proteome, and immunome studies on blood samples from 63 women in the 100 days before delivery. They identified a surge in IL-1R4 and steroid hormones in the weeks before delivery, which was coordinated with a switch from immune activation to regulation of inflammatory responses. A model was then constructed to predict time to labor independent of gestational age. These results may be helpful for development of more accurate methods to predict labor.Estimating the time of delivery is of high clinical importance because pre- and postterm deviations are associated with complications for the mother and her offspring. However, current estimations are inaccurate. As pregnancy progresses toward labor, major transitions occur in fetomaternal immune, metabolic, and endocrine systems that culminate in birth. The comprehensive characterization of maternal biology that precedes labor is key to understanding these physiological transitions and identifying predictive biomarkers of delivery. Here, a longitudinal study was conducted in 63 women who went into labor spontaneously. More than 7000 plasma analytes and peripheral immune cell responses were analyzed using untargeted mass spectrometry, aptamer-based proteomic technology, and single-cell mass cytometry in serial blood samples collected during the last 100 days of pregnancy. The high-dimensional dataset was integrated into a multiomic model that predicted the time to spontaneous labor [R = 0.85, 95% confidence interval (CI) [0.79 to 0.89], P = 1.2 × 10-40, N = 53, training set; R = 0.81, 95% CI [0.61 to 0.91], P = 3.9 × 10-7, N = 10, independent test set]. Coordinated alterations in maternal metabolome, proteome, and immunome marked a molecular shift from pregnancy maintenance to prelabor biology 2 to 4 weeks before delivery. A surge in steroid hormone metabolites and interleukin-1 receptor type 4 that preceded labor coincided with a switch from immune activation to regulation of inflammatory responses. Our study lays the groundwork for developing blood-based methods for predicting the day of labor, anchored in mechanisms shared in preterm and term pregnancies.

6. Data-Driven Modeling of Pregnancy-Related Complications

   Camilo Espinosa, Martin Becker, Ivana Maric, and 14 more authors

   Trends in Molecular Medicine, 2021

   Abs DOI

   A healthy pregnancy depends on complex interrelated biological adaptations involving placentation, maternal immune responses, and hormonal homeostasis. Recent advances in high-throughput technologies have provided access to multiomics biological data that, combined with clinical and social data, can provide a deeper understanding of normal and abnormal pregnancies. Integration of these heterogeneous datasets using state-of-the-art machine-learning methods can enable the prediction of short- and long-term health trajectories for a mother and offspring and the development of treatments to prevent or minimize complications. We review advanced machine-learning methods that could: provide deeper biological insights into a pregnancy not yet unveiled by current methodologies; clarify the etiologies and heterogeneity of pathologies that affect a pregnancy; and suggest the best approaches to address disparities in outcomes affecting vulnerable populations.

2020

1. VoPo leverages cellular heterogeneity for predictive modeling of single-cell data

   Natalie Stanley, Ina A. Stelzer, Amy S. Tsai, and 27 more authors

   Nature Communications, 2020

   Abs DOI

   High-throughput single-cell analysis technologies produce an abundance of data that is critical for profiling the heterogeneity of cellular systems. We introduce VoPo (https://github.com/stanleyn/VoPo), a machine learning algorithm for predictive modeling and comprehensive visualization of the heterogeneity captured in large single-cell datasets. In three mass cytometry datasets, with the largest measuring hundreds of millions of cells over hundreds of samples, VoPo defines phenotypically and functionally homogeneous cell populations. VoPo further outperforms state-of-the-art machine learning algorithms in classification tasks, and identified immune-correlates of clinically-relevant parameters.

2. OpenLUR: Off-the-shelf air pollution modeling with open features and machine learning

   Florian Lautenschlager, Martin Becker, Konstantin Kobs, and 4 more authors

   Atmospheric Environment, 2020

   Abs DOI

   To assess the exposure of citizens to pollutants like NOx or particulate matter in urban areas, land use regression (LUR) models are a well established method. LUR models leverage information about environmental and anthropogenic factors such as cars, heating, or industry to predict air pollution in areas where no measurements have been made. However, existing approaches are often not globally applicable and require tedious hyper-parameter tuning to enable high quality predictions. In this work, we tackle these issues by introducing OpenLUR, an off-the-shelf approach for modeling air pollution that (i) works on a set of novel features solely extracted from the globally and openly available data source OpenStreetMap and (ii) is based on state-of-the-art machine learning featuring automated hyper-parameter tuning in order to minimize manual effort. We show that our proposed features are able to outperform their counterparts from local and closed sources, and illustrate how automated hyper parameter tuning can yield competitve results while alleviating the need for expert knowledge in machine learning and manual effort. Importantly, we further demonstrate the potential of the global availability of our features by applying cross-learning across different cities in order to reduce the need for a large amount of training samples. Overall, OpenLUR represents an off-the-shelf approach that facilitates easily reproducible experiments and the development of globally applicable models.

3. Multiomics Characterization of Preterm Birth in Low- and Middle-Income Countries

   Fyezah Jehan, Sunil Sazawal, Abdullah H. Baqui, and 75 more authors

   JAMA network open, 2020

   HTML
4. MapLUR: Exploring a New Paradigm for Estimating Air Pollution Using Deep Learning on Map Images

   Michael Steininger, Konstantin Kobs, Albin Zehe, and 3 more authors

   ACM Transactions on Spatial Algorithms and Systems (TSAS), 2020
5. Integration of mechanistic immunological knowledge into a machine learning pipeline improves predictions

   Anthony Culos, Amy S Tsai, Natalie Stanley, and 8 more authors

   Nature Machine Intelligence, 2020

2018

1. EveryAware Gears: A Tool to visualize and analyze all types of Citizen Science Data

   Florian Lautenschlager, Martin Becker, Michael Steininger, and 1 more author

   In Proceedings of VGI geovisual analytics workshop, colocated with BDVA 2018, Oct 2018
2. pysubgroup: Easy-to-Use Subgroup Discovery in Python

   Florian Lemmerich and Martin Becker

   In Ecml/pkdd (3), 2018

   DOI
3. Understanding Human Navigation using Bayesian Hypothesis Comparison

   Martin Becker

   2018
4. Healing Time Correlates With the Quality of Scaring: Results From a Prospective Randomized Control Donor Site Trial

   Frank Werdin, Mayer Tenenhaus, Martin Becker, and 1 more author

   Dermatologic Surgery, 2018
5. Burrows’ Zeta: Exploring and Evaluating Variants and Parameters

   Christof Schöch, Daniel Schlör, Albin Zehe, and 3 more authors

   In Digital humanities, 2018

   HTML
6. Air Trails – Urban Air Quality Campaign Exploration Patterns

   Martin Becker, Florian Lautenschlager, and Andreas Hotho

   In Proceedings of VGI-ALIVE 2018 workshop, colocated with AGILE 2018, 2018

   HTML
7. Adaptive kNN using expected accuracy for classification of geo-spatial data

   Mark Kibanov, Martin Becker, Juergen Mueller, and 3 more authors

   In Proceedings of the 33rd annual ACM symposium on applied computing, 2018

   Abs DOI

   The k-Nearest Neighbor (kNN) classification approach is conceptually simple - yet widely applied since it often performs well in practical applications. However, using a global constant k does not always provide an optimal solution, e. g., for datasets with an irregular density distribution of data points. This paper proposes an adaptive kNN classifier where k is chosen dynamically for each instance (point) to be classified, such that the expected accuracy of classification is maximized. We define the expected accuracy as the accuracy of a set of structurally similar observations. An arbitrary similarity function can be used to find these observations. We introduce and evaluate different similarity functions. For the evaluation, we use five different classification tasks based on geo-spatial data. Each classification task consists of (tens of) thousands of items. We demonstrate, that the presented expected accuracy measures can be a good estimator for kNN performance, and the proposed adaptive kNN classifier outperforms common kNN and previously introduced adaptive kNN algorithms. Also, we show that the range of considered k can be significantly reduced to speed up the algorithm without negative influence on classification accuracy.

8. A White-Box Model for Detecting Author Nationality by Linguistic Differences in Spanish Novels

   Albin Zehe, Daniel Schlör, Ulrike Henny-Krahmer, and 2 more authors

   In Digital humanities, 2018

2017

1. MixedTrails: Bayesian hypothesis comparison on heterogeneous sequential data

   Martin Becker, Florian Lemmerich, Philipp Singer, and 2 more authors

   Data Mining and Knowledge Discovery, Sep 2017

   Abs DOI

   Sequential traces of user data are frequently observed online and offline, e.g., as sequences of visited websites or as sequences of locations captured by GPS. However, understanding factors explaining the production of sequence data is a challenging task, especially since the data generation is often not homogeneous. For example, navigation behavior might change in different phases of browsing a website or movement behavior may vary between groups of users. In this work, we tackle this task and propose MixedTrails , a Bayesian approach for comparing the plausibility of hypotheses regarding the generative processes of heterogeneous sequence data. Each hypothesis is derived from existing literature, theory, or intuition and represents a belief about transition probabilities between a set of states that can vary between groups of observed transitions. For example, when trying to understand human movement in a city and given some data, a hypothesis assuming tourists to be more likely to move towards points of interests than locals can be shown to be more plausible than a hypothesis assuming the opposite. Our approach incorporates such hypotheses as Bayesian priors in a generative mixed transition Markov chain model, and compares their plausibility utilizing Bayes factors. We discuss analytical and approximate inference methods for calculating the marginal likelihoods for Bayes factors, give guidance on interpreting the results, and illustrate our approach with several experiments on synthetic and empirical data from Wikipedia and Flickr. Thus, this work enables a novel kind of analysis for studying sequential data in many application areas.

2. Towards Sentiment Analysis on German Literature

   Albin Zehe, Martin Becker, Fotis Jannidis, and 1 more author

   In Joint german/austrian conference on artificial intelligence (künstliche intelligenz), 2017
3. Sedentary Behavior among National Elite Rowers during Off-Training—A Pilot Study

   Billy Sperlich, Martin Becker, Andreas Hotho, and 5 more authors

   Frontiers in Physiology, 2017

   Abs DOI

   The aim of this pilot study was to analyze the off-training physical activity (PA) profile in national elite German U23 rowers during 31 days of their preparation period. The hours spent in each PA category (i.e. sedentary: \textless1.5 MET; light physical activity: 1.5–3 MET; moderate physical activity: 3–6 MET and vigorous intense physical activity: \textgreater6 MET) were calculated for every valid day (i.e. \textgreater 480 min of wear time). The off-training PA during 21 weekdays and 10 weekend days of the final 11-wk preparation period was assessed by a wrist-worn multisensory device (Microsoft Band II (MSBII)). A total of 11 rowers provided valid data (i.e. \textgreater 480 min/day) for 11.6 week days and 4.8 weekend days during the 31 days observation period. The average sedentary time was 11.63±1.25 hours per day during the week and 12.49±1.10 hours per day on the weekend, with a tendency to be higher on the weekend compared to weekdays (p = 0.06; d = 0.73). The average time in light, moderate and vigorous PA during the weekdays was 1.27±1.15, 0.76±0.37, 0.51±0.44 hours per day and 0.67±0.43, 0.59±0.37, 0.53±0.32 hours per weekend day. Light physical activity was higher during weekdays compared to the weekend (p = 0.04; d = 0.69) Based on our pilot study of eleven national elite rowers we conclude that rowers display a considerable sedentary off-training behavior of more than 11.5 hours/day.

4. Learning Semantic Relatedness from Human Feedback Using Relative Relatedness Learning

   Thomas Niebler, Martin Becker, Christian Pölitz, and 1 more author

   In Proceedings of the ISWC 2017 posters & demonstrations and industry tracks co-located with 16th international semantic web conference (ISWC 2017), 2017

   HTML
5. Experimental Assessment of the Emergence of Awareness and Its Influence on Behavioral Changes: The Everyaware Lesson

   Pietro Gravino, Alina Sirbu, Martin Becker, and 2 more authors

   In Participatory sensing, opinions and collective awareness, 2017

   DOI
6. Eleven-Week Preparation Involving Polarized Intensity Distribution Is Not Superior to Pyramidal Distribution in National Elite Rowers

   Gunnar Treff, Kay Winkert, Mahdi Sareban, and 3 more authors

   Frontiers in Physiology, 2017

   Abs DOI

   Polarized (POL) training intensity distribution (TID) emphasizes high-volume low-intensity exercise in zone (Z)1 (\textless first lactate threshold) with a greater proportion of high-intensity Z3 (\textgreater second lactate threshold) compared to Z2 (between first and second lactate threshold). In highly trained rowers there is a lack of prospective controlled evidence whether POL is superior to pyramidal (PYR; i.e. greater volume in Z1 vs. Z2 vs. Z3) TID. The aim of the study was to compare the effect of POL vs. PYR TID in rowers during an 11-wk preparation period. Fourteen national elite male rowers participated (age: 20 ± 2 years, maximal oxygen uptake (⩒O2max): 66±5 mL/min/kg). The sample was split into PYR and POL by varying the percentage spent in Z2 and Z3 while Z1 was clamped to 93% and matched for total and rowing volume. Actual TIDs were based on time within heart rate zones (Z1 and Z2) and duration of Z3-intervals. The main outcome variables were average power in 2000 m ergometer-test (P2000m), power associated with 4 mmol/L [blood lactate] (P4[BLa]), and ⩒O2max. To quantify the level of polarization, we calculated a Polarization-Index as log (%Z1 x %Z3/%Z2). PYR and POL did not significantly differ regarding rowing or total volume, but POL had a higher percentage of Z3 intensities (6±3% vs. 2±1%; p \textless .005) while Z2 was lower (1±1% vs. 3±2%; p \textless .05) and Z1 was similar (94±3% vs. 93±2%, p = .37). Consequently, Polarization-Index was significantly higher in POL (3.0±0.7 a.u. vs. 1.9±0.4 a.u.; p \textless .01) P2000m did not significantly change with PYR (1.5±1.7%, p = .06) nor POL (1.5±2.6%, p = .26). ⩒O2max did not change (1.7±5.6%, p = .52 or 0.6±2.6, p = .67) and a small increase in P4[BLa] was observed in PYR only (1.9±4.8%, p = .37 or -0.5±4.1%, p = .77). Changes from pre to post were not significantly different between groups in any performance measure. POL did not prove to be superior to PYR, possibly due to the high and very similar percentage of Z1 in this study.

7. Comparing Hypotheses About Sequential Data: A Bayesian Approach and Its Applications

   Florian Lemmerich, Philipp Singer, Martin Becker, and 5 more authors

   In Joint european conference on machine learning and knowledge discovery in databases, 2017
8. Collective Sensing Platforms

   Martin Atzmueller, Martin Becker, and Juergen Mueller

   In Participatory sensing, opinions and collective awareness, 2017

   DOI
9. Applications for Environmental Sensing in EveryAware

   Martin Atzmueller, Martin Becker, Andrea Molino, and 3 more authors

   In Participatory sensing, opinions and collective awareness, 2017

   DOI

2016

1. Prediction of Happy Endings in German Novels

   Albin Zehe, Martin Becker, Lena Hettinger, and 3 more authors

   In Proceedings of the workshop on interactions between data mining and natural language processing 2016, Jul 2016

   Abs HTML

   Identifying plot structure in novels is a valuable step towards automatic processing of literary corpora. We present an approach to classify novels as either having a happy ending or not. To achieve this, we use features based on different sentiment lexica as input for an SVM- classifier, which yields an average F1-score of about 73%.

2. Extracting Semantics from Unconstrained Navigation on Wikipedia

   Thomas Niebler, Daniel Schlör, Martin Becker, and 1 more author

   KI - Künstliche Intelligenz, Jun 2016

   DOI
3. SparkTrails: A MapReduce Implementation of HypTrails for Comparing Hypotheses About Human Trails

   Martin Becker, Hauke Mewes, Andreas Hotho, and 3 more authors

   In International conference companion on world wide web, 2016

   DOI
4. Mining Subgroups with Exceptional Transition Behavior

   Florian Lemmerich, Martin Becker, Philipp Singer, and 3 more authors

   In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, 2016

   Abs DOI

   We present a new method for detecting interpretable subgroups with exceptional transition behavior in sequential data. Identifying such patterns has many potential applications, e.g., for studying human mobility or analyzing the behavior of internet users. To tackle this task, we employ exceptional model mining, which is a general approach for identifying interpretable data subsets that exhibit unusual interactions between a set of target attributes with respect to a certain model class. Although exceptional model mining provides a well-suited framework for our problem, previously investigated model classes cannot capture transition behavior. To that end, we introduce first-order Markov chains as a novel model class for exceptional model mining and present a new interestingness measure that quantifies the exceptionality of transition subgroups. The measure compares the distance between the Markov transition matrix of a subgroup and the respective matrix of the entire data with the distance of random dataset samples. In addition, our method can be adapted to find subgroups that match or contradict given transition hypotheses. We demonstrate that our method is consistently able to recover subgroups with exceptional transition models from synthetic data and illustrate its potential in two application examples. Our work is relevant for researchers and practitioners interested in detecting exceptional transition behavior in sequential data.

5. FolkTrails: Interpreting Navigation Behavior in a Social Tagging System

   Thomas Niebler, Martin Becker, Daniel Zoller, and 2 more authors

   In International on conference on information and knowledge management, 2016

   Abs DOI

   Social tagging systems have established themselves as a quick and easy way to organize information by annotating resources with tags. In recent work, user behavior in social tagging systems was studied, that is, how users assign tags, and consume content. However, it is still unclear how users make use of the navigation options they are given. Understanding their behavior and differences in behavior of different user groups is an important step towards assessing the effectiveness of a navigational concept and of improving it to better suit the users’ needs. In this work, we investigate navigation trails in the popular scholarly social tagging system BibSonomy from six years of log data. We discuss dynamic browsing behavior of the general user population and show that different navigational subgroups exhibit different navigational traits. Furthermore, we provide strong evidence that the semantic nature of the underlying folksonomy is an essential factor for explaining navigation.

2015

1. MicroTrails: comparing hypotheses about task selection on a crowdsourcing platform

   Martin Becker, Kathrin Borchert, Matthias Hirth, and 3 more authors

   In International conference on knowledge technologies and data-driven business, Oct 2015

   HTML
2. Genre Classification on German Novels

   Lena Hettinger, Martin Becker, Isabella Reger, and 2 more authors

   In 26th international workshop on database and expert systems applications (DEXA), Sep 2015

   DOI
3. Participatory Patterns in an International Air Quality Monitoring Initiative

   Alina Sirbu, Martin Becker, Saverio Caminiti, and 17 more authors

   PLoS ONE, Aug 2015

   Abs HTML

   The issue of sustainability is at the top of the political and societal agenda, being considered of extreme importance and urgency. Human individual action impacts the environment both locally (e.g., local air/water quality, noise disturbance) and globally (e.g., climate change, resource use). Urban environments represent a crucial example, with an increasing realization that the most effective way of producing a change is involving the citizens themselves in monitoring campaigns (a citizen science bottom-up approach). This is possible by developing novel technologies and IT infrastructures enabling large citizen participation. Here, in the wider framework of one of the first such projects, we show results from an international competition where citizens were involved in mobile air pollution monitoring using low cost sensing devices, combined with a web-based game to monitor perceived levels of pollution. Measures of shift in perceptions over the course of the campaign are provided, together with insights into participatory patterns emerging from this study. Interesting effects related to inertia and to direct involvement in measurement activities rather than indirect information exposure are also highlighted, indicating that direct involvement can enhance learning and environmental awareness. In the future, this could result in better adoption of policies towards decreasing pollution.

4. VizTrails: An Information Visualization Tool for Exploring Geographic Movement Trajectories

   Martin Becker, Philipp Singer, Florian Lemmerich, and 3 more authors

   In Conference on hypertext & social media, 2015

   DOI
5. Photowalking the City: Comparing Hypotheses About Urban Photo Trails on Flickr

   Martin Becker, Philipp Singer, Florian Lemmerich, and 3 more authors

   In International conference on social informatics, 2015

   DOI
6. Exploratory Subgroup Analytics on Ubiquitous Data

   Martin Atzmueller, Juergen Mueller, and Martin Becker

   In Mining, modeling, and recommending ’things’ in social media, 2015

   Abs

   This paper presents exploratory subgroup analytics on ubiquitous data: We propose subgroup discovery and assessment approaches for obtaining interesting descriptive patterns and provide a novel graph-based analysis approach for assessing the relations between the obtained subgroup set. This exploratory visualization approaches allows for the comparison of subgroups according to their relations to other subgroups and to include further parameters, e.g., geo-spatial distribution indicators. We present and discuss analysis results utilizing real-world data given by geo-tagged noise measurements with associated subjective perceptions and a set of tags describing the semantic context.

7. ConDist: A Context-Driven Categorical Distance Measure

   Markus Ring, Florian Otto, Martin Becker, and 3 more authors

   In Proceedings of the european conference on machine learning and knowledge discovery in databases (ECML PKDD), 2015

   Abs DOI

   A distance measure between objects is a key requirement for many data mining tasks like clustering, classification or outlier detection. However, for objects characterized by categorical attributes, defining meaningful distance measures is a challenging task since the values within such attributes have no inherent order, especially without additional domain knowledge. In this paper, we propose an unsupervised distance measure for objects with categorical attributes based on the idea that categorical attribute values are similar if they appear with similar value distributions on correlated context attributes. Thus, the distance measure is automatically derived from the given data set. We compare our new distance measure to existing categorical distance measures and evaluate on different data sets from the UCI machine-learning repository. The experiments show that our distance measure is recommendable, since it achieves similar or better results in a more robust way than previous approaches.

2014

1. Ubicon and its applications for ubiquitous social computing

   Martin Atzmueller, Martin Becker, Mark Kibanov, and 7 more authors

   New Review of Hypermedia and Multimedia, 2014

   Abs DOI

   The combination of ubiquitous and social computing is an emerging research area which integrates different but complementary methods, techniques and tools. In this paper, we focus on the Ubicon platform, its applications, and a large spectrum of analysis results. Ubicon provides an extensible framework for building and hosting applications targeting both ubiquitous and social environments. We summarize the architecture and exemplify its implementation using four real-world applications built on top of Ubicon. In addition, we discuss several scientific experiments in the context of these applications in order to give a better picture of the potential of the framework, and discuss analysis results using several real-world data sets collected utilizing Ubicon.

2. Subjective vs. Objective Data: Bridging the Gap

   Martin Becker, Andreas Hotho, Juergen Mueller, and 3 more authors

   2014

   Abs HTML

   Sensor data is objective. But when measuring our environment, measured values are contrasted with our perception, which is always subjective. This makes interpreting sensor measurements difficult for a single person in her personal environment. In this context, the EveryAware projects directly connects the concepts of objective sensor data with subjective impressions and perceptions by providing a collective sensing platform with several client applications allowing to explicitly associate those two data types. The goal is to provide the user with personalized feedback, a characterization of the global as well as her personal environment, and enable her to position her perceptions in this global context. In this poster we summarize the collected data of two EveryAware applications, namely WideNoise for noise measurements and AirProbe for participatory air quality sensing. Basic insights are presented including user activity, learning processes and sensor data to perception correlations. These results provide an outlook on how this data can further be used to understand the connection between sensor data and perceptions.

2013

1. Awareness and Learning in Participatory Noise Sensing

   Martin Becker, Saverio Caminiti, Donato Fiorella, and 10 more authors

   PLOS ONE, Dec 2013

   Abs DOI

   The development of ICT infrastructures has facilitated the emergence of new paradigms for looking at society and the environment over the last few years. Participatory environmental sensing, i.e. directly involving citizens in environmental monitoring, is one example, which is hoped to encourage learning and enhance awareness of environmental issues. In this paper, an analysis of the behaviour of individuals involved in noise sensing is presented. Citizens have been involved in noise measuring activities through the WideNoise smartphone application. This application has been designed to record both objective (noise samples) and subjective (opinions, feelings) data. The application has been open to be used freely by anyone and has been widely employed worldwide. In addition, several test cases have been organised in European countries. Based on the information submitted by users, an analysis of emerging awareness and learning is performed. The data show that changes in the way the environment is perceived after repeated usage of the application do appear. Specifically, users learn how to recognise different noise levels they are exposed to. Additionally, the subjective data collected indicate an increased user involvement in time and a categorisation effect between pleasant and less pleasant environments.

2. Tag Recommendations for SensorFolkSonomies

   Juergen Mueller, Stephan Doerfel, Martin Becker, and 2 more authors

   In Proceedings of the fifth ACM RecSys workshop on recommender systems and the social web co-located with the 7th ACM conference on recommender systems (RecSys 2013), 2013

   Abs HTML

   With the rising popularity of smart mobile devices, sensor data-based applications have become more and more popular. Their users record data during their daily routine or specifically for certain events. The application WideNoise Plus allows users to record sound samples and to annotate them with perceptions and tags. The app is being used to document and map the soundscape all over the world. The procedure of recording, including the assignment of tags, has to be as easy-to-use as possible. We therefore discuss the application of tag recommender algorithms in this particular scenario. We show, that this task is fundamentally different from the well-known tag recommendation problem in folksonomies as users do no longer tag fix resources but rather sensory data and impressions. The scenario requires efficient recommender algorithms that are able to run on the mobile device, since Internet connectivity cannot be assumed to be available. Therefore, we evaluate the performance of several tag recommendation algorithms and discuss their applicability in the mobile sensing use-case.

3. Difference-Based Estimates for Generalization-Aware Subgroup Discovery

   Florian Lemmerich, Martin Becker, and Frank Puppe

   In Proceedings of the european conference on machine learning and knowledge discovery in databases (ECML PKDD), 2013

   DOI
4. A generic platform for ubiquitous and subjective data

   Martin Becker, Juergen Mueller, Andreas Hotho, and 1 more author

   In Conference on pervasive and ubiquitous computing adjunct publication, 2013

   Abs HTML

   An increasing number of platforms like Xively or ThingSpeak are available to manage ubiquitous sensor data enabling the Internet of Things. Strict data formats allow interoperability and informative visualizations, supporting the development of custom user applications. Yet, these strict data formats as well as the common feed-centric approach limit the flexibility of these platforms. We aim at providing a concept that supports data ranging from text-based formats like JSON to images and video footage. Furthermore, we introduce the concept of extensions, which allows to enrich existing data points with additional information, thus, taking a data point centric approach. This enables us to gain semantic and user specific context by attaching subjective data to objective values. This paper provides an overview of our architecture including concept, implementation details and present applications. We distinguish our approach from several other systems and describe two sensing applications namely AirProbe and WideNoise that were implemented for our platform.

2012

1. Ubicon: Observing Physical and Social Activities

   Martin Atzmueller, Martin Becker, Stephan Doerfel, and 7 more authors

   In International conference on green computing and communications, Nov 2012

   Abs DOI

   The connection of ubiquitous and social computing is an emerging research area which is combining two prominent areas of computer science. In this paper, we tackle this topic from different angles: We describe data mining methods for ubiquitous and social data, specifically focusing on physical and social activities, and provide exemplary analysis results. Furthermore, we give an overview on the Ubicon platform which provides a framework for the creation and hosting of ubiquitous and social applications for diverse tasks and projects. Ubicon features the collection and analysis of both physical and social activities of users for enabling inter-connected applications in ubiquitous and social contexts. We summarize three real-world systems built on top of Ubicon, and exemplarily discuss the according mining and analysis aspects.

2. Generic Pattern Trees for Exhaustive Exceptional Model Mining

   Florian Lemmerich, Martin Becker, and Martin Atzmueller

   In European conference on machine learning and knowledge discovery in databases, 2012

   Abs DOI

   Exceptional model mining has been proposed as a variant of subgroup discovery especially focusing on complex target concepts. Currently, efficient mining algorithms are limited to heuristic (non exhaustive) methods. In this paper, we propose a novel approach for fast exhaustive exceptional model mining: We introduce the concept of valuation bases as an intermediate condensed data representation, and present the general GP-growth algorithm based on FP-growth. Furthermore, we discuss the scope of the proposed approach by drawing an analogy to data stream mining and provide examples for several different model classes. Runtime experiments show improvements of more than an order of magnitude in comparison to a naive exhaustive depth-first search.

2011

1. Constraint Based Descriptive Pattern Mining

   Martin Becker

   University of Wuerzburg, 2011

   HTML