Our group develops bioinformatics and machine-learning based approaches to understand complex human diseases through the integration of multi-omics and clinical data. Our research combines bioinformatics, machine learning, and systems biology to uncover molecular mechanisms underlying disease progression and identify predictive biomarkers for precision medicine.
A central focus of our work is understanding biological heterogeneity across tissues, cell types, and patients in order to identify robust molecular signatures and regulatory mechanisms. Our current applications span neurodegeneration, chronic inflammation, and cancer.
Complex tissues and patient cohorts are highly non-uniform, and critical disease drivers are frequently masked by biological noise. We develop structured bioinformatics workflows to deconstruct this heterogeneity across bulk, single-cell, and spatially resolved dimensions. By dissecting complex microenvironments into distinct cellular lineages and tracking altered cell states, we map the underlying gene regulatory networks and cell-to-cell communication axes. This multi-scale approach allows us to isolate true pathological mechanisms, transforming heterogeneous clinical samples into reproducible molecular signatures and actionable therapeutic targets.
Moving beyond standard "black-box" predictive algorithms, we implement interpretable machine learning frameworks designed to extract clear, causal features from complex patient data. By focusing on explainable AI, we uncover the specific molecular checkpoints and regulatory hinges associated with disease progression and distinct clinical phenotypes. This ensures our computational models provide transparent, biologically verifiable insights that wet-lab scientists and clinicians can confidently validate.
A cornerstone of our translational strategy is mapping multi-disease comorbidities at the systemic level. Centered around our project AI-PREDICT, we employ deep learning and multimodal data integration to decode the shared molecular pathways linking neurodegeneration and chronic gut inflammation. By targeting the gut-brain axis to predict comorbidity progression between IBD and Parkinsons disease, we aim to uncover unified precision biomarkers and discover novel therapeutic windows.
Data-driven discoveries are only as valuable as their reproducibility. In alignment with global open-science frameworks, we champion FAIR (Findable, Accessible, Interoperable, and Reusable) data principles and robust data stewardship across all our pipelines. We build open-source software packages, standardized data packages, and fully containerized workflows. This sustainable infrastructure ensures our multi-omics datasets and computational tools are immediately accessible, understandable, and reusable by the global scientific community.