Research

The Biomedical Data Research Laboratory (BDRL) develops computational methods to understand how metabolism shapes human health and disease. We integrate multi-omics data, systems biology, and artificial intelligence to connect molecular changes with cellular functions and whole-body physiology. Our goal is to discover disease mechanisms and translate these insights into therapeutic strategies and precision nutrition.

1. AI-powered systems biology and multi-omics modeling

We develop interpretable computational models to study biological processes through metabolic, transcriptional regulatory, and signaling networks. By integrating single-cell and spatial transcriptomics with proteomics, metabolomics, and other data, we investigate how molecular regulation and cellular environments shape biological function.

A central challenge is to infer biological processes from incomplete measurements, often collected at a single time point. We address this challenge by combining biological knowledge, mathematical constraints, and information across samples and experimental conditions. Our methods emphasize identifiability, uncertainty assessment, and biological interpretation, with the goal of generating hypotheses that can be tested experimentally.

2. Metabolic landscapes, disease mechanisms, and therapeutic opportunities

We investigate how metabolism varies across cell types, tissues, organs, and disease states. Our research brings transcriptomic and proteomic information into metabolic analysis to characterize pathway activity, metabolic capabilities, and interactions among cells and tissues.

A major focus is identifying metabolic stresses that contribute to disease progression and understanding the adaptations that allow cells and tissues to survive these stresses. We integrate metabolic network analysis, flux modeling, and computational perturbations to prioritize mechanisms and potential interventions for experimental validation.

Our disease-focused studies include cancer, cancer cachexia, and metabolic disorders. In cancer, we examine how metabolic interactions within the tumor microenvironment influence tumor progression, immune function, and treatment response, including opportunities to improve immunotherapy.

3. Precision nutrition informed by molecular and physiological variation

We seek to understand why individuals respond differently to nutrients, foods, and dietary interventions. Our approach integrates omics, clinical and real-world data, and evidence from scientific literature to connect nutritional exposures with metabolic states and health outcomes.

We aim to identify the mechanisms underlying variation in nutritional requirements and responses, and to develop computational approaches that guide personalized nutritional interventions. This work connects molecular biology with nutrition science to support hypothesis generation, intervention design, and evaluation.

4. Whole-body metabolic modeling and digital twins

We are developing whole-body metabolic models to investigate how organs exchange nutrients and metabolites and coordinate systemic metabolism. These models connect cellular and tissue-level processes with physiological states such as feeding, fasting, physical activity, and disease.

Our long-term goal is to build metabolic digital twins that integrate individual molecular, clinical, and physiological measurements. We aim to use these models to generate testable predictions about disease-associated metabolic changes and responses to nutritional or therapeutic interventions.

5. AI for biological reasoning and scientific discovery

We develop AI and statistical methods that connect complex biological data with mechanistic knowledge. Our methodological interests include interpretable representation learning, biological knowledge graphs, graph-based models of metabolic and regulatory processes, spatial and temporal modeling, and transfer learning across biological systems.

Our existing AI co-scientist platform, MetaKnogic, provides a foundation for integrating scientific literature and biological knowledge to support mechanism discovery and hypothesis generation. We are advancing its ability to reason from experimental observations, compare competing explanations, and propose informative follow-up experiments. Across these efforts, we emphasize biological context, transparent reasoning, and rigorous evaluation against experimental evidence.

Ongoing funding supports:

• NSF DBI CAREER: Mining biological functions from single-cell multi-omics data (Chi Zhang, PI). The major goal of this project is to develop AI-powered systems biology models for single-cell multi-omics data and natural language processing approaches to construct biological networks grounded in published evidence.

• NSF CISE CAREER: Disentangled learning of high dimensional biomedical data in the presence of inherent heterogeneity (Sha Cao, PI). Building on the previously funded CRII project, this project develops statistical and machine learning methods to identify, infer, and quantify subspace structures and biological heterogeneity in high-dimensional biomedical data.

• NIH NIGMS MIRA R35: Development of data driven and AI empowered systems biology to study human diseases (Chi Zhang, PI). The major goal of this research program is to develop interpretable AI and systems biology approaches that integrate omics data with biological knowledge to model metabolic and immune activities, characterize disease mechanisms, and identify therapeutic vulnerabilities.

• NIH NIGMS MIRA R35: Advancing computational modeling of disease metabolism by integrating AI and systems biology (Sha Cao, PI). The major goal of this research program is to develop computational frameworks for studying disease-associated metabolic changes and interactions among cellular populations in the tissue microenvironment.

• American Cancer Society Research Scholar Award: Computational modeling of cancer specific metabolic variations for improving therapeutic target detection (Sha Cao, PI). The major goal of this project is to develop computational approaches to characterize cancer-specific metabolic changes and identify metabolic vulnerabilities that can guide therapeutic target discovery, with applications to breast and pancreatic cancer.

• American Cancer Society Research Scholar Award: Computational modeling of stromal-immune cell interactions in triple-negative breast cancer (Chi Zhang, PI). The major goal of this project is to identify stromal cell subtypes and determine how their interactions with immune cells regulate T-cell infiltration and cytotoxic activity in triple-negative breast cancer.

• NIH NLM R01: SCH: AI-empowered wearable multimodal sensors (AI-MEDALLION) for noninvasive monitoring of treatment effect and progression of Parkinson’s disease (Chi Zhang, lead PI). The major goal of this project is to combine AI-powered biomarker discovery, metabolic modeling, and wearable multimodal sensors to enable noninvasive monitoring of Parkinson’s disease progression and treatment responses.

• NSF FDT-BioTech: NOURISH: Nutritional Optimization Using Real-time Integrated Sensors and Health Digital Twins (Chi Zhang, co-PI and OHSU site PI). The major goal of this project is to integrate real-time wearable sensors, AI, and whole-body metabolic digital twins to characterize individual metabolic states and support personalized nutritional recommendations.

Selected previously funded projects:

• NIH NIGMS R01: Construction of cell type specific gene co-regulation signatures based on single cell transcriptomics data (Chi Zhang, site PI). The major goal of this project was to develop computational approaches to identify cell population-specific transcriptional regulatory relationships using single-cell RNA-sequencing data.

• Susan G. Komen Foundation Big data-guide metabolic targeting of metastasis-promoting immune cells (Chi Zhang, co-PI). The major goal of this project was to integrate computational analysis and experimental studies to identify metabolic vulnerabilities in immune cells that promote cancer metastasis.

• NIH NCI Developmental and Hyperactive Ras Tumor (DHART) SPORE (Chi Zhang, co-Investigator). Our group contributed statistical analysis, multi-omics integration, and data management to collaborative studies of Ras-driven tumors and therapeutic responses.

Selected collaborative programs and research partnerships:

• Pancreatic Ductal Adenocarcinoma Stromal Reprogramming Consortium (PSRC) — computational and statistical studies of pancreatic cancer stroma, metabolism, and therapeutic responses.

• OHSU Knight Cancer Institute and Center for Biomedical Data Science — collaborative research in cancer systems biology, AI, single-cell and spatial omics, and precision oncology.

• OHSU Brenden-Colson Center for Pancreatic Care — integrated computational and experimental studies of pancreatic cancer metabolism, metastasis, tumor–microenvironment interactions, and cachexia.

• Developmental and Hyperactive Ras Tumor (DHART) SPORE (dhartspore.org).

• Indiana University Melvin and Bren Simon Comprehensive Cancer Center (cancer.iu.edu).

• Indiana Alzheimer’s Disease Research Center (program website).

• IUSM/Purdue Alzheimer’s Disease Drug Discovery Center (program website).

• IUSM breast cancer research and triple-negative breast cancer immunotherapy collaborations (program website).

• Indiana University Center for Computational Biology and Bioinformatics (center website).