Overview
Human diseases are rarely caused by a single gene, pathway or cell type. Instead, they emerge from dynamic and context-dependent interactions across molecular, cellular, tissue and organ systems. These interactions also vary among patients, meaning that people with the same diagnosis may have different disease-driving mechanisms and treatment needs. At the same time, genomic, transcriptomic, epigenomic, spatial and clinical data are expanding faster than researchers can convert them into biological insights.
The Integrated Systems Biology and AI — Tailored Pharmacology and Precision Medicine Laboratory led by Hu Li, Ph.D., addresses this challenge by transforming complex biomedical data into mechanistic, testable and clinically meaningful models of disease. The lab uses state-of-the-art systems biology, network biology, systems pharmacology and artificial intelligence (AI) approaches. The long-term goal is to move beyond one-size-fits-all biomarkers and black-box predictions toward models that explain why disease occurs, identify where a biological system can be therapeutically redirected and support more-precise decisions for each patient.
A defining innovation of the lab is its hypothesis-centered approach to computation. Rather than applying generic algorithms after data have been collected, the team develops new biological and medical concepts and then designs computational approaches to test them. This strategy combines the pattern-recognition power of AI with established scientific knowledge. This helps produce results that are more focused, interpretable and useful for forming experimentally testable hypotheses. The lab's work on hypothesis-driven AI represents an important step toward using AI not only to classify or predict but also to reason about complex disease mechanisms and guide scientific discovery.
The Li lab has developed numerous systems biology algorithms and tools to decipher complexity of disease etiology and target identification:
- NetDecoder maps. Context-specific information from NetDecoder maps flows through biological networks and prioritizes influential disease genes and potential therapeutic targets.
- Machine Learning-Assisted Network Inference (MALANI). MALANI uncovers "dark" disease genes that may play important roles, even when they are neither mutated nor differentially expressed.
- Personalized Mutation Evaluator (PERMUTOR). PERMUTOR constructs disease networks and modules for individual patients, revealing how the importance of disease genes and potential treatment combinations can differ from person to person.
- Regulostat Inferelator (RSI) and Gene Utility Model (GUM). RSI and GUM characterize intrinsic response programs, context-dependent gene utility and regulatory networks that influence drug-response phenotypes.
Other platforms expand the role of AI in biological knowledge discovery:
- Artificial neural network encoder (ANNE). ANNE uses weight engineering to recover meaningful gene relationships encoded within trained neural networks, helping transform neural-network models from prediction tools into sources of biological insight.
- Spatially Informed AI (SPIN-AI). SPIN-AI identifies genes whose expression predicts the spatial organization of cells, offering a new way to investigate disease microenvironments.
- Learning-Based Invariant Feature Engineering (LIFE) platform. The LIFE platform explores symmetry and invariant gene-expression relationships as potential organizing principles of healthy and diseased phenotypes.
- Artificial Clinic Intelligence (ACI). ACI is a generative AI-powered platform for modeling clinical trial enrichment to inform clinicians on key clinical parameters that are enriched in prospective patients prior to accrual.
- Mini Galaxy. Mini Galaxy is a new ecosystem of AI-powered platforms to investigate and model complex diseases across multiple molecular, cellular, tissue and systems-level scales.
- Digital Organism Artificial Intelligence (DO-AI). DO-AI is a generative AI framework that instantiates a digital organism by inferring a person's multiorgan molecular state from omics measurements obtained from one or more accessible tissues.
- Senescence-AI (Sen-AI). Sen-AI is a multifaceted AI suit that uses deep learning to identify how genes contribute to context-specific cellular senescence programs in naturally occurring senescent cells in aging and disease across wide range of tissue types.
Together, these approaches seek biological information that conventional analyses centered only on mutation frequency, differential expression or spatial variability may miss.
The lab also develops multiscale conceptual frameworks that connect molecular networks with the behavior of cells, tissues, organs and the whole body:
- Manifold Medicine and Manifold Epigenetic Model (MEMo). MEMo expands the dimensions considered in treatment and regenerative health.
- The Multiorgan Locked-State Model (LoSM). LoSM examines how interorgan communication may stabilize healthy or pathological body-wide states. More recently, the Mini-Galaxy Model has framed cells as multilayered networks whose visible and hidden gene properties collectively shape disease, biomarker discovery, therapeutic targeting and drug repurposing. Emerging work further extends this vision to generative AI and virtual patient modeling for clinical-trial enrichment.
The Hu Li Lab is advancing a vision of precision medicine in which each patient is understood as an interconnected and evolving biological system. By revealing the networks that drive disease and treatment response, the lab seeks to help scientists discover previously hidden mechanisms, help clinicians develop more rational therapeutic strategies and help patients receive care tailored to the biology of their disease.