Research Ma Group

AI for Precision Medicine and Smart Health

The AI for Precision Medicine and Smart Health Lab led by Dr. Tianle Ma seeks to develop novel AI and machine learning tools for precision medicine and smart health.

Research Areas

Unify Data-driven and Knowledge-driven AI Approaches for Precision Medicine

In the era of big data, we often face the challenge of integrating heterogeneous data and knowledge sources. Multi-modality machine learning is needed to enable knowledge discovery from a wide variety of different types of data (e.g., discrete, continuous, text, spatial, temporal, graphs) and knowledge sources (i.e., ontologies, literature). We have developed several machine learning and deep learning frameworks for biomedical data and knowledge integration. We continue pursuing research in this field and build a robust and generalizable AI for precision medicine and smart health.

Integrative Analysis of Single-cell Multi-omics Data for Systems Biology

A cell can be modeled as a heterogeneous network of various molecular entities. Only recently can we measure molecular states and interactions with single-cell resolution. Single-cell multi-omics data bring us unprecedented opportunities to reveal molecular networks underlying complex cellular systems. It is in urgent need to develop computational methods that can effectively mine single-cell multi-omics data. We are developing systems biology and integrative network approaches that can reconstruct cellular networks from molecular measurements underlying various conditions such as healthy and diseased states.

Modeling and Mining Heterogeneous Information Networks

In cellular systems, information flow underlies material and energy flow. We can construct an information diffusion network that models cellular communications. In this network, various types of nodes represent different molecular entities (such as genes, proteins, miRNAs, and so on). Various edges represent different molecular interactions (e.g., protein-protein interactions, protein-miRNA interactions, etc.). Both nodes and edges have their own attributes. Some attributes are from domain knowledge (e.g., gene annotations), while others are from molecular measurements (e.g., gene expressions). We are developing novel computational methods to identify causal links (edges) in the molecular circuitry and quantify the downstream effects if some molecular entities are perturbed (for example, a genetic mutation is introduced, or the abundances of certain molecules are abnormal). The computational toolbox developed in information and graph theory especially for modeling and mining heterogeneous information networks can be utilized to perform these inference and reasoning tasks. We are interested in creating such an integrative network model with inference and reasoning capabilities specifically to address the unique challenges in analyzing single-cell multi-omics and genome-wide CRISPR screen data.