Skip to main content
Guest  homeNews home
Story
1 of 10

RIT researchers use machine learning to better understand the pathways of disease

Cancer, Alzheimer’s, and other diseases follow a pathway in the human body. It starts at the molecular and cellular levels, and through a series of complex interactions can lead to the development and progression of disease.

At RIT, a new project funded by the National Institutes of Health (NIH) is using artificial intelligence to map the full journey of illnesses and discover entirely new disease pathways. If successful, the RIT research could transform how scientists understand disease and speed the discovery of new drugs and treatments for some of today’s most pressing health challenges.

Rui Li, associate professor in RIT’s Golisano College of Computing and Information Sciences, received a nearly $1.8 million grant from the NIH for his project to advance research at the intersection of statistical machine learning and computational biology. Working with students, Li will design new machine learning models and techniques that enable disease pathway discoveries.

The challenge for Li is to create a holistic view of a complex molecular network of genes that are not well connected. If those genes have problems, it can have a cascade of effects. Atypical activities of these molecules, and their interactions, can contribute to disease.

“Using machine learning, we can identify new disease-associated molecules and biomarkers that could be potential therapeutic targets,” said Li. “This information is crucial for developing new drugs and could help caregivers.”

A second challenge is that diseases manifest differently in different body tissues, such as nervous tissue or muscle tissue. For example, disease-associated molecules that could collectively lead to one disease, like Parkinson’s, will manifest differently in different parts of the brain.

Li explained that current deep learning techniques are limited because they encode independently. Experts can make an average, but it will lose the uniqueness of the pathway making it hard to know how one molecule affects another.

“I propose a hierarchical model to capture the uniqueness at the local level, while also sharing complementary information at the global level,” said Li. “We will allow it to aggregate information from its neighboring nodes, which mimics what happens in the human body.”

Li’s recent NIH grant will support a new methodology developed to model molecular network data and its topological structure. For efficiency, these methods will integrate deep learning techniques with probabilistic inference.

The effort is powered by RIT’s Lab for Use-inspired Computational Intelligence, which Li directs. Four Ph.D. student researchers— Mahendra Singh Thapa, Paribesh Regmi, Sicy Ruochen Shi, and Jeevan Thapa—are currently helping analyze more than 100,000 molecular interactions.

“I’m fascinated by how perfectly our biological systems work, from tiny molecules to the whole organism,” said Thapa, a computing and information sciences Ph.D. student from Nepal. “I enjoy using machine learning to uncover hidden patterns and explore the mysteries of biology.”

The NIH-funded research will have two stages. First, the lab will conduct comprehensive studies of about 500 diseases, focusing initially on cancer, cardiovascular disease, and immune system disorders. Next, the work will expand to larger-scale applications across diverse disease types.

“I’m really excited to be working in this area and with Professor Li because it combines my passion for machine learning with important questions in computational biology,” said Regmi, a computing and information sciences Ph.D. student from Nepal. “I enjoy the challenge of developing algorithms that can not only advance AI research but also contribute to real-world impact in health and medicine.”

The five-year award comes through the NIH Maximizing Investigators’ Research Award program, which supports new and early-stage investigators developing innovative approaches to biomedical challenges. The project is called "Large-scale Disease Pathway Discovery by Integrating Tissue-specific Molecular Networks via Hierarchical Bayesian Inference on Graph Neural Networks".

Latest All News