Using AI to Accelerate the Discovery of New Medicines

Lead Researchers:  Prof. Ram Samudrala and Prof. Zackary Falls
Institution: University at Buffalo

The Challenge: Making Drug Discovery Faster and More Effective

Developing new medicines is a lengthy, expensive and complex process. Traditional drug discovery approaches often focus on how a particular compound interacts with a single protein associated with a disease. While effective in some circumstances, this approach can overlook the complex biological interactions that influence whether a drug will work, cause side effects or prove useful in treating other conditions.

In reality, drugs interact with multiple proteins and biological systems simultaneously. Understanding these relationships could help researchers identify promising treatments more quickly, predict potential side effects and discover new uses for existing medications.

The challenge is the enormous number of possible interactions. Evaluating thousands of compounds against thousands of proteins generates vast amounts of data, requiring sophisticated artificial intelligence and substantial computing power.

Researchers at the University at Buffalo are developing an approach that could help change how scientists discover and design medicines.

The Researchers’ Approach: Understanding Billions of Molecular Interactions

Ram Samudrala, Ph.D., professor of biomedical informatics at the University at Buffalo’s Jacobs School of Medicine and Biomedical Sciences, is leading the development of CANDO — Computational Analysis of Novel Drug Opportunities.

Working with Zackary Falls, Ph.D., assistant professor of biomedical informatics and one of CANDO’s principal developers, Samudrala is advancing an AI-powered platform designed to calculate and analyze billions of interactions among and between compounds/drugs, proteins and other biological molecules.

Rather than examining one drug and one protein at a time, CANDO takes a broader, systems-level approach. It uses machine learning and computational modeling to understand how compounds interact with multiple proteins and how those interactions relate to different diseases.

CANDO was initially developed to identify new therapeutic uses for existing drugs, a process known as drug repurposing. Researchers have since expanded its capabilities to predict potential side effects and drug interactions, identify promising new compounds, and design molecules with specific therapeutic properties.

The team is also integrating multiple sources of biological and chemical information into increasingly sophisticated AI models. These advances could eventually support more personalized approaches to medicine, including identifying treatments best suited to patients with particular genetic characteristics.

How Empire AI Accelerates the Research

The continued development of CANDO requires significant computational resources. Analyzing billions of molecular interactions, training increasingly sophisticated AI models and integrating diverse biological datasets place substantial demands on computing infrastructure.

Empire AI’s advanced computing resources can allow the researchers to dramatically expand the scale and complexity of their work.

With Empire AI, the team can explore much larger collections of molecular interactions, develop more powerful predictive models and evaluate potential drug candidates at a scale that would be difficult to achieve using conventional university computing systems.

These capabilities could allow researchers to move more quickly from identifying promising molecular interactions to selecting compounds for experimental testing. Empire AI also creates opportunities to incorporate emerging advances in artificial intelligence and structural biology into CANDO’s drug-discovery framework.

By enabling more comprehensive analysis of complex biological systems, Empire AI can help the researchers pursue questions that previously would have been constrained by available computing power.

Potential Impact: New Medicines and More Personalized Treatments

The ultimate goal of CANDO is to make drug discovery faster, more efficient and more effective.

By identifying promising compounds earlier in the development process, researchers hope to reduce the time and expense associated with bringing new medicines to patients.

CANDO research has already led to the computational identification, experimental validation and optimization of promising therapeutic candidates, including  novel compounds targeting non-small cell lung cancer and opioid use disorder. Work involving the platform has also contributed to biotechnology efforts exploring treatments for these and other conditions, including diseases associated with aging.

As CANDO’s capabilities expand, the platform could help researchers identify promising treatments, predict potential side effects and toxicity, discover new uses for existing medicines, and advance personalized medicine by accounting for genetic and biological differences among patients.

The technology’s potential also extends beyond medicine. The same approach to understanding molecular interactions could help researchers develop new research probes, reagents, detergents, pesticides, sensors and other useful compounds.

With Empire AI’s computing power, Samudrala and Falls have an opportunity to expand the scale and sophistication of CANDO — accelerating the search for new treatments while advancing a more comprehensive understanding of how medicines interact with the human body.