Foundation Professor Guowei Wei is Leading Drug Discovery with AI

A new AI-powered program developed by an interdisciplinary team at Michigan State University, led by MSU Research Foundation Professor Guowei Wei, enhances drug discovery by translating three-dimensional molecular information into data usable by drug-interaction models, improving their ability to predict drug efficacy.

The team behind the new TopoFormer program was led by Guowei Wei (center) of Michigan State University and included Dong Chen (left), also of MSU, and Jian Liu, a visiting scholar from the Chongqing University of Technology in China. Credit: Finn Gomez/MSU College of Natural Science

A new AI-powered program developed by an interdisciplinary team at Michigan State University will allow researchers to enhance their drug discovery efforts. TopoFormer translates three-dimensional information about molecules into data that typical AI-based drug-interaction models can use, expanding those models’ abilities to predict how effective a drug might be. The effort is being led by MSU Research Foundation Professor Guowei Wei.

“With AI, you could make drug discovery faster, more efficient, and cheaper,” said Wei, who holds appointments in the Departments of Biochemistry and Molecular Biology, Mathematics, and Electrical and Computer Engineering.

In the United States, developing a single drug typically takes about a decade and costs around $2 billion, Wei explained. Half of this time is spent on clinical trials, while the other half goes into discovering new therapeutic candidates.

TopoFormer could significantly reduce development time and costs, potentially lowering drug prices for consumers. This is especially crucial for rare diseases, where limited patient numbers require higher drug prices to cover costs.

The first step in drug discovery is identifying which proteins a disease affects. Researchers then seek molecules that can counteract the disease's effects. Conventional models use chemical sequences from proteins and potential drugs to predict interactions, guiding drug development and clinical trials. However, these models often miss critical interactions that depend on molecular shape and three-dimensional structure.

Ibuprofen, discovered in the 1960s, exemplifies this. It has two molecules with the same chemical sequence but different 3D structures. Only one arrangement effectively binds to pain-related proteins.

“Current deep learning models can’t account for the shape of drugs or proteins when predicting interactions,” Wei noted.

TopoFormer addresses this gap. As a transformer model, it translates 3D information about protein-drug interactions into one-dimensional data that existing models can understand. “Topo” stands for “topological Laplacian,” referring to mathematical tools developed by Wei’s team to convert 3D structures into 1D sequences.

Trained on tens of thousands of protein-drug interactions, TopoFormer records each interaction as a piece of code, or a “word.” These words form sentences describing the drug-protein complex's shape, which other models can then read to predict new drug interactions more accurately.

TopoFormer is implemented in Python, and its source code is available on GitHub under the MIT Open Source License.

Wei is a MSU Research Foundation Professor, a title granted to highly accomplished current or incoming faculty members recommended by their college or dean. These distinguished researchers excel in their fields, furthering scholarly, disciplinary, or research areas crucial to MSU. Recipients retain the title throughout their tenure and typically receive scholarly support for the first five years after recognition. More than 60 professors have been honored with the MSU Research Foundation Professor title.

Read the full story at natsci.msu.edu