AI identifies key gene sets that cause complex diseases
For decades, scientists have struggled to understand why diseases like asthma or cancer develop. The problem is that there’s rarely a single faulty gene to blame. Instead, a hidden network of genes often works together
Image by Cybernews
For decades, scientists have struggled to understand why diseases like asthma or cancer develop. The problem is that there’s rarely a single faulty gene to blame. Instead, a hidden network of genes often works together.
Now, biophysicists at Northwestern University have developed a computational tool that can identify combinations of genes responsible for complex diseases such as diabetes, cancer, and asthma.
“Many diseases are determined by a combination of genes, not just one,” said Adilson Motter, senior author and professor at Northwestern.
Unlike disorders caused by one gene, complex diseases stem from interactions between multiple genes. The huge number of possible combinations has made this a long-standing challenge.
Motter compares it to an airplane crash. In most cases, multiple failures need to occur, and different combinations can lead to similar outcomes.
Traditional genetic studies tend to focus on one gene at a time, which often misses how groups of genes work together to influence disease.
To address this, the team developed a powerful artificial intelligence model called TWAVE. Instead of analyzing DNA sequences, TWAVE learns from gene expression data. This data shows which genes are turned on or off in cells and reveals groups that can collectively trigger disease.
“We are not looking at DNA sequences, but gene expression,” said co-author Wytock.
“Our model learns what healthy and diseased profiles look like based on clinical trial data.”
Gene expression shows the active state of a cell, including the effects of the environment, which DNA alone cannot capture.
In testing, TWAVE identified known gene groups and also discovered new ones. It showed that two people with the same illness might have different genetic patterns, which could lead to more personalized treatments.
“Different genes can cause the same disease in different people,” said Motter. “This information can help guide personalized therapy.”
The study was funded by the National Cancer Institute, the National Science Foundation, and the Simons Foundation. It will appear soon in the Proceedings of the National Academy of Sciences.