AI searches for problematic genes

Conventional approaches to finding the genetic causes of diseases are most often looking for a single mutation. But most real diseases are much more complicated. According to the author of the study, Professor Adilson Motter, “it’s like a plane crash — it rarely happens that only one cause is to blame. More often, it’s a chain of failures.” The new method helps to find this whole chain.

The team has developed a tool called TWAVE. This is an AI model that can recognize which genes are turned on and off in a cell during a disease. Moreover, she looks not at the DNA itself, but at how the genes manifest themselves — that is, at the activity inside the cell.

It is important that TWAVE can mimic the condition of a diseased and healthy cell. Thanks to this, it is possible to trace exactly how the work of genes changes in different diseases. And — most importantly — which of them really trigger the disease, and not just exist in the background.

It turned out that the same diagnosis can have different genetic causes in different people. This means that the approach to treatment should also be different. One patient needs some drugs, another needs others, even if they have the same diagnosis. This is the approach that underlies personalized medicine.

In addition, the new method helps to circumvent sensitive issues related to privacy. It doesn’t use the genetic code directly — only data about how cells work. This is safer and more accurate, especially when it comes to the interaction of genes with the external environment, nutrition, stress, or lifestyle.

Scientists have already tested TWAVE on a number of diseases and found genes that could not be detected using previous methods. The next step is to use this data in clinical practice in order to select therapy on a point—by-point basis.