Bioinformatics and the Missense Genome: a 10-Year-Old Scenario for Predicting Pathogenic and Benign Diseases

About 10 years ago, Žiga Avsec was a PhD physics student who found himself taking a crash course in genomics via a university module on machine learning. The lab that he was working on was studying rare diseases and trying to figure out the genetic cause of a rare disease.

Avsec says it was a needle in a haystack problem. There were millions of potential culprits lurking in the genetic code—DNA mutations that could wreak havoc on a person’s biology. Of particular interest were so-called missense variants: single-letter changes to genetic code that result in a different amino acid being made within a protein. Small changes in the body can have a large and long-reaching effect, because the structure of the body is made of two building blocks, mino acids and proteins.

Of the 4 million missense variants that have been spotted in humans, only 2 percent have been categorized as either pathogenic or benign, through years of painstaking and expensive research. It can take months to study the effect of a single missense variant.

According to Marsh, computational predictions have a minimal role in the diagnosis of genetic diseases and physicians groups say that these tools should only demonstrate the link between a bug and a disease. Avsec said that AlphaMissense was able to find a greater amount of missense genes than have been done before. I think people will be more willing to trust these models as they get better.

Joseph Marsh, Computational Biologist at theMRC Human Genetics Unit in Edinburgh, UK, agrees that its impact will not match the impact of AlphaFold. “It’s exciting. The predictor we have right now is probably the best. It will be the best predictor over the next two or three years. There’s a good chance it won’t be.”

A bioinformatician at Emory University in Atlanta, Georgia emphasizes that AlphaMissense must be rigorously evaluated before ever being applied in the real-world.

For several years, the performance of prediction methods has been compared against experimental data that has yet to be released with the help of the Critical Assessment of Genome Interpretation (CAGI). It is my worst nightmare to see a doctor taking a prediction and trying to run with it without having any evidence of its validity, or an evaluation by the CAGI.