Scientists from the Berlin Institute of Health and the Charité University Hospital in Berlin have specifically changed the control ranges of twenty disease-relevant genes. This enabled them to identify those changes that have the greatest influence on disease processes.
Many diseases are caused by mutations in the genetic material. These lead to a process in which vital protein molecules are not produced correctly: Sometimes the molecules themselves are altered in such a way that they can no longer perform their function. More often, however, changes in the genetic control areas lead to the wrong amount of proteins being produced. Either the cell produces far too much, far too little or no protein at all. "This leads to cancer, for example, if a protein that stimulates cell division is produced in an extremely high quantity," explained Martin Kircher, head of the Berlin Institute of Health (BIH or in German: Berliner Institut für Gesundheitsforschung) junior research group “Computational Genome Biology”, and main author of the publication.
These results now make it possible to predict which changes in patient’s genetic makeup are really responsible for the course of a certain disease, enabling the possibility of targeted therapy.
However, it is precisely in cancer cells that many mutations often occur, some of which do nothing, while others actually cause or drive the disease by influencing the amount of protein produced. So before physicians start a therapy that targets the effects of certain mutations, they should know how important they are for the disease.
Events in the cell strongly influenced by individual mutation and control ranges
The scientific team working with Dr. Kircher selected twenty disease-relevant genes and modified them building block by building block, in other words, they modified each base of their DNA. For this purpose, they developed a method through which the changes could be generated at high throughput and tested in parallel. They examined how the respective change affected protein production via cell culture.
"About 85% of the changes had no measurable effect, and of the remaining 15% about two-thirds reduced the amount of protein produced," said Dr. Kircher. And it depends to a large extent on the individual mutation and the control range itself how intensively it influences what happens in the cell: "If one base is exchanged for another, this usually has less influence than a complete elimination”, he added.
The studied control ranges come from genes altered in patients with cancer, heart failure, hereditary high cholesterol or various rare diseases. The results of the more than 30,000 mutation analyses have been made freely available on the Internet. Dr. Kircher now hopes that this data treasure trove will also be used: "It would be great if physicians who have analyzed the genetic material of their patients could check our database to see what effect the found mutation has on the probabilities of disease occurrence, and thus be able to estimate whether the change found in that specific patient could undergo a specific targeted therapy,'' he explained.
Such an effort, for twenty, or for possibly hundreds of thousands to millions of control ranges, raises the question of whether such a prediction cannot also be done with machine learning or artificial intelligence. There are already several computer programs that try to do precisely that matching between which mutation causes which effect. Dr. Kircher and his colleagues therefore also investigated how well different computer programs were able to predict the changes observed in cell culture. "Unfortunately, this was very disappointing," reported the bioinformatician, adding that "the predictions rarely coincided with our observations. Sometimes they even predicted the exact opposite”. The scientists now hope that their data treasure may also serve to improve the prediction programs.
Kircher M et al., Nature Communications 2019; 10: 3583. doi:10.1038/s41467-019-11526-w