Google DeepMind has created artificial intelligence that can determine whether millions of genetic mutations are harmless or can cause disease. Company researchers say which identified 89% of all key variations. The new development could speed up the diagnosis of rare diseases and help develop new drugs.
DNA is made up of four blocks: adenine (A), cytosine (C), guanine (G) and thymine (T). When a human embryo develops, the order of these letters is read to produce proteins, which are the building blocks of the cells and tissues that make up the various parts of the body.
A new program called AlphaMissense, looks at so-called nonsense mutations. These are variations of the DNA code in which one letter is misspelled. There are 9,000 such mutations in the typical human genome. In many cases they may be harmless, but this change is also associated with diseases such as cystic fibrosis or cancer.
There are about 71 million possible variations of absurdity, but only 0.1% were classified by human experts. Only in this small number of cases was it known whether these were benign mutations or whether they were capable of causing some kind of pathology. Google DeepMind, an Alphabet subsidiary specializing in artificial intelligence and scientific research, would completely change analysis in this sense.
Google DeepMind will change the understanding of DNA mutations
The company says its AlphaMissense was able to classify 89% of the variants. The artificial intelligence model assigned “pathogenicity index” from 0 to 1 for each of the 71 million possible missense mutations. It was based on information about the effects of other closely related mutations that were known. The higher the score, the more likely it is that a particular mutation will occur that could cause or be associated with a disease.
Of the total number of cases 57% were classified as probably benign.. 32% were identified as potentially pathogenic. Of the remaining 11%, Google’s DeepMind artificial intelligence was unable to detect its impact.
The study was published in the journal The science. DeepMind researchers teamed up with Genomics England, a government body examining a set of genetic data collected by the United Kingdom’s National Health Service. The article states that AlphaMissense achieved an accuracy level of 90%.
Google DeepMind has provided free access to AlphaMissense predictions to the scientific community. “Researchers can now focus their efforts on new areas that they were unaware of that we have flagged as potentially causing disease,” said Pushmeet Kohli, Google DeepMind spokesman. BBC.
Awaiting confirmation
Ben Lehner, head of the human genetics group at Britain’s Wellcome Sanger Institute, said Google’s DeepMind artificial intelligence predictions still needed to be verified by other scientists. In principle, the expert noted, this seems to be a good means of identifying which changes in DNA cause diseases and which do not.
“One of the problems with the DeepMind model is that it is extremely complex.”Lehrer told The keeper. “A model like this may be more complex than the biology it is trying to predict…Will doctors be comfortable making decisions about patients that they don’t understand and can’t explain?”
Google DeepMind also acknowledged that its predictions need to be interpreted in light of other sources of evidence. In his statement, he clarified that his results “They are not intended for direct clinical use.”
However, the company highlighted its potential to improve the diagnosis of rare genetic diseases. “Ultimately, we hope that AlphaMissense, along with other tools, will allow researchers to better understand diseases and develop new life-saving treatments.”
Source: Hiper Textual
