A new artificial intelligence (AI) tool has predicted how likely some DNA changes are to cause disease in humans.
The AI tool looked at one particular type of genetic variant, known as 'missense' variants, and enabled researchers to categorise 89 percent of all possible examples that can occur in the protein-coding regions of the human genome as either 'likely pathogenic' (disease-causing) or 'likely benign'. Four million missense variants have been observed in humans, and previously, only an estimated two percent have been classified.
Dr Pushmeet Kohli, joint corresponding author of the paper published in Science, told BBC News that previously, researchers had to search for potentially disease-causing regions across billions of chemical building blocks that make up DNA. 'Researchers can now focus their efforts on the new areas, that they were not aware of and we have highlighted as potentially disease-causing,' he said.
Missense variants are changes in DNA that result in a different amino acid being produced in a certain position within a protein. On average, each person has more than 9000 missense variants, while most are benign, others can alter the resulting protein's shape and ability to function. These changes are known to cause diseases in humans, such as sickle cell anaemia, cystic fibrosis or cancer. Classifying missense variants is an important step in understanding which of these protein changes could give rise to disease.
Dr Žiga Avsec and Dr Jun Cheng led the research to develop the AI tool, named AlphaMissense, at Google DeepMind in London. The tool was used to predict the pathogenicity of all 216 million possible single amino acid changes across all known human proteins. This resulted in a catalogue of 71 million missense variant predictions.
'Ultimately, we hope that AlphaMissense, together with other tools, will allow researchers to better understand diseases and develop new life-saving treatments', wrote Dr Avsec and Dr Cheng.
AlphaMissense is an adapted version of AlphaFold, an AI tool released by DeepMind in 2020 that predicted structures for nearly all proteins known to science from their amino acid sequences. The improved AlphaMissence tool also considers how frequently different missense variants are seen in humans and other 'closely related primate populations'.
The AlphaMissense tool operates within a protein language, which Dr Cheng, described to the Guardian as 'similar to human language. If we substitute a word in an English sentence, a person familiar with English can immediately see whether the word substitution will change the meaning of the sentence or not.'
The predictions from AlphaMissense have been validated by Genomics England to be consistent with known variant pathogenicity data. Dr Ellen Thomas, deputy chief medical officer at Genomics England told BBC News: 'The new tool is really bringing a new perspective to the data. It will help clinical scientists make sense of genetic data so that it is useful for patients and for their clinical teams'.
However, Drs Avsec and Cheng cautioned that the tool does not allow for instant diagnosis, and clinical expertise is still needed.
Sources and References
-
A catalogue of genetic mutations to help pinpoint the cause of diseases
-
Accurate proteome-wide missense variant effect prediction with AlphaMissense
-
Google DeepMind AI speeds up search for disease genes
-
Google DeepMind AI tool assesses DNA mutations for harm potential
-
DeepMind is using AI to pinpoint the causes of genetic disease
Leave a Reply
You must be logged in to post a comment.