An average human genome has 9,000 genetic mutations. While most are safe, some can lead to cancer, cystic fibrosis, or impairment to the development of the brain.
Researchers at Google’s DeepMind artificial intelligence division have created a ground-breaking tool that can determine whether genetic abnormalities are likely to cause harm, aiding in the study of uncommon diseases.
An average human genome has 9,000 genetic mutations. While the majority are safe, some can lead to cancer, cystic fibrosis, or impairment to the development of the brain.
Out of a potential 71 million such mutations, scientists have only so far identified 4 million in humans. Only 2% of them have been identified as benign or disease-causing.
How does the AI tool function?
The Google AI program AlphaMissense generates predictions regarding so-called missense mutations, which occur when the genetic (DNA) code contains one incorrect letter. After evaluating the mutation, the AI algorithm could accurately anticipate 90% of them, or 89% of them.
The program predicted that 32% of missense mutations were likely detrimental and that 57% were likely harmless when the researchers raised the program’s precision to 90%. The remaining 7% were anticipated to be uncertain.
Scientists provided the AI system DNA information from people and closely related primates to teach it about missense mutations, some of which are quite frequent and hence benign, and others which are uncommon and possibly harmful.
The machine familiarized itself while the scientists were teaching it by analyzing millions of protein sequences and learned what a “healthy” protein looked like.
The AI tool creates a score for each mutation after being fed a mutation, indicating the likelihood that it will cause disease (also known as being pathogenic).
“Human language is pretty comparable to this. A person who is familiar with the English language can immediately tell whether a word substitution will alter the sense of a phrase, according to Google DeepMind’s Jun Cheng.
Scientists can now access the database created by AlphaMissense in conjunction with a study that was published in the journal Science.
Researchers are optimistic.
The discovery, according to scientists, may pave the way for the creation of novel therapies and the assessment of the danger of a fatal mutation before it reaches an irreversible point.
We must stress that the forecasts were never truly developed or meant to be used only for clinical diagnosis, added Jun.
However, we do believe that our predictions could help us uncover novel disease-causing genes and possibly boost the rate at which uncommon diseases are diagnosed.
The onus is still on the human doctors to comprehend the cause and eventually apply the knowledge to operate on the patient, even though the AI tool can determine whether a genetic mutation is hazardous.