Genetic mutations entail alterations in our DNA sequence, occurring as cells replicate during the process of cell division. These mutations serve as the primary wellspring of human genetic diversity and bear significance in evolutionary and medical genetics.
An alteration impacting our genes can potentially lead to a genetic disorder, although the presence of a mutation does not guarantee the development of such a disorder.
Hence, DeepMind, the artificial intelligence division of Google, has introduced AlphaMissense, a machine learning model, with the purpose of discerning which DNA variations within our genomes are probable culprits in causing diseases.
Additionally, DeepMind is unveiling a comprehensive catalog comprising 71 million potential variants capable of influencing the functionality of human proteins. According to a press release from Google DeepMind, some of these variants can contribute to the onset of conditions like cystic fibrosis, sickle-cell anemia, or cancer.
DeepMind is optimistic about the potential of its novel AI model to enhance the diagnosis of rare genetic disorders and facilitate the identification of previously unknown disease-causing genes. This advancement could empower medical professionals to gain deeper insights into diseases and advance the development of life-saving treatments.
The researchers emphasized in the press release that AI tools with the ability to accurately forecast the impact of genetic variants have the capacity to expedite research in various domains, spanning from molecular biology to clinical and statistical genetics.
Eyebrows Raised on DeepMind’s Work
DeepMind’s recent work has drawn attention, with questions arising about the commercial viability of AlphaMissense. Alex Zhavoronkov, the founder of Insilico Medicine, an AI-focused pharmaceutical company, remarked, ‘DeepMind is staying true to its nature, excelling in public relations and making commendable strides in the field of AI.’
Google reports that the AI model has successfully classified 89% of the extensive 71 million ‘missense’ variants, which involve single DNA letter changes affecting the protein generated by a gene, as potentially disease-causing.
Stephen Hsu, a physicist at Michigan State University specializing in genetic challenges through AI techniques, articulated, ‘The objective here is to provide insights into whether a protein alteration is detrimental to the human possessing it, instead of merely predicting the protein’s structure. Many of these genetic alterations remain enigmatic in terms of their disease-causing potential.’
Anticipating the effects of specific genetic variants is a resource-intensive and costly process. The application of AI tools can expedite research in this domain significantly. The researchers elaborated, ‘By employing AI predictions, scientists can obtain preliminary results for numerous proteins simultaneously, aiding in the efficient allocation of resources and accelerating more intricate investigations.
Building Upon a Pre-Existing Deepmind AI Model
AlphaMissense is developed upon the foundation of DeepMind’s earlier model, AlphaFold, renowned for its ability to forecast three-dimensional models of protein structures.
To equip AlphaMissense with its capabilities, researchers conducted a refinement process by utilizing AlphaFold’s expertise, focusing on identifying variations observed in both human and closely related primate populations.
Although AlphaMissense serves a distinct purpose compared to AlphaFold, it harnesses extensive databases containing related protein sequences and considers the structural context of genetic variants to generate a numerical score, typically ranging from 0 to 1. This score is a good indicator that the disease is caused by a genetic change.
The team believes that the predictions made by their new model have the potential to answer important questions in many branches of genomics and biological sciences.
“The results of Google DeepMind’s research were documented in the journal Science.
Conclusion
Genetic mutations lead to human diversity and disease. DeepMind’s AlphaMissense AI can predict DNA damage and provide a list of protein mutations. It promises to revolutionize rare genetic disease testing and gene discovery in science. Its economic value is still questionable, but its impact is potentially huge.
