Google launched AI, which can help understand how our genes work. Known as AlfagenNew language model Predictable changes in DNA, which affect molecular processesas genetic activity. Google believes that this will help scientists understand the influence of genetic variations on diseases.
In the message on the blog, Deepmind Google proposed more details of the alfagenoma, AI, which predicts the influence of individual mutations on DNA. According to engineers, the model You can analyze DNA sequences to predict molecular propertiesFor example, the beginning and end of genes, RNA splashing or protein binding sites.
New model Use neural networks to detect templates throughout the sequence And make detailed forecasts of each DNA letter. There is alfagen Compatible with sequences up to 1 million letters (or pairs of bases), a characteristic that allows you to capture more details, such as areas that regulate genes from afar.
DeepMind engineers indicate that this long sequence context was the main restriction of coercion, the previous model. Unlike his predecessor, alfagenom It covers indistinguishable areas that contain options related to diseasesIn addition, AI, trained in data from scientific consortiums, such as Encode, GTEX, 4D Nucleome and Fantom5, which cover hundreds of cellular types of people and mice.
How Alfagen works, the new Google AI, which analyzes DNA
In practical terms, Alfagen takes a DNA sequence as an entrance and analyzes its molecular properties. The process predicts where the genes begin and end in different types of cells, as well as their molecular properties. AI can guess where RNA splashing occurs, how many RNA occurs or where certain proteins are connected to DNA.
Subsequently, Alfagen compares a normal sequence with the mutant to evaluate its effects. II is so strong that You only need a second to qualify the influence of mutation In all molecular properties. DeepMind engineers mention some genetic diseases, such as muscle atrophy, will be caused by errors in RNA splashing.
Alfagen It has already been used to model a possible mutation associated with cancerThe field in the internal test was predicted that mutations in certain parts of the genome are introduced for protein, which is activated by the TAL1 gene. This activation can contribute to the development of acute T -limforatal leukemia.
“For the first time, we created a unique model that unites many different problems that are related to the understanding of the genome,” said Pushmet Kokhli, vice -president of DeepMind for research.
A Alfagenoma API will be available today For research laboratories, universities and public health organizations. Deepmind also invited academic employees, governments and sectors of cooperation in their development. Alfagen is the successor of AlfafoldAI, which predicts a three -dimensional structure of proteins for which they won the Nobel Prize in Chemistry in 2024.
Source: Hiper Textual
