Google has developed an AI system called AlphaGenome designed to analyze the functional potential of non-coding DNA. The software evaluates various biological parameters, including gene expression, transcription initiation, chromatin accessibility, histone modifications, and splice site usage.
The system currently operates using data from human and mouse sequences. It is trained on a limited set of cell types that have been previously studied by biologists. The tool aims to help researchers determine the potential significance of mutations found in non-coding genomic regions.
Google used AlphaGenome to evaluate every possible single-base change across the human reference genome. By testing every substitution for each position, the system pre-calculates potential mutation consequences, allowing scientists to search for specific functional impacts across the entire genome.
While the system provides insights into genomic sequences, its practical utility remains subject to further adoption by the scientific community. Future development aims to enable reliable analysis for genomes without existing training data, such as those of Neanderthals or Denisovans.
