Most non-coding DNA is junk, but identifying the functional parts has been a challenge for biologists. Proteins that bind DNA aren't picky, and their behavior varies by cell type. Traditional software tools struggle with this complexity.

Google developed Alpha Genome, an AI system that evaluates DNA sequences for potential function. It predicts gene expression, transcription factor binding, and other regulatory features. The system is currently trained on human and mouse data, covering a limited set of cell types.

Despite these limits, Alpha Genome can help researchers assess whether changes in non-coding regions are significant. Its predictions are often as good as or better than specialized tools, offering a useful hypothesis generator for genetic studies.