Mutating every DNA letter of a genome shows surprising effects — and the limits of AI
2026-09-01 08:00Models🔥 28.9 heat score
1sources
1days unfolding
28.9heat score
4mentions
SummaryAI generated
The researchers identified nearly 44,000 single nucleotide and amino acid mutations in the entire genome of bacteriophage ΦX174. They found that half of the single nucleotide mutations and 60% of the amino acid mutations were harmful to the virus’s survival, and the mechanisms of about one-fourth of the harmful mutations remained unknown. Some mutations, however, enhanced the virus’s adaptability in E. coli. Advanced AI models performed poorly in predicting the biological consequences of these mutations, especially in terms of protein interaction disruption and structural damage, highlighting their dependence on experimental data and insufficient understanding of complex gene functions.
Researchers systematically examined nearly 44,000 single-nucleotide and amino acid mutations across the entire genome of bacteriophage ΦX174, finding that half of the nucleotide mutations and 60% of amino acid mutations were detrimental to viral survival, with about one-quarter of these harmful mutations remaining unexplained. Despite decades of study, one-quarter of the mutations still lead to viral death, and the underlying mechanisms remain elusive. Advanced AI models perform poorly in predicting the biological consequences of these mutations, particularly in detecting disruptions to protein interactions or deep structural damage. The study revealed that some mutations actually enhanced the virus's adaptability within its host, E. coli. These findings highlight the current limitations of biological AI tools, including their heavy reliance on experimental data and their insufficient understanding of complex gene functions.