Comprehensive genome mutagenesis of phage ΦX174 reveals unexpected lethal mutations and AI shortfalls
Scientists altered almost every base in the ΦX174 virus genome, discovering many changes are fatal and that AI models could not reliably predict these outcomes.
A collaborative effort headed by Ben Lehner at the Wellcome Sanger Institute, alongside Huijin Wei of the Centre for Genomic Regulation and Xianghua Li of Kings College London, systematically mutated virtually every nucleotide and amino-acid residue in the 5,386-base circular genome of bacteriophage ΦX174, generating more than 44,000 distinct variants. These mutants were cultured with Escherichia coli for two to three infection cycles, enabling fitness measurement via DNA sequencing.
The experiment found that about 50 % of single-nucleotide changes and 60 % of amino-acid alterations reduced viral viability, while a small number of mutations enhanced replication. Many lethal mutations remained unexplained, and leading AI systems for predicting harmful variants performed poorly on this dataset. The results highlight the necessity for richer experimental data to improve AI-driven biological predictions. The work was posted on bioRxiv in July.
Why it matters
The study shows AI's current limits in forecasting viral mutations, stressing the need for more experimental data in biology.
In this story
