Algorithm Predicts Best Therapies

Nature Genetics has published a study co-led by Dr. Fran Supek (IRB Barcelona) and Dr. Ben Lehner (CRG) describing RTDetective, a machine-learning tool trained on 140,000 measurements of read-through at premature stop codons. Using this data, the algorithm predicts, for every possible premature stop in the human transcriptome, which of six small-molecule drugs will most effectively restore full-length protein production.

Because premature stop codons underlie about 20% of monogenic diseases and inactivate important tumour-suppressor genes, RTDetective could immediately help match patients to the best read-through therapy—for example in cystic fibrosis, Duchenne muscular dystrophy and certain cancers. By prioritizing compounds before clinical trials, it may also speed up the development of new nonsense-suppression treatments.

More information: ‘Detective’ algorithm predicts best drugs for genetic disorders and cancer