AlphaFold 3, the latest iteration from DeepMind, predicts protein structures with approximately 80% accuracy, a significant improvement over previous models. It can also model DNA, RNA, ligands, and their modifications, completing the process in seconds, compared to the years it takes humans using advanced microscopy.
While AlphaFold 3 offers high accuracy, it doesn't explain the underlying reasons for protein folding. Its impact on drug discovery remains unclear; despite solving the protein-folding problem being a step forward, it won't magically improve drugs' success in clinical trials.
Concerns exist regarding AlphaFold 3's limited free use and the lack of public access to its inner workings. Given its potential impact on healthcare and the fact its training data included publicly funded research, alternative revenue models are suggested that would avoid paywalls or exorbitant prices.