Structure Biology has laid out a blueprint for success in AI for science

Structural biology became the first major scientific discipline to reap the rewards of AI when Alphafold de facto solved the longstanding protein structure prediction problem in 2020. Its AI algorithm is truly far superior for the “messy” protein folding process and is now virtually the only viable approach left. A pivotal factor to its success is the wealth of protein structures (~170K) carefully curated and made public by the community. Undoubtedly, high quality, public databases are indispensable prerequisites for any other scientific field to write its own AI success story. Another critical factor is the rigorous and transparent evaluation of competing algorithms. This practice is imperative because of the barriers to replicating training and test procedures outlined in publications.