我不是业内的，道听图说一点

my kid was involved in a UC berkeley bioengineering research project on a multitarget work (drug is one of several applications?). Too many combinations of compounds to make (on wet bench outsourced actually) before testing out effects. The project stalled for over 1-2 years. They have access to Google protein model (not sure if it is Alphafold). Later someone used one useful ML algorithim and narrowed down # of compounds to make significantly. Got really good results. They tried different ML algorithims before but not useful. Paper submitted to Science if I remeber correctly. It might be just a research with no impact on application.