Major BioAI advance on designing therapeutic peptides: PepMimic beat a leading method in overall performance on many tests. ✅ It designed 384 small protein pieces for five disease targets. About 8% stuck to their targets very strongly, and 26 had exceptionally strong stickiness. That is far better than picking at random. ✅ It could also copy other AI-made binders for two targets, working about 14% of the time overall. ✅ In mice, selected pieces homed in on tumors very well: about 9 times more in tumors with PD-L1 and about 16 times more in tumors with TROP2. ✅Overall, its success rate was tens of thousands of times higher than old-style random screening.