Deep learning for antimicrobial peptides; plus Zymeworks’ anti-HER2 bispecific and more

University of Warsaw and Medical University of Gdańsk researchers reported in Nature Communications a deep learning model, dubbed HydrAMP, that generates diverse highly active anitmicorbial peptides by separating the learnt representation of a peptide from its expected antimicrobial properties to create new peptide designs.

Wet-lab experiments validated the activity of nine generated peptides against five bacterial strains, including Gram-positive, Gram-negative and antibiotic-resistant strains, and demonstrated low toxicity of the peptides in mammalian red blood cells...