BoltzGen cyclic-minipeptides targeting Acinetobacter baumannii lipoprotein-transporters

Alternative antibiotic candidates were explored against pathogenic Gramnegative Acinetobacter baumannii lipoprotein-transporters. For that, novel deep-learning BoltzGen in Colab A100 notebooks co-generated thousands of new cyclic-minipeptide-conformers. In particular, Alphafold2-modeled lipoprotein outer membrane localization (Lol) CE heterodimers were targeted by 3-to 21-amino acid minipeptide sequences. Boltzgen automatically targeted previously identified LolCE lipoprotein-binding cavities, co-generating fitting LolCE // cyclic-minipeptide-conformers. The new cyclic-minipeptides should facilitate their chemical synthesis, reduce protease degradation, increase chemical stability, and increase resistance to single LolCE mutations. To favor their possible experimental validation, the candidates were computationally filtered by hundreds of known toxicities and bacterial membrane permeability predictions. Despite exploring only a minimal percentage of the natural amino acid peptide wide-space, these candidates constitute novel strategies and provide proof-of-concept possibilities to deeper explorations.