engleski

Genetic Algorithm-enhanced Parallel Chemical Space Exploration Utilising Multiple Peptide Libraries

Finding novel compounds with specific traits, such as antimicrobial or cell-penetrating activity, presents a challenging endeavour for chemists. Large portions of the chemical space can be systematically explored using random peptide libraries. However, the challenging characterization of such mixtures due to mass and sequence overlapping of peptide permutations presents a crucial limitation. We present a novel approach for parallel chemical space exploration based on multiple peptide libraries. The approach utilises a 3-objective NSGA-II genetic algorithm with an early-stopping criterion based on Pareto front hyperarea. By incorporating expert input as a guideline at the start of the process, the algorithm seeks to find the specified number of subsets of the initial peptide library while maximising the number of peptides within the libraries, as well as intra-library mass diversity and cross-library sequence diversity. Optimised libraries are presented in the form of Pareto fronts, which gives a broad spectrum of potential solutions to choose from. Preliminary results presented in Table 1 suggest that the algorithm is able to traverse large distances in chemical space in a relatively short time, and therefore maximise search space coverage while keeping the execution time and resource expenditures to a minimum. The algorithm was early-stopped after completing 712 iterations. The three final libraries, which resided on the best Pareto front, were chosen on the basis of offering wide search space coverage (100% cross-library sequence diversity), while having sufficient, 95% intra-library mass diversity. Given that the peptide design is NP-hard, we believe the proposed approach could be of value for improving drug design success rates.

Genetic algorithm ; Peptides ; Peptide library

nije evidentirano