Abstract
High-throughput virtual screening (HTVS) has emerged as a pivotal strategy for identifying high-affinity peptides targeting functional proteins, which are crucial for diagnostic and therapeutic applications. In the HTVS of peptides, expanding the library capacity to enhance peptide sequence diversity, thereby screening out excellent affinity peptide candidates, remains a significant challenge. This study presents a de novo design strategy that leverages directed mutation driven HTVS to evolve vast virtual libraries and screen peptides with ultrahigh affinities for various target proteins. Utilizing a computer-generated library of 10(4) random 15-mer peptide scaffolds, we employed a self-developed algorithm for parallelized HTVS with Autodock Vina. The top 1% of designs underwent random mutations at a rate of 20% for six generations, theoretically expanding the library to 10(14) members. This approach was applied to various protein targets, including a tumor marker (alpha fetoprotein, AFP) and virus surface proteins (SARS-CoV-2 RBD and norovirus P-domain). Starting from the same 10(4) random 15-mer peptide library, peptides with high affinities in the nanomolar range for three protein targets were successfully identified. The energy-saving and high-efficient design strategy presents new opportunities for the cost-effective development of more effective high-affinity peptides for various environmental and health applications.