BPS2026 – Construction of de novo β-barrel nanopores using in silico evolution

Nanopore sensing is an electrical single-molecule detection method. The nanopore DNA sequencer has been commercialized, and the next application of nanopore sensing is amino acid sequencing of proteins. However, a significant issue is the lack of suitable biological nanopore proteins with appropriate pore sizes and chemical properties for amino acid identification. To approach this challenge, we tried the rational (de novo) design of nanopores. In our previous report, we designed a β-barrel peptide nanopore, named SVG28, which forms β-barrel pore structures by the self-assembly of β-hairpin peptide monomers. SVG28 formed pore structures and could detect single-molecule polypeptides, but its pore-forming ability was still inferior to that of natural pore-forming proteins. In this study, we sought to obtain SVG28 variants with enhanced pore-forming ability using in silico protein evolution, which yields high-functioning variants by repeatedly predicting protein structures, screening based on prediction scores, and introducing further mutations. As the first round of SVG28 evolution, we generated 28,000 sequences by introducing several mutations into SVG28 and screened pore-forming variants using AlphaFold2. To experimentally validate the reliability of the prediction, six pore-forming variants were selected and confirmed to have membrane-disrupting ability in the liposome leakage assay compared to SVG28. Then, 30 cycles of in silico evolutionary rounds were performed. For each round, 100 sequences were produced. To reduce computational cost, we developed a lightweight pore-forming classification model using molecular descriptors and random forest. Before structure prediction with AlphaFold2, we employed this model to narrow down the sequences. As a result of evolution, we could obtain SVG28 variants with higher pore-forming potential.