R-CCS Cafe 第221回 第2部

An antibody-based drug molecule disrupts the function of its target by tightly binding to the target. It is possible to derive an antibody capable of binding to a desired target either by implementing biotechnology methods or by alternative means. To fully optimize the binding ability of such a naïve antibody, targeted changes of specific amino acids in the structure are necessary. We previously demonstrated that implementing a double point mutation strategy, that is changing two amino acids at the same time, enhanced the binding ability. In the study, we designed the double-point mutation based on careful visual inspection of interface structures, generated by a conventional structural modeling program, of multiple antibody mutant and antigen complexes. Such an operation requires manual labor and substantial experience of structure-based drug design. In addition, we found that greater precision in structure modeling was required than was possible with the conventional program. Herein, we describe our recently developed automatic structure modeling protocol that is capable of building the precise structure of double-point amino acids to the 1-Å level. This was achieved by a combination of machine-learning and our novel feature generation method that reflects our experience of structure-based drug design.