[20p-P11-1~25] 12.6 Nanobiotechnology

[20p-P11-12] Machine-learning assisted interaction analyses of amino acid residues in Chignolin

〇Yuji Mochizuki^1,2, Yuto Komeiji³, Tsuyoshi Iyama¹, Akira Okusawa⁴, Takeshi Makimura⁴, Takaya Nakanishi⁴, Shigenori Tanaka⁵ (1.Rikkyo Univ., 2.Univ. Tokyo, 3.AIST, 4.Knowledge Communication Co. Ltd., 5.Kobe Univ.)

Keywords:Fragment Molecular Orbital (FMO) method,Machine Learning,Interaction Energy

Nowadays, the fragment molecular orbital (FMO) calculations can be performed on a series of protein structures with statistical fluctuations, and then a large dataset of inter fragment interaction energy (IFIE) is generated. Such big data are frequently hard to manually understand their chemical meanings. We thus have introduced the machine learning (ML) assisted approach to automatically analyze the IFIE dataset. In this presentation, we show the first example of Chignolin mini-protein consisting of only 10 amino acid residues (100 structure samples). The ML-based analyses are reasonable in a sense of chemical intuition. We have a vision that ML techniques for FMO calculations of proteins are applicable not only to data analyses but also to predictions and designs.