GenomeGAN: in silico genome design with generative adversarial networks

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K22897
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 62:Applied informatics and related fields
• Research Institution: Keio University
• Principal Investigator: Sato Kengo 慶應義塾大学, 理工学部(矢上), 講師 (20365472)
• Project Period (FY): 2019-06-28 – 2022-03-31
• Keywords: バイオインフォマティクス / ゲノム合成 / 敵対的生成ネットワーク / 深層強化学習 / RNA配列設計 / RNA二次構造

Outline of Final Research Achievements

In order to generate a genome sequence with specific traits, we tackled the RNA sequence design problem of designing an RNA sequence that forms specific secondary structures. By using deep reinforcement learning as a learning method for optimizing the search of sequence space, more efficient generation of sequences for the target secondary structure is achieved. The optimization method for converting discrete nucleotide sequences into a differentiable representation using Activation Maximization was applied to the RNA sequence design problem. We improved IPknot, a method for predicting RNA secondary structure including pseudoknot structures, to achieve linear computational time with respect to sequence length.

Free Research Field

バイオインフォマティクス

Academic Significance and Societal Importance of the Research Achievements

合成生物学は，生命を再構成することによってその完全な理解を目指す究極のアプローチであると同時に，生物の工学的な応用に繋がることからその産業的な価 値も極めて高い.しかし，生命として完全に機能するゲノム配列を設計して，人工的な生命を合成することは困難を極める挑戦的な課題である.