Deep representation learning for drugs and proteins with neural networks

• Project/Area Number: 17H07392
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Single-year Grants
• Research Field: Life / Health / Medical informatics
• Research Institution: National Institute of Advanced Industrial Science and Technology
• Principal Investigator: Tsubaki Masashi 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 研究員 (80803874)
• Research Collaborator: Asho Hideki ; Kanemura Atsunori
• Project Period (FY): 2017-08-25 – 2019-03-31
• Project Status: Completed (Fiscal Year 2018)
• Budget Amount: ¥2,990,000 (Direct Cost: ¥2,300,000、Indirect Cost: ¥690,000); Fiscal Year 2018: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000); Fiscal Year 2017: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
• Keywords: 深層学習 / 創薬 / 機械学習 / 人工知能

Outline of Final Research Achievements

Received the Best Paper Award at the workshop of Advances in Neural Information Processing Systems (NIPS 2017), the largest international conference in machine learning. It was also adopted by Bioinformatics, an international journal in the field of bioinformatics. At the same time, the developed software was released to the public. In fact, it was decided to conduct joint research with a pharmaceutical company. I was able to create a series of flows from industrial design to the application of the method which is a part of the foundation, to the dissertation and the release of software, and from there to joint research with companies, to industrial applications.

Academic Significance and Societal Importance of the Research Achievements

この研究成果の学術的な意義としては、まず、グラフ構造のような離散データについても、深層学習の有効性を検証できたという点である。特に、これまで特徴量や記述子を使って、データを一旦変換した上で、つまり情報を人間の観点から削減した上で機械学習手法を適用していたものが、データのより原始的な情報を入力として扱えるようになった。また、社会的な意義としては、これまで新薬の開発が難しかった病気などに対して、コンピュータのアプローチから迫ることができる点である。特に、機械学習手法はシュミレーションなどの異なり、予測の精度が非常に速いことが大きな利点である。

Research Products

• [Journal Article] Quantitative estimation of properties from core-loss spectrum via neural network (2019)
  • Author(s): Shin Kiyohara, Masashi Tsubaki, Kunyen Liao, and Teruyasu Mizoguchi
  • Journal Title: Journal of Physics: Materials Volume: 1 Pages: 1-2
  • Peer Reviewed / Open Access / Int'l Joint Research
• [Journal Article] Compound-protein Interaction Prediction with End-to-end Learning of Neural Networks for Graphs and Sequences (2018)
  • Author(s): Masashi Tsubaki, Kentaro Tomii, and Jun Sese
  • Journal Title: Bioinformatics Volume: 35 Pages: 309-318
  • Peer Reviewed / Open Access / Int'l Joint Research
• [Journal Article] Fast and Accurate Molecular Property Prediction: Learning Atomic Interactions and Potentials with Neural Networks (2018)
  • Author(s): Masashi Tsubaki and Teruyasu Mizoguchi
  • Journal Title: The Journal of Physical Chemistry Letters Volume: 9 Pages: 5733-5741
  • Peer Reviewed / Open Access / Int'l Joint Research
• [Journal Article] Mean-field theory of Graph Neural Networks in Graph Partitioning (2018)
  • Author(s): Tatsuro Kawamoto, Masashi Tsubaki, and Tomoyuki Obuchi
  • Journal Title: Advances in Neural Information Processing Systems Volume: 1 Pages: 1-2
  • Peer Reviewed / Open Access / Int'l Joint Research
• [Presentation] 深層学習を用いた化合物とタンパク質の相互作用予測 (2018)
  • Author(s): 椿真史
  • Organizer: 創薬インフォマティクス研究会
  • Invited
• [Presentation] End-to-end Learning of Graph Neural Networks for Latent Molecular Representations (2017)
  • Author(s): Masashi Tsubaki
  • Organizer: Advances in Neural Information Processing Systems (NIPS 2017) Workshop, Machine Learning for Molecules and Materials,
  • Int'l Joint Research
• [Remarks] 化合物の薬剤活性予測ソフトウエア
  • URL: https://github.com/masashitsubaki/GNN_molecules
• [Remarks] 化合物とタンパク質の相互作用予測ソフトウエア
  • URL: https://github.com/masashitsubaki/CPI_prediction
• [Remarks] 化合物の立体構造データからの物性値予測ソフトウエア
  • URL: https://github.com/masashitsubaki/QuantumGNN_molecules