機械学習とベイズ推論の融合による免疫受容体タンパク質の設計手法の開発

结合机器学习和贝叶斯推理开发免疫受体蛋白的设计方法

• 批准号: 22K18003
• 负责人: 郭 中梁
• 金额: $ 3万
• 依托单位: Aichi Cancer Center Research Institute
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Early-Career Scientists
• 财政年份: 2022
• 资助国家: 日本
• 起止时间: 2022-04-01 至 2024-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-22K18003/
• 关键词: TCR; タンパク質設計; 機械学習; サンプリング; ベイズ推論

本研究は，機械学習とベイズ推論を組み合わせて，高精度かつ効率的な免疫受容体タンパク質の設計手法を確立することを目指している．T細胞受容体（TCR）は，がん免疫においてがん細胞を認識する重要な役割を果たすが，患者に合わせた適切なTCRを見つけることが困難である．本研究では，タンパク質設計問題を統計のサンプリング問題と捉え，機械学習モデルを用いてTCR配列から結合能を予測し，TCR配列空間をサンプリングして高い結合能を持つTCR配列を発見する．当該年度では，申請者はベイズ推論に基づくTCR配列のサンプリング手法を開発し，その性能を確認した．従来のサンプリング手法では，高い結合能を持つTCR配列を見つけられるものの，類似した配列が多く存在し，結合能の高いローカルモードにとらわれる問題があった．この問題に対処するため，クラスタリングを導入した逐次モンテカルロ法（SMC）を開発し，全体の探索と結合能の高い領域の優先的な探索のバランスを制御することに成功した．ただし，この手法はpHLAとTCRの結合能予測モデルに依存し，小規模データや結合能データが存在しないpHLAには適用できない．申請者は，シミュレーションと機械学習の両方から解決策を模索していた．シミュレーションモデルは予測精度が低く，時間もかかるため，本研究には適さないことが判明した．機械学習モデルは高い精度を実現できるが，タンパク質の立体構造情報が必要であり，配列からの高精度予測方法は未確立である．現在，申請者はマルチモーダル学習とタンパク質立体構造予測モデルを活用し，配列から高精度な結合能予測モデルの開発を進めている．

In this study, mechanical engineering experiments were conducted to determine the accuracy of immune tolerance test. T-cell acceptor (TCR) was used in the design of high-precision immune receptacle design. In this study, the immune system was used to cut the important equipment, and the patient was used to cut the immune system. This study was conducted. The mechanical engineering equipment uses the combination of TCR and TCR to maintain the performance of TCR. In the year of the year, the applicant will discuss the basic theory of TCR. Please make sure that the performance is good. In recent years, you have been able to improve the performance of the system. The combination of high-performance and high-performance is able to support the TCR configuration of the system. The type of configuration is similar to that of the system, and the combination can be used to improve the performance of the problem. The problem is to improve the performance of the system, and the combination of high-performance and high-performance can be used to improve the performance of the system. The combination of high-performance and high-performance makes sure that the performance is good. In recent years, the combination of high-performance and high-performance has been implemented. The combination of high-performance and high-performance can maintain the performance of the TCR configuration system. The combination of all exploration systems can be used in the first place in the field of exploration and control systems. The combination of pHLA and TCR techniques can be used to determine the dependence of the system, and the combination of small-scale models can verify that there are significant pHLA applications. Applicants. In this study, the accuracy of the accuracy is low, and the accuracy is low. In this study, the accuracy of the accuracy is low. In this study, the accuracy of the accuracy is low, and the accuracy is low. In this study, the accuracy of the mechanical engineering is improved, and it is necessary to improve the accuracy of the system. The high-precision prediction method does not ensure the accuracy of the system. At present, the applicant is required to create a real-time operating system for the operating system, and the combination of the high-precision combination of the system can enable the operation of the system.