@techreport{oai:ipsj.ixsq.nii.ac.jp:02008126,
 author = {清水,正浩 and 杉田,昌岳 and 柳澤,渓甫 and 秋山,泰 and Masahiro Shimizu and Masatake Sugita and Keisuke Yanagisawa and Yutaka Akiyama},
 issue = {26},
 month = {Mar},
 note = {分子動力学（Molecular Dynamics, MD）シミュレーションは，複雑な分子の挙動を原子レベルを明らかにすることができる一方，高いエネルギー障壁によるサンプリング不足が生じやすい．この問題に対し，REST法とREUS法を組み合わせたREST/REUS法は有効であるが，レプリカ数や溶質温度集合T，拘束中心位置および拘束の強さの集合Uなど調整すべきパラメータが多く，手動調整は困難である．そこで本研究では，短時間のREST/REUS MDから得られるスナップショットのポテンシャルエネルギーとPMFに基づいて，隣接レプリカ間の交換確率が目標交換確率に近づくようNT, NU, T, Uを反復的に更新する自動調整手法を開発した．環状ペプチドB03，B08，B09の膜透過過程予測に適用した結果，production runにおける交換確率の平均絶対誤差はREST法においては平均1.98%pt，REUS法においては平均2.13%ptとなり，実用的精度で目標交換確率へ調整できた．, Molecular dynamics (MD) simulations can elucidate the behavior of complex molecules at the atomic level; however, they are prone to insufficient sampling due to high energy barriers. To address this issue, the combined REST/REUS method, which integrates replica exchange with solute tempering (REST) and replica exchange umbrella sampling (REUS), is effective. Nevertheless, REST/REUS involves many parameters that must be tuned―such as the numbers of replicas, the solute temperature set T, and the set U of restraint center positions and force constants―making manual tuning difficult. In this study, we developed an automated tuning method that iteratively updates NT, NU, T, and U based on the potential energies and the potential of mean force (PMF) obtained from short REST/REUS MD simulations, so that the exchange probabilities between adjacent replicas approach a target exchange probability. When applied to predicting the membrane permeation processes of cyclic peptides B03, B08, and B09, the method achieved practical accuracy in tuning toward the target exchange probability: in the production run, the mean absolute error of the exchange probability was 1.98%pt on average for REST and 2.13%pt on average for REUS.},
 title = {REST/REUS分子動力学法のパラメータ自動調整手法の開発},
 year = {2026}
}