@techreport{oai:ipsj.ixsq.nii.ac.jp:00092653,
 author = {大野, 亮仁 and 藤, 博幸 and 山名, 早人 and Yorihito, Ohno and Hiroyuki, Toh and Hayato, Yamana},
 issue = {11},
 month = {Jun},
 note = {G タンパク共役受容体 (G-protein-coupled receptor，以下 GPCR) は，内在性リガンドと結合することで細胞外からの様々なシグナルを細胞内に伝達しており，新薬開発の重要なターゲットとして注目されている．しかし，GPCR と化合物の組合せは膨大であるため，計算機による正確な結合予測手法が求められている．先行研究として，GPCR を構成するアミノ酸配列全長が持つ化学的性質と化合物の化学的性質を用いて結合を予測する手法がある．しかし，GPCR には立体構造が既知のものがあり，その細胞外側の領域にリガンド結合部位が決まっている．よって，リガンド結合部位のアミノ酸が結合に強く影響を与えると考えたため，リガンド結合部位のアミノ酸に注目すべきと考えた．本研究では，全長配列を使用する代わりに，リガンド結合部位のアミノ酸のみを利用することで予測の改善を試みた．特徴量として結合部分のアミノ酸と化合物の化学記述子を用い，SVM により GPCR と化合物の結合を予測したところ，アミノ酸配列全長を用いた時に比べ Accuracy が 3.6%，F 値は 0.038，AUC は 0.002 向上した．, G-protein-coupled receptors (GPCRs) are involved in the transduction of signals carried by the endogenous ligands into cytosolic regions, which are regarded as important targets to develop new drugs. Accurate prediction of interaction between GPCRs and chemical compounds is keenly required for drug development, because the number of the combinations of GPCR and the compounds is too large to be examined by experiments. Therefore, such computational approaches have been extensively investigated. One of the preceding studies by Okuno et al. had succeeded to achieve high performance in prediction by using the entire amino acid sequence of a GPCR and the chemical feature of a chemical compound. However, the amino acid residues involved in the ligand binding are quite limited. We estimate that the residues could strongly affect the binding. So, we identified the amino acid residues constituting ligand binding region from the 3D structure of GPCR. Then, we examined whether the use of the residues, instead of entire amino acid sequence, can improve the prediction. Support vector machine (SVM) was used for the prediction. Experimental result showed that the accuracy was improved by 3.6%, Fvalue was improved by 0.038% and AUC was improved by 0.002%, comparing to the approach by Okuno et al.},
 title = {SVMによるGタンパク共役受容体と化学化合物の相互作用予測},
 year = {2013}
}