化学発癌物質の計算機による構造活性相関 : 芳香族アミン(第2報)ラット, 肝臓データセット
スポンサーリンク
概要
- 論文の詳細を見る
A computer-assisted structure-activity study using pattern recognition methods was applied to handle a set of 153 aromatic amines containing 37 carcinogens and 116 noncarcinogens. All pharmacological data were obtained from rat experimental animals and their liver active site. A set of 48 descriptors are generated by the ADAPT system program which was specially developed for structure-activity studies. These 48 descriptors are divided into 6 small subsets. Each subset contains 40 descriptors. Important descriptors of each subset that could support to dichotomize the aromatic amines into carcinogen and noncarcinogen are selected by the variance method. These selected descriptors are not influenced by any perturbed factors caused by statistics or pattern recognition methods. Most of these important descriptors are related to molecular sizes and shapes. Various pattern recognition methods are applied to classify the aromatic amine data set mentioned above. The linear learning machine, Bayesian quadratic discriminant and the iterative least-squares linear discriminant development routine obtained good classification values. the prediction value by the linear learning machine using the leave two out method achieved 90.1% as the highest value.
- 公益社団法人日本薬学会の論文
- 1984-05-25
著者
関連論文
- グリーンケミストリーとADMEWORKS (特集 環境経営) -- (Green Products(グリーンプロダクト))
- バイオ,蛋白,化合物の統合情報:ドラッグインフォマティクス (特集 ライフサイエンス)
- 化学発癌物質の計算機による構造活性相関 : 芳香族アミン(第2報)ラット, 肝臓データセット
- 4A-2 完全分類を実現するK-step Yard sampling method (KY法)の提案 : Ames試験データ7000サンプルへの適用実験(セッション4A)