Multiple kernel support vector machine can generate weights of feature matrices for toxicity prediction

• Item type: SIG Technical Reports(1)
• 公開日: 2017-09-19
• タイトル: Multiple kernel support vector machine can generate weights of feature matrices for toxicity prediction
• タイトル / 言語: en
• タイトル: Multiple kernel support vector machine can generate weights of feature matrices for toxicity prediction
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: Theoretical Cell Science Lab, Center for iPS Cell Research and Application, Kyoto University
• 著者所属: Theoretical Cell Science Lab, Center for iPS Cell Research and Application, Kyoto University
• 著者所属: Theoretical Cell Science Lab, Center for iPS Cell Research and Application, Kyoto University
• 著者所属(英): en
• 著者所属(英): Theoretical Cell Science Lab, Center for iPS Cell Research and Application, Kyoto University
• 著者所属(英): en
• 著者所属(英): Theoretical Cell Science Lab, Center for iPS Cell Research and Application, Kyoto University
• 著者所属(英): en
• 著者所属(英): Theoretical Cell Science Lab, Center for iPS Cell Research and Application, Kyoto University
• 著者名: Hiroki, Takahashi; Junko, Yamane; Wataru, Fujibuchi
• 著者名(英): Hiroki, Takahashi; Junko, Yamane; Wataru, Fujibuchi
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: In order to reduce drug discovery period and costs, development of prediction system for toxicities and effects of medicine by artificial intelligence (AI) is expected. In our laboratory, a toxicity prediction system with multiple kernel support vector machine (MK-SVM) was constructed using SHOGUN library. In this study, sub-kernel matrices are created from three feature matrices (qRT-PCR expression values, correlations between genes by Bayesian network, and structure-activity relationships of each compound quantitated by E-Dragon) and a kernel matrix generated by the linear sum of these sub-kernel matrices was used for SVM prediction. Weights of each sub-kernel matrix in the linear sum indicate the contribution degree of each feature matrix in the classification. Therefore, focusing on the weights, we discuss whether each feature matrix can correctly predict toxicity or not.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: In order to reduce drug discovery period and costs, development of prediction system for toxicities and effects of medicine by artificial intelligence (AI) is expected. In our laboratory, a toxicity prediction system with multiple kernel support vector machine (MK-SVM) was constructed using SHOGUN library. In this study, sub-kernel matrices are created from three feature matrices (qRT-PCR expression values, correlations between genes by Bayesian network, and structure-activity relationships of each compound quantitated by E-Dragon) and a kernel matrix generated by the linear sum of these sub-kernel matrices was used for SVM prediction. Weights of each sub-kernel matrix in the linear sum indicate the contribution degree of each feature matrix in the classification. Therefore, focusing on the weights, we discuss whether each feature matrix can correctly predict toxicity or not.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA12055912
• 書誌情報: 研究報告バイオ情報学（BIO）; 巻 2017-BIO-51, 号 4, p. 1-4, 発行日 2017-09-19
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8590
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会