オーラルセッション
- 第3日 5月17日（金） 10:50～11:10 D会場（202）
-
3D-O1-1050 PDF
理論質量値を用いた機械学習による皮膚常在菌プロテオタイピング
Bacteria’s features differ depending on their strains or subtypes, therefore it is critical to discover biomarkers to distinguish them. In this Study, we created discrimination model of Cutibacterium acnes with theoretical mass as the first step of developing the method of automatically discovering biomarkers. Although Cutibacterium acnes are classified into 6 subtypes (IA1, IA2, IB, IC, II, III), we excluded type IC because we could not get strain of the type. We turned 7 ribosomal proteins into training data and created the model using random forest. In the results, the attendance rate of discriminating measured peak lists was 63.9%. The results had the deviation by subtypes, however, we think that it is sufficiently possible to discriminate subtypes with theoretical mass because that rate of type IA1, IB and II were extremely accurate.