演題概要

オーラルセッション

第3日　5月17日（金）　10:30～10:50　D会場（202）

3D-O1-1030　　PDF

深層学習に基づくペプチド由来イオンピークの新規検出手法

(¹DBCLS・ ²京大院薬・ ³京大CiRA・ ⁴理研AIP・ ⁵北大ICReDD)
^o守屋勇樹¹・ 田畑剛^2,3・ 岩崎未央³・ 河野信¹・ 五斗進¹・ 石濱泰²・ 瀧川一学^4,5・ 吉沢明康²

Both the detection of monoisotopic peak masses and the charge numbers of peptide ions by "peak-picking" in the MS1, and obtaining a peak list of fragment ions from the MS2 are crucial steps for protein identification in the mass spectrometry (MS)-based proteomics. Although the peak mass and the charge of peptide ion dramatically affect identification of peptide in the database search, detection of accurate mass and charge are often difficult because multiple peaks derived from different peptide ions are overlapping in complex on the mass spectrum. In this study, we tried to detect the peak mass and the charge based on deep learning. As the result, we succeeded in detecting correct peaks that were missed by conventional peak-picking methods.