基于人工智能网络的肽段保留时间预测
详细信息 查看官网全文
- 英文论文题名：Peptide Elution Time Prediction for Reversed-Phase Liquid Chromatography-MS by Artificial Neural Network
- 论文作者：陆文渊 ; 刘晓慧 ; 张扬 ; 杨芃原
- 英文论文作者：Wenyuan Lu ; Xiaohui Liu ; Yang Zhang ; Pengyuan Yang ; Institute of Biomedical Sciences ; Fudan University
- 年：2016
- 作者机构：复旦大学生物医学研究院;
- 论文关键词：人工智能网络 ; 肽段 ; 保留时间
- 会议召开时间：2016-07-01
- 会议录名称：中国化学会第30届学术年会摘要集-第二十三分会：复杂样品分离分析
- 语种：中文
- 分类号：Q51
- 学会代码：ZGHY
- 会议名称：中国化学会第30届学术年会-第二十三分会 ; 复杂样品分离分析
- 会议地点：中国辽宁大连
- 主办单位：中国化学会
- 学会名称：中国化学会
- 页数：1
- 文件大小：655k
- 原文格式：D
- 会议级别：全国
摘要
本文采用了基于人工智能网络的肽段保留时间预测方法,用以对蛋白质组学中的LC-MS实验结果进行训练与预测。该方法不仅考虑了肽段序列中氨基酸位置对保留时间的影响,同时还考虑到了肽段长度以及疏水性等性质,这些肽段性质可被用来提高保留时间预测的准确性。该人工智能网络采用了三层架构,其中输入层有652个节点,隐藏层有22个节点,以及输出层的1个节点。为了增加人工智能网络的训练可信度,同时减少不必要的拟合误差,本文选用了长度小于25个氨基酸的肽段作为训练集与测试集,同时样本都进行了一维色质联用的三次重复实验。该方法将人宫颈癌细胞样本中的8850条肽段作为训练集,最终得到的训练模型的决定系数为0.92。
We applied an improved artificial neural network(ANN)-based method for predicting peptide retention times in reversed-phase liquid chromatography(RPLC).In addition to the peptide amino acid composition,this study compared several other peptide descriptors to improve the predictive capability,such as peptide length and hydrophobicity.An ANN architecture that consisted of at most 652 input nodes,22 hidden nodes,and 1 output node was used to fully consider the amino acid residue sequence in each peptide.The network was trained using 8850 unmodified peptides identified from a LC-MS/MS analyses of Hela protein digested standards.Then the predictive capability of the model was tested using confidently identified peptides from three repeated LC-MS/MS analyses of Heart sample.The model demonstrated a correlation fit(square of R) of 0.92.
We applied an improved artificial neural network(ANN)-based method for predicting peptide retention times in reversed-phase liquid chromatography(RPLC).In addition to the peptide amino acid composition,this study compared several other peptide descriptors to improve the predictive capability,such as peptide length and hydrophobicity.An ANN architecture that consisted of at most 652 input nodes,22 hidden nodes,and 1 output node was used to fully consider the amino acid residue sequence in each peptide.The network was trained using 8850 unmodified peptides identified from a LC-MS/MS analyses of Hela protein digested standards.Then the predictive capability of the model was tested using confidently identified peptides from three repeated LC-MS/MS analyses of Heart sample.The model demonstrated a correlation fit(square of R) of 0.92.
引文