基于Bi-LSTM和Transformer的谱图预测模型  

Bi-LSTM and Transformer based model for spectrum prediction

作　　者：朱宇翀 陈德华[1] 潘乔[1] ZHU Yuchong;CHEN Dehua;PAN Qiao(School of Computer Science and Technology,Donghua University,Shanghai 201620,China)

机构地区：[1]东华大学计算机科学与技术学院,上海201620

出　　处：《智能计算机与应用》2025年第3期203-206,共4页Intelligent Computer and Applications

摘　　要：数据非依赖采集(DIA)近年来发展迅速,在蛋白质组学中也有着广泛的应用。DIA数据的蛋白质鉴定通常需要使用由数据依赖采集(DDA)得到的谱图数据库。然而该数据库含有的信息有限,为了在搜索过程中覆盖更多的蛋白质,目前采用深度学习模型的预测结果对该数据库进行补充。针对谱图预测任务,不同模型在不同数据集上的表现存在差异,且仅有少量模型展示了其在四维(4D)质谱数据上的性能。本文比较不同序列模型在4D-DIA血浆数据上的表现,提出了一个新的模型结构,该模型使用门控结合了双向长短期记忆网络(Bi-LSTM)和Transformer的特征,在较长的氨基酸序列上拥有更好的表现。Data Independent Acquisition(DIA)has developed rapidly in recent years and is widely applied in proteomic studies.A sample-specific spectral library from Data Dependent Acquisition(DDA)is used to identify proteins in DIA data.Because of the limitation of the DDA spectral library,predictions results of deep learning models are used to enrich the library,thereby covering more proteins during the searching.For spectrum prediction,different models perform inconsistently on different datasets,and only a small number of models have demonstrated their performance on 4D mass spectrometric data.The performance of different sequence models on 4D-DIA plasma data is compared and a new model structure is proposed.The model employs gate combining features extracted by Bi-LSTM and Transformer,which works better on long amino acid sequences.

关 键 词：数据非依赖采集技术 谱图预测 双向长短期记忆网络 TRANSFORMER 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]