Label-free Quantitative Algorithm Based On Glycopeptides Mass Spectrometry Data

Glycopeptide Mass Spectrometry(MS)data-oriented label-free quantitative is one of important components in bioinformatics.MS data cannot be utilized to achieve glycopeptide label-free quantitative directly because of its complexity.However,the increasingly development of computer technology makes it possible to quantify label-free glycopeptide based on MS data.In order to solve the problems of quantifying glycopeptide by computer-aided method based on MS data,this thesis conducts following research:(1)To solve the mistake of the peak clustering extraction because of peaks overlapping,a new 3D-based Multiple-charges Overlap Peaks Extraction algorithm(3D-MOPE)combining with isotope distribution rules is proposed.The algorithm firstly solves isotope peaks overlapping in first stage of MS(MS1),and then it separates and extracts overlap peaks in the three-dimensional space which is constructed by combining retention time,mass/charge ratio and ions' intensity of glycopeptide ions.The experiment results show that 3D-MOPE algorithm is 16%,13% and 2% higher in terms of peak cluster extraction accuracy than Max Peak,Pep Quan and Max Quant algorithms,respectively.(2)A new Accurate Peaks Alignment algorithm Fusing Cross Peaks Features(APAFCPF)is proposed to solve the peak alignment problem in MS data.The algorithm contains two procedures: a time-weighted global coarse granularity alignment model based on random disturbance is proposed to alleviate effect by outliers when establishing the relationships between different MS data.What's more,a cross fusion formula on peak features and a dynamic programming algorithm are designed to align peak clusters accurately.The experiment results show that APAFCPF algorithm are 0.08,0.14 and 0.06 higher than IPTW,SFA-MS and DTW algorithms in terms of F1 on DS3 dataset,respectively.(3)In order to improve the accuracy of quantitative results,a glycopeptide mass spectrometry data-oriented label-free quantitative algorithm(GpMS-LFQ)is proposed.The abundance of glycopeptides is calculated based on peak cluster information by combining the previous work in chapter2 and chapter3.Then,a multi-module normalization method is designed to solve systematic errors,finally,the quantitative results of glycopeptide are obtained.Three experiments are conducted to evaluate different glycopeptide quantitative models,the results show that GpMS-LFQ is 0.51,0.27,0.20 higher than Max Quant algorithm and 0.24,0.02,0.08 higher than Byonic algorithm in terms of F1,respectively.In addition,a new glycopeptide label free quantitative tool(Gp QT)is developed based on proposed algorithm.In summary,the thesis focuses on glycopeptide MS data-oriented label-free quantitative algorithm,which can provide the basis for disease diagnosis and biomarker discovery.