Biomarker Discovery Study Of Psychological Stress Disorders Based On 4D-DIA Mass Spectrometry

Psychological stress disorders are illnesses with significant symptoms caused by a range of physical,psychological and social factors,usually accompanied by intense anxiety,tension,depression,anger,fear and other emotions,which seriously affect the normal productivity and life of the patient.In recent years,the prevalence of stress-related disorders has been increasing year on year.Current diagnoses of psychological stress disorders are mostly in the form of scales,questionnaires or interviews.Peripheral blood markers that can objectively diagnose psychological stress disorders have not yet been identified,and the scales are significantly non-specific,and the proficiency of psychiatrists and serious gaps also affect the diagnosis of the disorders.There is therefore an urgent need to discover protein biomarkers that can provide timely warning of psychological stress disorders.With the gradual rise of proteomics and mass spectrometry,more and more researchers are discovering protein biomarkers of diseases through mass spectrometry,but most of them are based on statistical learning methods,which cannot better analyse high-throughput,high-sensitivity mass spectrometry data and are not objective enough when validating the discovered biomarkers.To address these issues,this paper proposes a new biomarker discovery method for psychological stress disorders based on clinical 4D-DIA mass spectrometry data from a mental health centre in Shanghai.The method performs spectrum library construction through a deep learning approach.The biomarker discovery is also carried out by the proposed two-stage screening method and validated using the proposed target correlation algorithm.Finally,a biomarker discovery and application system for psychological stress disorders is designed and built based on the method of this paper.The research covers the following three main areas:1)Protein identification based on mass spectrometry data.The complex format of mass spectrometry data and the difficulty of identification have led to the selection of an LSTM-based spectral library construction method and a max DIA-based protein resolution method in this paper.In order to provide high quality quantitative protein data for downstream biomarker discovery services,a graph neural network-based protein missing value filling algorithm is proposed.The algorithm constructs a sample similarity network graph based on Euclidean distance,performs sampling of sample nodes in the graph by random walks,and performs graph reconstruction and edge prediction based on the learned feature vectors to obtain missing values for neighbourhood calculations.Finally,experiments on a public mass spectrometry dataset of depression demonstrate the effectiveness of this paper’s approach for the target task of mass protein identification.2)Differential protein analysis based on two-stage screening and confidence validation.This paper presents a two-stage screening and confidence validation-based differential protein analysis method.The screening part includes a primary screening of differential proteins based on a fusion method and a secondary screening of differential proteins based on interpretable SVM to maximize the discovery of biomarkers that can better distinguish between the two types of samples.Meanwhile,in order to validate the biomarkers obtained from the analysis,this paper proposes a target correlation scoring algorithm to obtain the disease and protein correlation scores by mining the large amount of a priori knowledge contained in the knowledge graph.Experiments are also conducted on a publicly available dataset of depression,demonstrating the superiority of the proposed method over current mainstream methods.3)Psychological stress disorder biomarker discovery and application system construction.This paper first analyzes data from the Psychological Stress Disorders Scale data,and then applies the method proposed in this paper to psychological stress disorders 4D-DIA mass spectrometry data to discover protein biomarkers such as actin,platelet factor 4,and kininogen-1.This paper also designs and implements a biomarker discovery and application system for psychological stress disorders based on the biomarker discovery needs of mass spectrometry data,which effectively improves the efficiency of mass spectrometry-based biomarker discovery.