Application Research Of Intelligent Computing In Protein Analysis

Protein is an important component of every cell in the human body.It is the main performer of life activities and regulates the entire life process.Proteomic assay and data analysis can monitor the changes in the type and quantity of proteins when the physiological state of body changes,helping us learn the mechanism and the state of body,which is of great significance.In addition to playing an important role in the hu-man body,protein can also help increase the efficiency of production in the chemical,medical,pharmaceutical and other fields.However,due to the nature of enzymes(pro-teins),they are generally not directly used in complex chemical systems in vitro,which is caused by instability of the enzyme,harsh reaction conditions,or other factors.At present,the characteristics of enzymes are mainly improved by directed evolution,and the computational method is an important tool for improving the efficiency and usability of directed evolution to produce proteins.This paper mainly studies the computational methods required in the data analysis after proteomic assay and the construction of homologous protein library in directed evolution.The main works include:(1)this study proposed a MSCNN(Mass Spec-trometry Convolutional Neural Network)preprocessing model based on convolutional neural network for the untargeted analysis of DIA(Data Independent Acquisition)data.Compared with the method in DIA-Umpire,our model could make full use of the fea-tures of peptides in MS1(First stage of Mass Spectrometry)and MS2(Second stage of Mass Spectrometry)and more effectively remove the noise peaks in DIA data,which is of great help for subsequent analysis.(2)This study respectively proposed a uniform extraction preprocessing algorithm and a protein quantification model based on least square error for the preprocessing and protein quantification problems in the targeted analysis of DIA data on the RTF platform,whose effects were respectively demonstrated by the change of the TIC(Total Intensity Current)and the relative concentration ratio of two samples.(3)This study proposed a multi-objective optimization solution for the reverse translation process in the construction of homologous protein library.We mod-eled the reverse translation process as a multi-objective problem and solved it with the classic multi-objective optimization algorithm NSGA-?.Finally,the simulation exper-iment of fluorescent protein parent sequences showed that our method achieved a great performance improvement compared with the conventional segmentation algorithm.