Missing Data Imputation And Normalization For The Mass Spectrometry Quantification Data

In recent years,the development of high resolution LC-MS/MS has promoted the evolution of shotgun proteomics technology,which let scientists go further in biomarker discovering with corresponding statistical methods.But during the proteomics data analysis,two challenges in proteomics quantitative results tend to reduce its reliability and accuracy:one is the massive missing values always ignored by people,the other is the serious batch effect resulted from different experiments or platforms.Here,we present a unique imputation method called KDI(Knowledgeable-based Data Imputation)dealing with the first challenge.Our results prove that the imputation result of KDI is with higher accuracy and more stable than the KNN and Stepwise approach.Facing the other challenge,we try to apply the microarray data normalization methods(Quantile,TMM and sva)to the proteomics quantification result.Moreover,we creatively combine the KDI with these three normalization method and find the best preprocessing step(the combination of imputation and normalization)solving the two challenges mentioned above.Our results show that the preprocessing step(the combination of KDI and sva)can dramatically reduce the missing value and batch effect,which can significantly improve the biological analysis.Thus,our strategy provides a convenient data preprocessing scheme to the related scientists.