Left-Censored Missing Value Imputation Approach for MS-Based Proteomics Data with GSimp

Area of Science:

• Biochemistry
• Proteomics
• Bioinformatics

Background:

• Missing values are common in mass spectrometry (MS)-based omics, particularly proteomics.
• Values below the limit of detection or quantification (LOD/LOQ) are often missing not at random (MNAR).
• MNAR data can cause biased statistical analysis and hinder downstream applications.

Purpose of the Study:

• To introduce GSimp, a novel imputation method for MS-proteomics data.
• To address the challenge of missing not at random (MNAR) values in omics studies.
• To provide a tool for accurate statistical estimation in MS-based omics.

Main Methods:

• Development of GSimp, a Gibbs sampler-based imputation approach.
• Focus on imputing left-censored missing values specific to MS-proteomics.
• Detailed explanation of MNAR principles and GSimp's application.

Main Results:

• GSimp effectively handles left-censored missing values in MS-proteomics datasets.
• The method is designed to mitigate bias caused by MNAR data.
• Facilitates more reliable downstream analyses in omics research.

Conclusions:

• GSimp offers a robust solution for imputing MNAR values in MS-proteomics.
• Accurate imputation is crucial for unbiased statistical inference in omics.
• This approach supports the advancement of MS-based omics studies.