Assessment of label-free quantification and missing value imputation for proteomics in non-human primates

Area of Science:

• Proteomics
• Bioinformatics
• Comparative Genomics

Background:

• Label-free quantification (LFQ) relies heavily on data acquisition and downstream processing, including database quality.
• Non-human primate (NHP) proteomics faces challenges due to limited genomic annotation, impacting protein and Post-Translational Modification (PTM) discovery.
• Missing data in LFQ analyses hinders biological interpretation and statistical power.

Purpose of the Study:

• To enhance protein and PTM discovery in NHP by utilizing the MetaMorpheus proteomics search engine.
• To evaluate various imputation methods for accurate inference of missing protein abundance values in NHP LFQ data.

Main Methods:

• Employed the MetaMorpheus proteomics search engine to maximize protein and PTM identification in NHP samples.
• Conducted a comparative analysis of different imputation methods, including Generalized Ridge Regression (GRR), Random Forest (RF), local least squares (LLS), and Bayesian Principal Component Analysis (BPCA).
• Utilized a generic approach for missing data imputation without distinguishing the source of missingness.

Main Results:

• Identified 1622 proteins and 10,634 peptides, including 58 PTMs, in NHP brain frontal cortex samples.
• Only 293 proteins were quantified across all samples, highlighting the necessity of imputation.
• Demonstrated that imputation methods like GRR, RF, LLS, and BPCA accurately estimate missing protein abundance values.

Conclusions:

• The study provides strategies for improving LFQ in NHP proteomics.
• MetaMorpheus and advanced imputation techniques offer solutions to database limitations and missing data challenges in NHP LFQ.