Data Imputation in Merged Isobaric Labeling-Based Relative Quantification Datasets.

Area of Science:

• Proteomics
• Clinical Mass Spectrometry
• Bioinformatics

Background:

• Shotgun clinical proteomics using isobaric labeling (iTRAQ, TMT) generates large datasets.
• Missing values are common in these datasets, complicating statistical analysis and biomarker discovery.
• Accurate imputation is crucial for linking protein expression to clinical traits.

Purpose of the Study:

• To evaluate data imputation methods for mass spectrometry-based proteomics.
• To assess the effectiveness of microarray-derived imputation techniques in shotgun clinical proteomics.
• To identify robust methods for handling missing values in quantitative proteomic datasets.

Main Methods:

• Tested three data imputation approaches originally developed for microarray data.
• Applied methods to quantitative proteomic datasets using isobaric tags (iTRAQ and TMT).
• Evaluated performance on datasets with varying percentages of missing values.

Main Results:

• k-Nearest Neighbors (kNN) imputation demonstrated success in handling missing values.
• The kNN method effectively imputed data in datasets with up to 50% missing values.
• Imputation improved the integrity of proteomic data for downstream analysis.

Conclusions:

• k-Nearest Neighbors imputation is a viable strategy for addressing missing values in quantitative proteomics.
• This approach enhances the reliability of clinical proteomics studies.
• Successful imputation facilitates the identification of disease-associated proteins for diagnostic and prognostic applications.