PROTRIDER: protein abundance outlier detection from mass spectrometry-based proteomics data with a conditional...

Area of Science:

• Proteomics
• Genomics
• Computational Biology

Background:

• Detecting gene regulatory aberrations is crucial for rare disease diagnostics and tumor characterization.
• Existing methods for RNA expression outlier detection are abundant, but protein expression outlier detection is underdeveloped.

Purpose of the Study:

• To propose and assess modeling approaches for calling protein expression outliers from mass spectrometry data.
• To develop a robust method for identifying aberrantly expressed proteins.

Main Methods:

• Conditional autoencoders were employed to control for hidden confounders and technical covariates while modeling missing values.
• The Student's t-distribution was utilized to model heavy-tailed differences between experimental and fitted log-transformed intensities.
• PROTRIDER was compared against baseline methods including Z-scores, PCA, and Isolation Forests.

Main Results:

• PROTRIDER demonstrated superior performance compared to baseline approaches.
• The method revealed significant enrichments of AlphaMissense pathogenic variants in protein expression outliers.
• Conditional autoencoders effectively handled confounders and missing data in proteomics datasets.

Conclusions:

• PROTRIDER provides a reliable method for identifying aberrantly expressed proteins using mass spectrometry data.
• The findings support the application of PROTRIDER in rare disease diagnostics and cancer proteomics.
• The study highlights the benefits of advanced modeling techniques for analyzing complex proteomics data.