Bridging the gap between transcriptome and proteome measurements identifies post-translationally regulated genes

Area of Science:

• Molecular Biology
• Systems Biology
• Computational Biology

Background:

• Messenger RNA (mRNA) abundance is often used as a proxy for protein levels, but this correlation is not universal due to post-transcriptional and post-translational regulation.
• High-throughput omic measurements provide valuable data, but bridging the gap between transcriptome and proteome levels remains a challenge.

Purpose of the Study:

• To develop a data-driven machine learning approach to predict protein concentrations from mRNA data in Saccharomyces cerevisiae.
• To identify mRNA-protein pairs potentially regulated by post-translational modifications.

Main Methods:

• Applied feature selection using sparsity-inducing regression (l1 norm regularization) to identify key predictive features.
• Reduced 37 features to 5: mRNA, ribosomal occupancy, ribosome density, tRNA adaptation index, and codon bias.
• Developed a linear predictor to model the relationship between selected features and protein concentrations.

Main Results:

• Achieved accurate prediction of protein concentrations with R² = 0.86 using the reduced feature set.
• Identified proteins with inaccurately predicted concentrations (outliers) as candidates for post-translational regulation.
• Demonstrated significantly higher evidence of post-translational modification in outliers compared to random subsets (P < 0.02).

Conclusions:

• Machine learning models can effectively bridge the transcriptome-proteome gap.
• Analysis of prediction outliers offers a novel method to uncover post-translational regulatory mechanisms.
• Outliers in machine learning models can yield significant biological insights.