Analysis of mRNA expression and protein abundance data: an approach for the comparison of the enrichment of features in the cellular population of proteins and transcripts

Dov Greenbaum, Ronald Jansen, Mark GersteinBioinformatics 2002 · DOI · OpenReview

Abstract

This paper introduces a unified mathematical framework to reconcile the weak correlations typically observed between mRNA levels and protein abundance in high-throughput biological data. By shifting the analytical focus from individual gene-to-gene correlations to the "enrichment" of broad biological features—such as amino acid composition, molecular weight, and cellular function—the authors demonstrate that the transcriptome and the "translatome" (the population of proteins weighted by their cellular abundance) are significantly more synchronized than previously recognized. This approach provides a robust method for integrating noisy, disparate datasets to reveal global biological trends.

Introduction

In the early era of high-throughput biology, researchers faced a significant paradox: while the Central Dogma suggests that mRNA levels should serve as a reliable predictor of protein levels, experimental data often showed weak and conflicting correlations. This discrepancy was attributed to two primary factors: biological complexity (including post-transcriptional regulation, varying translation rates, and protein degradation) and inherent experimental noise and bias in techniques like SAGE, DNA microarrays, and 2D-gel electrophoresis.

This study addresses these challenges by developing a formalism to integrate diverse datasets and understand the global characteristics of highly expressed genes and proteins in the model organism Saccharomyces cerevisiae.

Methodology

The "Translatome" and Enrichment Formalism

The authors introduce the concept of the translatome, defined as the cellular population of proteins where each gene is weighted by its actual abundance. This is distinct from the "proteome," which often refers to the static list of all possible proteins encoded by the genome.

To compare these populations, the study utilizes an enrichment formalism. Instead of gene-by-gene comparisons, the researchers calculate the weighted average of specific features ($F$) within a population. Enrichment ($\Delta$) is then defined as the relative difference in a feature's prevalence between two populations (e.g., comparing the translatome to the genome).

Conceptual framework for the 'translatome' and the enrichment formalism used to compare genomic, transcriptomic, and proteomic populations.

Data Integration and Bias Correction

To handle diverse datasets, the authors employed iterative scaling using non-linear regression ($Y = AX^B$) to normalize mRNA expression data across different experimental platforms. Furthermore, to isolate biological signals from sampling bias—particularly since protein studies like 2D gels only capture a small subset of the genome—the framework explicitly restricts comparisons to identical gene sets across different data types.

Results

A key finding of the study is that the cell is highly economical in its protein production. The translatome is significantly enriched in low molecular weight proteins; the average molecular weight per open reading frame (ORF) is approximately 20% lower in the translatome than in the genome. This suggests an evolutionary pressure favoring shorter, less energetically expensive proteins for high-volume production.

Furthermore, highly expressed proteins exhibit distinct compositional shifts. They are enriched in small amino acids such as Valine, Glycine, and Alanine. Statistically, these enrichments were confirmed through randomization controls ($p < 10^{-3}$), proving they are not the result of random noise.

Enrichment of specific amino acids in the yeast translatome compared to the genome.

Functional and Localization Synchronization

Despite weak correlations at the single-gene level, the transcriptome and translatome show substantial agreement when viewed through functional categories. Both populations are consistently enriched in proteins related to protein synthesis, cell structure, and energy production. Additionally, there is a massive enrichment of cytoplasmic proteins in the translatome relative to nuclear ones, reflecting the high cellular demand for metabolic machinery in the cytosol.

Discussion and Conclusion

The study demonstrates that "averaging beats noise." While individual gene measurements are prone to high error rates, clear biological trends emerge when data is aggregated into broad feature categories. The results indicate that the depletion of molecular weight in highly expressed products occurs primarily during transcription, with translation providing a secondary, detectable bias toward shorter proteins.

While limited by the relatively small protein samples available from 2D-gel data at the time, this framework provides a foundation for future proteomic research. As mass spectrometry and other technologies improve, this enrichment-based approach can be used to identify "regulatory outliers"—genes where mRNA and protein levels diverge significantly—which likely represent critical biological control points.

Summary updated

BibTeX
@article{greenbaum2002analysis,
  author = {Dov Greenbaum and Ronald Jansen and Mark Gerstein},
  title = {{Analysis of mRNA expression and protein abundance data: an approach for the comparison of the enrichment of features in the cellular population of proteins and transcripts}},
  journal = {Bioinformatics},
  year = {2002},
  doi = {10.1093/bioinformatics/18.4.585},
}
Figure