In the standard model of molecular biology, messenger RNA (mRNA) is treated as a faithful intermediary—a transcript that directly dictates the concentration of functional proteins. However, a growing body of evidence suggests that this "Central Dogma" view is often a poor predictor of cellular reality. Across diverse biological systems, the correlation between mRNA levels and protein abundance is frequently as low as 40% to 60%.
This discrepancy is not merely technical noise; it represents a sophisticated layer of biological regulation. Researchers are now moving beyond transcriptomics to model the "decoupling" of these two layers, recognizing that the proteome—the actual machinery of the cell—is governed by post-transcriptional and post-translational "switches" that mRNA data alone cannot capture.
The Failure of the Transcriptomic Proxy
For decades, transcriptomics has served as a high-throughput proxy for cellular function because mRNA is easier to sequence than proteins are to identify. But as Ronald Jansen and Dov Greenbaum noted in their foundational study, Relating Whole-Genome Expression Data with Protein-Protein Interactions (Genome 2002), even highly synchronized mRNA profiles may not result in stoichiometric protein concentrations.
This decoupling occurs because the cell exerts tight control over the "life cycle" of a protein after the gene has been transcribed. Two primary mechanisms drive this:
- Translation Efficiency: The rate at which ribosomes turn a single mRNA molecule into protein can vary wildly.
- Degradation Rates: Proteins have vastly different half-lives; some are stable for days, while others are tagged for destruction (ubiquitination) within minutes of synthesis.
As a result, relying on mRNA as a proxy for cellular communication or metabolic state—as highlighted by Xin Shao and Xiaoyan Lu in New avenues for systematically inferring cell-cell communication: through single-cell transcriptomics data (Protein 2020)—can lead to an incomplete or even misleading map of the signaling environment.
Post-Translational Switches and Functional Activity
Even when protein abundance does correlate with mRNA, the functional activity of those proteins may not. Post-translational modifications (PTMs) act as rapid chemical switches that alter a protein's behavior without changing its concentration.
In Dynamic Responses of Barley Root Succinyl-Proteome to Short-Term Phosphate Starvation and Recovery (Frontiers in Plant Science 2021), Juncheng Wang demonstrated that plants use lysine succinylation to rapidly reprogram metabolism during stress, a process invisible to transcriptomic analysis. Similarly, Luning Liu found that in pomelo fruit, a "tug-of-war" exists where increased transcript levels of antioxidant genes are offset by post-translational degradation, leading to physiological failure despite a "healthy" genetic response.
This divergence is particularly critical in disease. In cancer, Andreas Mock has shown that mRNA levels often fail to reflect the activity of survival factors like BCL-XL or signaling pathways like Notch, which are heavily regulated by protein cleavage and localization.
Translating the Transcriptome
To bridge this gap, researchers are developing computational "Rosetta Stones" that can translate abundant transcriptomic data into predicted proteomic profiles. These models aim to capture the "missing regulation" that occurs between the gene and the functional protein.
- scTranslator: Linjing Liu and colleagues developed A pre-trained large generative model for translating single-cell transcriptome to proteome (2023), a Transformer-based model that leverages both bulk and single-cell data to infer the levels of over 14,000 proteins from standard RNA sequences.
- scLinguist: Using a sub-quadratic "Hyena" architecture, Zhaoyu Fang and Jiayuan Ding created scLinguist: A pre-trained hyena-based foundation model for cross-modality translation in single-cell multi-omics (2025) to perform zero-shot translation of RNA to protein, effectively filling the gap when expensive paired multi-omic data is unavailable.
The Genetic Architecture of Protein Levels
The decoupling is also evident at the genetic level. By mapping protein quantitative trait loci (pQTLs), researchers like Bing He have discovered that many genetic variants influencing protein abundance do not affect mRNA levels. In Genome-wide pQTL analysis of protein expression regulatory networks in the human liver (BMC 2020), He identified thousands of regulatory variants that were invisible to previous mRNA-based studies, revealing that a significant portion of human genetic variation acts specifically on the post-transcriptional machinery.
This research direction suggests that the future of systems biology lies not in choosing between RNA or protein, but in modeling the complex, non-linear relationship between them. By treating the discrepancy between the two as a feature rather than a bug, we can better identify the true drivers of development, regeneration, and disease.