Unraveling the complex relationship between mRNA and protein abundances: a machine learning-based approach for imputing protein levels from RNA-seq data.

The correlation between messenger RNA (mRNA) and protein abundances has long been debated. RNA sequencing (RNA-seq), a high-throughput, commonly used method for analyzing transcriptional dynamics, leaves questions about whether we can translate RNA-seq-identified gene signatures directly to protein changes. In this study, we utilized a set of 17 widely assessed immune and wound healing mediators in the context of canine volumetric muscle loss to investigate the correlation of mRNA and protein abundances. Our data reveal an overall agreement between mRNA and protein levels on these 17 mediators when examining samples from the same experimental condition (e.g. the same biopsy). However, we observed a lack of correlation between mRNA and protein levels for individual genes under different conditions, underscoring the challenges in converting transcriptional changes into protein changes. To address this discrepancy, we developed a machine learning model to predict protein abundances from RNA-seq data, achieving high accuracy. Our approach also effectively corrected multiple extreme outliers measured by antibody-based protein assays. Additionally, this model has the potential to detect post-translational modification events, as shown by accurately estimating activated transforming growth factor β1 levels. This study presents a promising approach for converting RNA-seq data into protein abundance and its biological significance.