Revealing determinants of translation efficiency via whole-gene codon randomization and machine learning

Abstract

It has been known for decades that codon usage contributes to translation efficiency and hence to protein production levels. However, its role in protein synthesis is still only partly understood. This lack of understanding hampers the design of synthetic genes for efficient protein production. In this study, we generated a synonymous codon-randomized library of the complete coding sequence of red fluorescent protein. Protein production levels and the full coding sequences were determined for 1459 gene variants in Escherichia coli. Using different machine learning approaches, these data were used to reveal correlations between codon usage and protein production. Interestingly, protein production levels can be relatively accurately predicted (Pearson correlation of 0.762) by a Random Forest model that only relies on the sequence information of the first eight codons. In this region, close to the translation initiation site, mRNA secondary structure rather than Codon Adaptation Index (CAI) is the key determinant of protein production. This study clearly demonstrates the key role of codons at the start of the coding sequence. Furthermore, these results imply that commonly used CAI-based codon optimization of the full coding sequence is not a very effective strategy. One should rather focus on optimizing protein production via reducing mRNA secondary structure formation with the first few codons.

CodonBERT: Large Language Models for mRNA design and optimization

“…The mRFP Expression dataset [23] profiles protein production levels for several gene variants in E. coli . The Fungal expression dataset [39, 43] includes coding sequences longer than 150 bp from a wide range of fungal genomes.…”

“…While mRNA sequence heavily influences cellular RNA stability [21,22], secondary structure can also impact mRNA stability in solution and modulate protein expression [13,23,24,25]. For example, replacing a codon with a synonymous codon can alter the local base-pairing interactions and affect nearby structural motifs [26,27].…”

From Context to Code: Rational De Novo DNA Design and Predicting Cross-Species DNA Functionality Using Deep Learning Transformer Models

“…For example, the initially transcribed region at the 5′-end of the 5′ UTR influences the abortive profile 21 ; the transcriptional and translational characteristics are shared across the promoter and the 5′ UTR sequences 22,23 . Second, when it comes to the CDS, no single design meets all the criteria used for generating codon-optimised variants 14,15,24 . Therefore, to capture these unique features, we design the 5′ regulatory sequences as an integrated DNA element and consider not just one, but a range of codon-optimised CDS variants in combination.…”

“…One such data set we used is from a recent publication 24 , which delved into the role of codon variations within CDSs and how these affect gene expression in E. coli . The constructs used in this research shared a fixed promoter and 5′ UTR, but had differing CDSs, generated from a synonymous codon-randomised library using a red fluorescent protein.…”

“…Second, when it comes to the CDS, no single design meets all the criteria used for generating codon-optimised variants 14,15,24 . Therefore, to capture these unique features, we design the 5 ′ regulatory sequences as an integrated DNA element and consider not just one, but a range of codon-optimised CDS variants in combination.…”