Predicting gene sequences with AI to study codon usage patterns

Abstract Selective pressure acts on the codon use, optimizing multiple, overlapping signals that are only partially understood. We trained artificial intelligence (AI) models to predict the codons given their amino acid sequence in the eukaryotes Saccharomyces cerevisiae and Schizosaccharomyces pombe and the bacteria Escherichia coli and Bacillus subtilis , to study the extent to which we can learn patterns in naturally occurring codons to improve predictions. We trained our models on a subset of the proteins, and evaluated their predictions on large, separate sets of proteins of varying lengths and expression levels. Our models significantly outperformed naïve frequency-based approaches, demonstrating that there are dependencies between codons that can be learned to better predict evolutionary-selected codon usage. The prediction accuracy advantage of our models is greater for highly expressed genes and it is greater in bacteria than eukaryotes, supporting the hypothesis that there is a monotonic relationship between selective pressure for complex codon patterns and effective population size. Also, in S . cerevisiae and bacteria, our models were more accurate for longer proteins, suggesting that the AI system may have learned patterns related to co-translational folding. Gene functionality and conservation were also important determinants that affect the performance of our models. Finally, we showed that using information encoded in homologous proteins has only a minor effect on prediction accuracy, perhaps due to complex codon-usage codes in genes undergoing rapid evolution. In summary, our study employing contemporary AI methods offers a new perspective on codon usage patterns and a novel tool to optimize codon usage in endogenous and heterologous proteins. Significance statement Can one predict codon sequences used by an organism to encode a given amino acid sequence? This is difficult, because there are exponentially many codon sequences that can encode the same amino acid sequence and evolution is stochastic. Indeed, codons frequencies vary, a phenomenon known as codon-bias, yet we improve upon frequency-based predictions using contemporary AI tools that learn complex patterns and capture interactions between codons. Because our predictions are tested fairly, on cases not seen during the training process, accurate predictions suggest that these learned patterns are not random, and may be related to the evolutionary process. Thus, studying where our predictions are more accurate, is expected to reveal novel insights related to the way evolution shapes coding regions.