Accuracy and data efficiency in deep learning models of protein expression

Abstract

Recent progress in laboratory automation has enabled rapid and large-scale characterization of strains engineered to express heterologous proteins, paving the way for the use of machine learning to optimize production phenotypes. The ability to predict protein expression from DNA sequence promises to deliver large efficiency gains and reduced costs for strain design. Yet it remains unclear which models are best suited for this task or what is the size of training data required for accurate prediction. Here we trained and compared thousands of predictive models of protein expression from sequence, using a large screen of Escherichia coli strains with varying levels of GFP expression. We consider models of increasing complexity, from linear regressors to convolutional neural networks, trained on datasets of variable size and sequence diversity. Our results highlight trade-offs between prediction accuracy, data diversity, and DNA encoding methods. We provide robust evidence that deep neural networks can outperform classic models with the same amount of training data, achieving prediction accuracy over 80% when trained on approximately 2,000 sequences. Using techniques from Explainable AI, we show that deep learning models capture sequence elements that are known to correlate with expression, such as the stability of mRNA secondary structure. Our results lay the groundwork for the more widespread adoption of deep learning for strain engineering across the biotechnology sector.

Deep Learning Concepts and Applications for Synthetic Biology

“…It will also be helpful to understand the trade-offs between performance and complexity for specific applications, and studies aimed at exploring this are likely to be valuable. For instance, Nikolados et al 103 compared models of increasing complexity to compare their ability to predict protein expression from DNA sequence. Finally, the amount of data available also places important constraints on whether deep learning approaches are appropriate, as deep models require large training sets.…”

Improving the generalization of protein expression models with mechanistic sequence information

“…The dataset employed in this study was originally reported in [22]. The data reorganized in a form suitable for machine learning analyses can be found in the Zenodo [50].…”

15 years of microbial biotechnology: the time has come to think big—and act soon

“…Metabolic models of a whole range of biological systems have been particularly useful (Gudmundsson and Nogales, 2021), and they hold a promise for guiding implementation of a suite of biotechnological interventions of the sort advocated below (García‐Jiménez et al ., 2021). Finally, of considerable interest to microbial biotechnology is the recent development of ML platforms to support roadmaps for engineering heterologous gene expression workflows (Reis and Salis, 2020; Nikolados et al ., 2021). While these approaches currently fail to provide mechanistic insights, they may solve practical problems that are not comprehensible with the principles known at the present time.…”