A poly-omics machine-learning method to predict metabolite production in CHO cells<strong> </strong>

Abstract

Abstract:The success of biopharmaceuticals as highly effective clinical drugs has 12 recently led industrial biotechnology towards their large-scale production. The 13 ovary cells of the Chinese hamster (CHO cells) are one of the most common 14 production cell line. However, they are very inefficient in producing desired 15compounds. This limitation can be tackled by culture bioengineering, but 16 identifying the optimal interventions is usually expensive and time-consuming. In 17 this study, we combined machine learning techniques with metabolic modelling 18 to estimate lactate production in CHO cell cultures. We trained our poly-omics 19 method using gene expression data from varying conditions and associated 20 reaction rates in metabolic pathways, reconstructed in silico. The poly-omics 21 reconstruction is performed by generating a set of condition-specific metabolic 22 models, specifically optimised for lactate export estimation. To validate our 23 approach, we compared predicted lactate production with experimentally 24 measured yields in a cross-validation setting. Importantly, we observe that 25 integration of metabolic predictions significantly improves the predictive ability 26 of our machine learning pipeline when compared to the same pipeline based on 27 gene expression alone. Our results suggest that, compared to transcriptomic-only 28 studies, combining metabolic modelling with data-driven methods vastly 29 improves the automatisation of cultures design, by accurately identifying optimal 30 growth conditions for producing target therapeutic compounds. 31