A machine learning poly-omics classifier to improve protein production in CHO cells

The success of biopharmaceuticals as highly effective clinical drugs has recently led industrial biotechnology towards their large-scale production. The ovary cells of the Chinese hamster (CHO cells) are one of the most common production cell line. However, they are very inefficient in producing desired compounds. This limitation can be tackled by culture bioengineering, but identifying the optimal interventions is usually expensive and time-consuming. In this project, we combined machine learning techniques with metabolic modelling to estimate lactate production in CHO cell cultures.
We integrated experimental data at gene level with data generated in silico via a publicly available genome-scale metabolic model of CHO cell within an integrated data-driven framework. We trained our poly-omics method using gene expression data from varying conditions and associated reaction rates in metabolic pathways, reconstructed in silico. We evaluated this approach through a computational validation based on cross-validation, estimating the average prediction error in general settings.
Importantly, we showed that metabolic predictions coupled with gene expression data can significantly improve estimations of lactate production based solely on gene expression. The proposed approach integrating machine learning and the state-of-the-art model of CHO cell metabolism was published and presented at IECM 2017 and highlighted on Phys.org.