Computer-aided modeling and analysis tools for bioprocess industrialization

Bioprocess development, scaleup and optimization is a significant undertaking. A lack of complete understanding of the underlying biological processes and a shortage of a trained workforce in this field has made the development process slow and inefficient, thus leading to continued reliance on non-renewable petroleum-based chemistries for fuels and chemicals. The overall goal of this project is to use machine learning and bioinformatics tools to provide a software product and a service offering to integrate and analyze large sets of biological data that provide actionable outputs to expedite bioprocess research, development, industrialization and optimization. We aim to provide clear-cut and reliable guidelines to the biotech industry and research community for building new hypotheses and carrying out efficient experimentation, by leveraging knowledge from the integration and analysis of all the relevant past data using a hybrid of cell metabolism modeling and machine learning. We will provide insights and the potential of unexplored spaces. Some of the information provided by us can be directly incorporated into improving existing bioprocess and designing novel controls. In Ph I, we will develop and demonstrate a workflow to build a machine learning-friendly, integrated and homogenized multi-omics database for E. coli K12. Machine learning-derived functional relationships and predictions using this database will enable us to get a non-steady state solution to the reconstructed dynamic genome-scale metabolic (dGEM) model, thus yielding a complete prediction of temporal transcript-, prote-, metabol- and flux- ome data for a given experiment condition. The integrated knowledge of the temporal genome-wide information will provide complete insight into the molecular level workings of the E. coli K12 cell factory in a given bioprocess, thus highlighting the opportunities for improvements. We will also do several useful abstractions from the simulation of the dynamic GEM model, that can be directly used to build hypotheses, design experiments, and test novel strategies for bioprocess development, scaleup and improvement. Once the workflow is developed and demonstrated for E. coli, which will complete our Ph I objective, we plan to build the same workflow for yeast with our funds. In Ph II, the workflow will be generalized to other cell factories, made more automated, have improved predictions by the addition of cell transcription and translation models, and will be packaged into a commercial software product and service offering. The proposal is in line with the goal of the DOE biological and environmental research program to gain a predictive understanding of biological processes for improved energy resilience and sustainability.