Development of a deep learning model for predicting the effects of microbial culture media on gene expression levels

Abstract

This thesis is about computational biology and focuses on two types of research and development, including biodata-based computer modeling and systematic management strategies for chemical and biomolecular engineering software development. Biodata-based computer modeling relates to the development of DeepMGR, a new deep learning model that predicts the expression levels of genes in Escherichia coli under specific culture conditions. The level of gene expression in a cell can vary significantly depending on culture conditions, and it is very important in terms of the cost of industrial biotechnology to accurately and efficiently understand the gene regulation network of microorganisms under specific culture conditions. For this, DeepMGR uses the protein sequence of a specific gene, the composition of the compound in the medium in which the cell grows, and the carbon source required for metabolic activity as input information, and processes the input information with multiple perceptrons and convolutional neural networks. Moreover, as a second research topic, version control systems, and unit testing are very important technologies in software development. These software engineering technologies should be actively introduced into biochemical engineering-related programs, and as an example, this study applied them to the metabolic modeling program, GMSM. The results will help to derive in-depth knowledge by effectively processing biodata that is accumulating day by day.