In metabolic engineering, one goal is to target modification of cell metabolism to maximize the production of a specific compound. To increase our production and realize optimization in metabolic engineering. Tools and algorithms are needed to efficiently help us design experiment through learning systems pattern.For these purposes, we developed in vivo model(HAWNA, PPIBoost) and in vitro model(two layer logistics growth model) to integrate gene expression profile, experimental data, protein sequence data and Go annotation data to help us to design a modification strategy to improve organism system.
Our in vivo model, which integrates literature data and public dataset to infer genetic regulatory network with protein-protein interaction network. Enable our team to find genes may have possible regulatory function with our cellulose synthetic pathway. Based on statistical model analysis, it will help our wet lab to design optimize strategy to promote cellulose production.
Our in vitro model, we used experimental data to fit two layers logistic growth model. In training process, our model back propagates microbial cell growth information and combines with culture condition variable information to estimate parameters. In predicting process, we input new culture conditions and our model can return a cell growth curve automatically. We used this model to optimize culture condition to promote biomass.
The model map below is a rough outline of our model, click on it to explore our statistical model in detail.