Our lab member Navneet Rai presented a poster describing the application of Optimal Experimental Design to Omics Experimentation at the Synthetic Biology: Engineering, Evolution & Design (SEED), New York.

Poster Title: Application of Optimal Experimental Design to Omics Experimentation

Abstract: Recent advances in high-resolution omics technologies and high-performance computing have paved the way for the development of powerful predictive models in biology. How can one design their experiments, so that the data gather can maximally increase the model performance? To answer this question, we have developed an optimal experimental design (OED) framework to identify a set of experiments that maximize prediction performance, given gene expression (GE) data. OED methods have been used in the past from aerospace engineering to protein design, but this is the first time they are used in omics prediction. In this study, we expose Escherichia coli populations to antiseptics and antibiotics combinations. We train gaussian process models of gene expression with partial data and then introducing an OED method to infer what experiments we should perform so that the models become more accurate in predicting gene expression in novel conditions. We then compare OED versus random and expert sampling, for picking most informative experiments until all 40 experiments are performed. Our results show that the prediction error is significantly lower when OED is used, and the same prediction accuracy is used with 40% less data points compared to expert sampling. Our work is a step towards automated experimentation for traversing the vast experimental space as it is the case of omics biology.