2. Machine learning assisted synthetic biology

 Synthetic biology is to develop artificial cellular systems that humans can utilize for their own purpose. As all the functions in biological systems rely on enzymatic reactions and protein-protein interactions, manipulating proteins is a critical technology in synthetic biology. We are focusing on protein and nucleic acid engineering by sequencing and screening large-scale libraries in order to innovate synthetic biology. For instance, we characterize protein structures using public deep learning models, such as AlphaFold2, to identify regions to be engineered. Then, the screening library can be experimentally constructed with random mutagenesis, and selection pressure is applied for directed evolution. Since the sequence space of randomized DNA is astronomical and next generation sequencing (NGS) can't cover it, we are using machine learning to evaluate the functions of unseen protein sequences. Particularly, customized deep learning models trained by NGS data help increase the chances to find superior clones in the libraries. Currently, we are engineering enzymes and antibodies through in silico deep learning modeling and experimental validation.