Deep must read! 5 Empowering Fields of Artificial Intelligence+Synthetic Biology! High attention should be paid to further education, scientific research, and industry (protein engineering, customized biological scenarios)
Synthetic biology provides great prospects for scientists to solve important social problems.
Artificial intelligence (AI) and machine learning (ML) have significant prospects in providing the predictive capabilities required for synthetic biology and can be applied throughout the entire process of synthetic biology.
Protein design field: AI driven precision design
O'Neill et al. proposed a toolbox of signal peptide elements that can be used to enhance the production of biopharmaceutical proteins in Chinese hamster ovary cells, resulting in ML assisted carrier design that increases the yield of a range of products by ≥ 1.8 times compared to standard industry systems.
Khamwachirapithak et al. applied ML to optimize bioethanol production in brewing yeast at both room temperature and high temperature. In the initial round of experiments, ethanol production increased by 63% at 30 ° C, and the ML assisted workflow increased by another 7% at 40 ° C.
Dallo et al. constructed, transformed, and analyzed 76 bacterial strains, using high-density gRNA tiling, correlation analysis, and machine learning to investigate the design principles for the gRNA effectiveness of CRISPRi in Synechococcus sp. PCC 7002 strain of cyanobacteria.
A fascinating attempt at protein engineering: the powerful capabilities of machine learning
Marchal et al. developed an ML assisted protein engineering workflow based on Gaussian processes and applied it to improve glycosylation of coenzyme A carboxylase enzymes. Among the ten variants tested in vitro, nine were active, indicating a significant improvement in success rate compared to previous random mutations.
Bricio et al. developed a protein engineering tool called POET using genetic programming, demonstrating its utility in engineering a novel peptide with improved MRI contrast.
Golinski et al. used artificial neural networks to explore patterns in high-throughput exploitable datasets of ligand scaffold Gp2 protein variants, enabling direct visualization of adaptive landscapes and revealing evolutionary bottlenecks leading to sequence competition subgroups with different exploitability.
Just now, Singapore has heavily invested in synthetic biology! United Eight deep layout!
Advanced Machine Learning Architecture: Customizing AI for Specific Biological Scenarios nisonoff et al. designed a principled probabilistic approach that integrates biophysical knowledge into Bayesian neural networks, making the model more dependent on biophysical prior information. The author demonstrated this method on several examples, including GFP fluorescence and GB1 binding prediction.
Van Lent et al. solved the combinatorial pathway optimization problem and proposed a framework based on mechanistic dynamics models for optimizing and applying ML methods in iterative metabolic engineering, such as for the design build test learn cycle.
He et al. proposed a new interpretable model architecture called Nucleic Transformer, based on self attention and convolution, demonstrating its utility in several model tasks, including E. coli promoter classification, virus genome recognition, enhancer classification, and chromatin profile prediction.
Billion dollar fund! Shanghai will focus on investing in these three future industries!
