Semisupervised Gaussian Process for Automated Enzyme Search

• DOI: 10.1021/acssynbio.5b00294
• Publication Date: 2016-03-23T18:14:41Z
• Authors: Joseph Mellor; Ioana Grigoras; Pablo Carbonell; Jean-Loup Faulon
• Keywords: Synthetic Biology

Synthetic biology is today harnessing the design of novel and greener biosynthesis routes for production added-value chemicals natural products. The pathways often requires a detailed selection enzyme sequences to import into chassis at each reaction steps. To address such requirements in an automated way, we present here tool exploring space enzymatic reactions. Given provides probability estimate that catalyzes reaction. Our first considers similarity known biochemical reactions with respect signatures around their centers. Signatures are defined based on chemical transformation rules by using extended connectivity fingerprint descriptors. A semisupervised Gaussian process model associated similar then estimate. uses information about both providing These estimates were validated experimentally application newly identified metabolite Escherichia coli order search enzymes catalyzing its Furthermore, show several pathway examples how ability assign potential assist bioengineering applications, experimental validation our proposed approach. best knowledge, approach processes dealing biological chemicals, use framework also context machine learning applied bioinformatics. However, catalyze depends affinity between substrates enzyme. This generally quantified Michaelis constant KM. Therefore, demonstrate regression predict KM given substrate-enzyme pair.