Peptide tags play an important role as chemical biology tools, which are genetically grafted onto a recombinant protein to allow specific enzymatic or chemical modifications. Post-translational enzymes have been utilized to site-specifically modify peptide tags for protein labeling purposes. The Burkart laboratory introduced 4’-phosphopantethienyl-transferase (PPTase), a post-translationally modifying carrier protein in natural product pathways, as a versatile tool that can site-specifically modify conserved serine residue. These PPTases consist of two distinct families, Sfp and AcpS type, which are both promiscuous with respect to the range of reporter groups that are delivered to the target serine residue. We developed a hybrid computational and biochemical method to discover orthogonal peptide substrates for both classes of PPTases. This method employs an iterative machine learning process where experimental data is accumulated into a mathematical algorithm that selects potential peptide substrates to be tested experimentally. Peptide substrates predicted by the algorithm were tested for their labeling viability and the information was fed back into the algorithm to refine further selections. In this manner, we have demonstrated that machine learning can be leveraged to guide peptide optimization for specific biochemical functions not immediately accessible by biological screening techniques, such as phage display and random mutagenesis.

Discovering <i>de novo</i> peptide substrates for enzymes using machine learning

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.

Discovering de novo peptide substrates for enzymes using machine learning