Custom 3D printed hardware and a deep learning predictive model for creating transcriptional fingerprints of bioactive molecules in bacteria
About
Background One of the most challenging aspects of drug discovery is identifying how prospective lead molecules work. This mechanistic determination is a time-consuming component of modern drug discovery pipelines, and streamlining it to a single assay is an immense technological advance. Here, researchers at McMaster University showcase their custom 3D printed hardware, that is used to create transcriptional fingerprints of bioactive molecules in bacteria. Amassing a substantial database of known antimicrobials, they utilize a deep learning model capable of predicting the mechanism of action of unknown chemical matter. Further, full timecourse gene expression is captured on top of the mechanistic predictions. Technology Overview The technology presented here exists in two parts, 3D printed acquisition hardware, and a deep learning predictive model using a comprehensive database. Small 3D-printed fluorescence imaging boxes (PFIboxes) are used to capture gene expression patterns using a fluorescence reporter library. Bacteria have unique responses to unique chemical matter, and comparing response fingerprints from unknown chemicals, to a databased of known chemicals, offers a unique means of mechanistic prediction. Response fingerprints are acquired over time, and describe transcriptional responses of bacteria, to bioactive chemicals.
Key Benefits
• Rapid transcriptional profiling of any chemical • Mechanistic predictions based on deep learning algorithm • Complete time course gene expression assays further inform of mechanism of action
Applications
PFIbox technology aims to streamline mechanism of action predictions for bioactive chemicals. While the primary focus is on the drug discovery space (in particular antimicrobials), any chemical or physical stress can be provided to obtain a complete bacterial response profile.