Session: Posters:

The global impact of Tuberculosis (TB), which causes ~10 million new infections and ~1 million deaths annually, is exacerbated by the lack of effective therapeutics and the emergence of multi-drug resistant strains of Mycobacterium tuberculosis (Mtb). Historically, clinically used TB drugs and late-stage candidate compounds have been discovered by whole-cell phenotypic screening campaigns, which account for the formidable permeability barrier presented by the mycobacterial cell wall. While this approach has proven successful, it is logistically demanding to execute an HTS of large compound libraries in a Biosafety Level 3 facility. Subsequent identification of the target and mechanism of action of hits, which is important for structure-based hit-to-lead optimization, presents a second daunting challenge. As a result, there remains an urgent need for new anti-TB drugs and many genetically validated drug targets in Mtb remain to be exploited. In this project, we sought to harness the power of AtomNet®, the world’s first deep convolutional neural network for structure-based drug discovery developed by Atomwise, Inc., to accelerate the discovery of novel scaffolds and lead compounds for TB drug development. We conducted virtual screens of 2.5 million commercially available compounds for predicted inhibitors of FusA1, an essential ribosomal accessory factor involved in ribosome translocation, and LdtMt2, an L,D-transpeptidase responsible for unusual 3’-3’cross-linking of the peptidoglycan layer of the cell wall. Primary screening of 94 compounds for each target against a bioluminescent reporter strain of Mtb revealed 10 novel scaffolds (5 per target) with dose-dependent whole-cell antimycobacterial activity. In addition to establishing structure-activity-relationship (SAR) profiles for prioritized scaffolds by assaying of analogs, demonstration of target-specific activity is also underway using both microbiological and biochemical strategies. The discovery of first-in-class inhibitors of underexploited targets with potent bactericidal activity against Mtb would validate this AI-driven strategy for the rapid, more cost-effective development of novel TB therapeutics.

Rashmi Gupta

Appearances