Antimicrobial resistance (AMR) is a major global threat, causing an estimated 700,000 deaths per year. One of the main ways that AMR genes spread among bacteria is through conjugation. Therefore, it is crucial to develop conjugation inhibitors (COINs) to combat AMR. Shared characteristics between known COIN molecules are a hydrophobic tail, unsaturations, and a polar head. Following these characteristics and accounting for the efficacy of 2-hexadecynoic acid and tanzawaic acids (TZAs) low toxicity, we designed a new set of TZA analogs.

In this work, we use AI, bioinformatics, and physics-based computational tools to study the efficacy of this new set of COIN molecules to target the type IV secretion system (T4SS), which is vital for bacterial conjugation. Our research involves predicting the structure of T4SS, identifying COIN binding pockets, analyzing binding poses of our analogs, and evaluating their relative binding affinities.

To predict the experimentally unknown T4SS structure, we compare structures from AI (Alpha Fold and RoseTTA Fold) and homology (Phyre2) protein structure prediction. We then run molecular dynamics (MD) simulations to assess the predicted structure's stability and the shape of possible binding pockets.

The location of binding pockets and the possible binding pose are determined using the DOCK software and MELD (Modeling Employing Limited Data) accelerated MD (MELDxMD) simulations. DOCK allows us to identify a set of possible binding poses for our COIN molecules. MELDxMD is then used to rank and refine these predictions. MELD is an MD accelerator that leverages external information to reduce the conformational space available to the system, thus allowing the simulation to converge to an answer faster. In this case, the external information comes from the set of possible dock poses, and MELDxMD identifies a unique binding pose for each COIN molecule.

Finally, we employ MELDxMD again to perform competitive simulations between pairs of COIN molecules. In this part of the project, we use the MELD framework to competitively bind two molecules to the same TSS4 binding pocket. We identify the best ligand from the relative probability of the two molecules to be in the bound state. Subsequently, the optimal molecules will be selected for synthesis and biological evaluations to identify the mechanisms of action of these new COIN molecules.

Renata Rezende Miranda

Appearances