(180c) Antibiotics Molecular Design Using Artificial Bee Colony Optimization

Hartenstein, M., University of Kansas
Gaumer, R., The University of Kansas
Camarda, K., University of Kansas
Antibiotics Molecular Design Using Artificial Bee Colony Optimization

Authors: Shweta Mapari, Matthew Hartenstein, Rex Gaumer, Kyle Camarda

In a recently published WHO report, a list of bacteria which pose the greatest threat to human health was given. The ESKAPE list also defines pathogens which have become resistant to multiple drugs and for which there exists a critical need for alternate therapies. Acinetobacter baumannii, which is on both of these lists, is a pathogen causing infections such as pneumonia and bacteremia. The goal of this research is to apply Computer Aided Molecular Design (CAMD) techniques to develop a novel antibiotic agent to treat this bacterial infection.

In order to treat A. baumannii, two molecular design problems were formulated, one for a beta lactam antibiotic and the other for a fluoroquinolone compound. CAMD is used to find analogous structures of small molecules which have reduced values of the Minimum Inhibitory Concentration (MIC) and toxicity for beta lactam and flouroquinolone drugs. We used piperacillin, amoxicillin, and ticarcillin as starting points for the beta lactam drugs and paufloxacin as a starting point for the fluoroquinolone drugs.

This work applies topological molecular descriptors to develop Quantitative Structure Activity Relationships (QSAR) which are employed to predict MIC values and toxicity. These expressions are combined with structural feasibility constraints to formulate an optimization problem. Because of the large number of potential chemical structures and the uncertainty in the structure-property correlations, stochastic algorithms are preferred to solve the resulting MINLP. One stochastic method which has shown promise to solve these problems is the Artificial Bee Colony algorithm, which relies on principles of swarm intelligence to find near-optimal solutions efficiently. The Artificial Bee Colony algorithm described in this paper is used to derive solutions which serve as lead compounds for a narrowed search for novel antibiotics, which would employ molecular simulation and experimental validation.