UC Irvine

The research described in this work consists of two distinct fields of study. The first part of the work describes the design, synthesis, and testing of small molecule azole synergizers, while the second part describes work on computational and machine learning methods towards reaction prediction.

Five isoquinolones and phthalazinones were previously shown to increase the activity of fluconazole against Candida albicans. A dozen analogues were synthesized and one compound, 2.41, was shown to have activity at 1 nM against a susceptible strain of C. albicans. 2.41 was shown to be potent against resistant strains of C. albicans but lacked activity against a resistant strain of C. glabrata. Previously synthesized spiroindolinone azole synergizers were tested against resistant strains of C. albicans and C. glabrata in the presence of fluconazole. There was activity against both strains. Over fifty new spiroindolinones were synthesized and three, 3.61, 3.78, and 3.71, were found to have activity at or below 100 nM against both resistant strains.

Reaction Predictor was a system developed by the Baldi lab in the early 2010’s to use deep learning to predict reaction mechanisms. In collaboration with the Baldi lab, Reaction Predictor was improved in accuracy and performance through various ways including writing 1000’s of training reactions, modifying the architecture and features of the neural net, and using simple chemistry rules. Reaction Predictor was able to achieve a 80% top-5 recovery rate on a separate, challenging benchmark set of reactions drawn from modern organic chemistry literature. Using pathway search, Reaction Predictor was able to identify plausible products from organic synthesis reactions as well as drug degradation.

Methyl anion affinity has previously been shown to correlate linearly with electrophilicity only within specific sets of functional groups. Using a solvation model in the calculation of methyl anion affinity was shown to give a correlation with electrophilicity over a broad class of charged and uncharged electrophiles. Methyl cation affinity calculated with a solvation model was shown to correlate with nucleophilicity over a broad class of charged and uncharged nucleophiles. Using these correlations we were able to estimate the span of electrophilicity and nucleophilicity in the universe.