Machine Learning Algorithm Identifies an Antibiotic Vocabulary for Permeating Gram-Negative Bacteria

Abstract

Drug discovery faces a crisis. The industry has used up the “obvious” space in which to find novel drugs for biomedical applications, and productivity is declining. One strategy to combat this is rational approaches to expand the search space without relying on chemical intuition, to avoid rediscovery of similar spaces. In this work, we present proof of concept of an approach to rationally identify a “chemical vocabulary” related to a specific drug activity of interest without employing known rules. We focus on the pressing concern of multidrug resistance in Pseudomonas aeruginosa by searching for submolecules that promote compound entry into this bacterium. By synergizing theory, computation, and experiment, we validate our approach, explain the molecular mechanism behind identified fragments promoting compound entry, and select candidate compounds from an external library that display good permeation ability.

Outer membrane permeability: Antimicrobials and diverse nutrients bypass porins in Pseudomonas aeruginosa

“…Porins and nutrients are grouped based on unsupervised clustering. ( C ) Permeability of 472 antimicrobial compounds with detectable antipseudomonal activity ( 15 ). Following the published analysis, compounds are grouped in permeability classes according to the MIC values for efflux-deficient P. aeruginosa strains with intact (P Δ6) or abolished (P Δ6-pore) outer membrane permeability barrier (class 0, MIC of PA Δ6-pore is less than 20% of PA Δ6; 1, 20 to 40%; 2, 40 to 60%, 3, 60 to 80%; 4 more than 80%).…”

“…2 C and D ; aconitate, citrate, and succinate). To extend these observations, we reanalyzed recent datasets for more than 500 antipseudomonal compounds in the 250- to 700-Da size range ( 14 , 15 ). In particular, we compared MIC values for efflux-deficient P. aeruginosa strains with intact outer membrane barrier (P Δ6, lacking six major efflux systems) or abolished barrier function because of insertion of water-filled pores with ∼2.4-nm diameter (P Δ6-pore).…”

“…Previous analysis of these data has revealed molecular fragments promoting translocation through the wild-type outer membrane (primary and secondary amine groups, benzene rings, and trifluoromethyl groups) ( 15 ). Our reanalysis showed an additional marked negative impact of carboxylate groups on membrane permeation ( Fig.…”

“…For P. aeruginosa , physicochemical parameters favoring translocation have been more difficult to identify ( 10 , 14 , 15 ). Both P. aeruginosa and A. baumannii have lower outer-membrane permeability than Enterobacteriaceae for hydrophilic molecules because they lack unspecific porins ( 16 ), making antimicrobial development particularly difficult for these critical pathogens.…”

The landscape of tumors-infiltrate immune cells in papillary thyroid carcinoma and its prognostic value

“…nih.gov/geo/). We used the "ComBat" algorithm (Mansbach et al, 2020) to reduce the batch effects from non-biological technical biases between GSE35570, GSE33630, and GSE60542. We combined these three datasets and referred to them as the GEO combined sets.…”