From Genotype to Phenotype: Raman Spectroscopy and Machine Learning for Label-Free Single-Cell Analysis.

Raman spectroscopy has made significant progress in biosensing and clinical research. Here, we describe how surface-enhanced Raman spectroscopy (SERS) assisted with machine learning (ML) can expand its capabilities to enable interpretable insights into the transcriptome, proteome, and metabolome at the single-cell level. We first review how advances in nanophotonics-including plasmonics, metamaterials, and metasurfaces-enhance Raman scattering for rapid, strong label-free spectroscopy. We then discuss ML approaches for precise and interpretable spectral analysis, including neural networks, perturbation and gradient algorithms, and transfer learning. We provide illustrative examples of single-cell Raman phenotyping using nanophotonics and ML, including bacterial antibiotic susceptibility predictions, stem cell expression profiles, cancer diagnostics, and immunotherapy efficacy and toxicity predictions. Lastly, we discuss exciting prospects for the future of single-cell Raman spectroscopy, including Raman instrumentation, self-driving laboratories, Raman data banks, and machine learning for uncovering biological insights.

Predicting permeation of compounds across the outer membrane of P. aeruginosa using molecular descriptors.

The ability Gram-negative pathogens have at adapting and protecting themselves against antibiotics has increasingly become a public health threat. Data-driven models identifying molecular properties that correlate with outer membrane (OM) permeation and growth inhibition while avoiding efflux could guide the discovery of novel classes of antibiotics. Here we evaluate 174 molecular descriptors in 1260 antimicrobial compounds and study their correlations with antibacterial activity in Gram-negative Pseudomonas aeruginosa. The descriptors are derived from traditional approaches quantifying the compounds' intrinsic physicochemical properties, together with, bacterium-specific from ensemble docking of compounds targeting specific MexB binding pockets, and all-atom molecular dynamics simulations in different subregions of the OM model. Using these descriptors and the measured inhibitory concentrations, we design a statistical protocol to identify predictors of OM permeation/inhibition. We find consistent rules across most of our data highlighting the role of the interaction between the compounds and the OM. An implementation of the rules uncovered in our study is shown, and it demonstrates the accuracy of our approach in a set of previously unseen compounds. Our analysis sheds new light on the key properties drug candidates need to effectively permeate/inhibit P. aeruginosa, and opens the gate to similar data-driven studies in other Gram-negative pathogens.

High rates of plasmid cotransformation in E. coli overturn the clonality myth and reveal colony development.

The concept of DNA transfer between bacteria was put forth by Griffith in 1928. During the dawn of molecular cloning of DNA in the 1980s, Hanahan described how the transformation of DNA plasmids into bacteria would allow for cloning of DNA fragments. Through this foundational work, it is widely taught that a typical transformation produces clonal bacterial colonies. Using low concentrations of several plasmids that encode different fluorescent proteins, under the same selective antibiotic, we show that E. coli bacteria readily accept multiple plasmids, resulting in widespread aclonality and reveal a complex pattern of colony development. Cotransformation of plasmids occurs by either CaCl2 or by electroporation methods. A bacterium rod transformed with three plasmids-each expressing a high level of a unique fluorescent protein-and replated on agar, appears to reassign a random number of the three fluorescent plasmids to its daughter cell during cell division. The potential to simultaneously follow multiple lineages of clonally related bacteria in a bacteria colony would allow for mosaic analysis of gene function. We show that clonally related bacterium rods self-organize in a fractal growth pattern and can remain linked during colony development revealing a potential target against microbiota growth.

Predictive Rules of Efflux Inhibition and Avoidance in Pseudomonas aeruginosa.

Antibiotic-resistant bacteria rapidly spread in clinical and natural environments and challenge our modern lifestyle. A major component of defense against antibiotics in Gram-negative bacteria is a drug permeation barrier created by active efflux across the outer membrane. We identified molecular determinants defining the propensity of small peptidomimetic molecules to avoid and inhibit efflux pumps in Pseudomonas aeruginosa, a human pathogen notorious for its antibiotic resistance. Combining experimental and computational protocols, we mapped the fate of the compounds from structure-activity relationships through their dynamic behavior in solution, permeation across both the inner and outer membranes, and interaction with MexB, the major efflux transporter of P. aeruginosa We identified predictors of efflux avoidance and inhibition and demonstrated their power by using a library of traditional antibiotics and compound series and by generating new inhibitors of MexB. The identified predictors will enable the discovery and optimization of antibacterial agents suitable for treatment of P. aeruginosa infections.IMPORTANCE Efflux pump avoidance and inhibition are desired properties for the optimization of antibacterial activities against Gram-negative bacteria. However, molecular and physicochemical interactions defining the interface between compounds and efflux pumps remain poorly understood. We identified properties that correlate with efflux avoidance and inhibition, are predictive of similar features in structurally diverse compounds, and allow researchers to distinguish between efflux substrates, inhibitors, and avoiders in P. aeruginosa The developed predictive models are based on the descriptors representative of different clusters comprising a physically intuitive combination of properties. Molecular shape (represented by acylindricity), amphiphilicity (anisotropic polarizability), aromaticity (number of aromatic rings), and the partition coefficient (LogD) are physicochemical predictors of efflux inhibitors, whereas interactions with Pro668 and Leu674 residues of MexB distinguish between inhibitors/substrates and efflux avoiders. The predictive models and efflux rules are applicable to compounds with unrelated chemical scaffolds and pave the way for development of compounds with the desired efflux interface properties.