Exploring the role of phage resistance by using label-free SERS for biochemical profiling of gram-negative pathogen Pseudomonas aeruginosa strains with machine learning approach.

Pseudomonas aeruginosa (P. aeruginosa) is an emerging gram-negative pathogen, accountable for diverse and chronic nosocomial infections, particularly in immunocompromised patients and frequently associated with multidrug resistant, posing a serious challenge to conventional antibiotic therapy. Bacterial infections are getting more difficult to treat as antibiotics are becoming less effective due to bacterial resistance. Development of advanced and novel treatment arises due to increasing bacterial resistance from P. aeruginosa. Bacteriophages are beneficial options to resist bacterial infections. This study presents a novel integration of surface-enhanced Raman spectroscopy (SERS) with machine learning algorithms to rapidly differentiate gram-negative pathogenic bacterial strains of P. aeruginosa, one phage-sensitive and one phage-resistant, with three replicates of each strain. Biochemical profiles comparison of P. aeruginosa strains was identified using SERS. For this purpose, SERS spectra are acquired using silver nanoparticles (Ag-NPs), as SERS substrate, from the pellets (cell mass) of these bacteriophage-sensitive and resistant bacterial strains, in triplicate. Protein contents in pellets of bacterial strains of P. aeruginosa are major distinguishing biomolecules. Substantial biochemical changes between sensitive and resistant bacteriophage strains of P. aeruginosa are found in this experimental study using SERS, such as proteins, lipids, carbohydrates, DNA and nucleic acids. Furthermore, chemometric analysis like principal components analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were used to evaluate the SERS spectral fingerprints of the bacterial pellets. The unsupervised PCA model was used for the classification of bacteriophage-resistant and sensitive bacterial sample data sets. PLS-DA is used to discriminate the SERS spectral data sets based on variance which were measured in terms of specificity 100%, sensitivity 99.8% and accuracy100% and AUC value 84% from receiver operating curve (ROC). This label-free, non-destructive and rapid strategy provides a promising analytical platform to support bacteriophage therapy development.