Purchase this article with an account.
Alexander Tuzhikov, Alexander Panchin, Darlene Miller, Jorge Maestre, Valery I Shestopalov; Can metagenomic analysis of microbiome help design ocular probiotics?. Invest. Ophthalmol. Vis. Sci. 2016;57(12):5866.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Homeostatic ocular surface (OS) microbiota contains predominantly bacteria and resides primarily on the conjunctiva and extends to the corneal epithelium. The composition of the OS microbiota changes dramatically during the onset of keratitis. The disease inflicts devastating consequences to vision, including blindness, and, therefore, early detection is crucial. The main purpose of this work was to evaluate bacterial relative abundance and map OS "microbiome core compositions" for both homeostatic and pathogenic cornea tissues. The homeostatic core is predicted to contain potential candidates for OS probiotics.
We applied TUIT classifier to the 16S rDNA sequencing data from 42 subjects. Bacterial abundance data were used to fit a number of machine learning models (including logistic regression, decision tree, and support vector machine) with CARET R package and the best models were merged to fit an ensemble predictive model. We then used Primer-BLAST to collect a comprehensive candidate identification set of primers for multiplex PCR analysis.
We have fitted and refined a predictive model, which is compatible with both PCR and sequencing data obtained from used contact lens surface. This model assigns a value (on a scale from 0 to 1), which reflects the calculated risk of keratitis development due to homeostatic OS microflora disruption. Based on the model, we were able to select 3 pivotal bacterial genera, including Propionibacteria, Pseudomonas and Staphylococcus that distinguish between homeostatic and pathogenic microbiome states. The identified genome assemblies for each genera allowed us to compile a set of specific primers for multiplex PCR, which may be used for rapid microflora detection in conjunctival swabs, used contact lenses and storage cases. Further analysis of candidate species genomes for low immunogenicity profile allowed identification of several candidate species for ocular probiotics.
Successful application of novel effective bioinformatics and machine learning algorithms allowed us to detect "core microbiomes" for healthy and pathogenic OS and identify the dynamics of the microbiome components. Our results indicate that further sampling and training of the model will achieve higher sensitivity and specificity. These data make the rational design of ocular probiotic formulation for beneficial microbiome manipulation a feasible task.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
This PDF is available to Subscribers Only