Predicting Conjunctivitis Outcomes via Genome Sequencing and Machine Learning
By Jean Shaw
Selected by Emily Y. Chew, MD
Journal Highlights
Ophthalmology Science, December 2022
Download PDF
To what degree do genetic variants determine clinical outcomes in adenoviral conjunctivitis? Nakamichi et al. explored this question via whole-genome viral sequencing and found that adenovirus (AdV) type D8 has three prevalent molecular substrains that predispose to varying risk of developing subepithelial infiltrates (SEIs), a complication of viral conjunctivitis.
For this study, the researchers evaluated banked conjunctival swab samples collected during the BAYnovation Study, a previously conducted study of a compound to treat adenoviral keratoconjunctivitis. They purified DNA from the swabs of 96 patients with AdV D8–positive samples and subjected it to viral sequencing. Viral variants were identified and correlated with clinical outcomes. Two machine learning models were independently trained to predict clinical outcomes using polymorphic sequences. The main outcome measures were viral DNA sequence and development of SEIs.
Full genome reconstructions were obtained for 71 AdV D8–positive samples. A total of 630 single-nucleotide variants were identified, including 156 missense mutations. Sequence clustering revealed three previously unappreciated viral clades within the AdV D8 type. The likelihood of developing SEIs differed between the clades, ranging from 46% to 83%. With regard to the two machine learning models, using a newly sequenced validation set of 16 cases, the models proved capable of predicting SEI development with >97% accuracy.
These findings add to prior evidence that sequence variants influence disease severity in adenoviral conjunctivitis, the researchers said. In addition, they noted, the use of machine learning may be applicable to other questions in viral pathogenesis—including the “determination of oncogenic potential of human papilloma viruses, understanding risks for reactivation of varicella zoster causing shingles, or understanding determinants of outcomes from SARS-CoV2.”
The original article can be found here.