Clinical utility and limitations of whole-genome sequencing via nanopore technology and Bayesian modeling for outbreak transmission inference.

Abstract

Background: Traditional molecular epidemiological analysis has been limited in its ability to evaluate pathogen homology, making detailed transmission pathway estimation in outbreaks challenging. Bayesian reconstruction methods using whole-genome sequencing and epidemiological data enable estimation of temporal transmission relationships from within-host diversity. This study examined the practical utility and limitations of transmission pathway inference via whole-genome analysis with nanopore sequencers for SARS-CoV-2 and multidrug-resistant Acinetobacter (MDRA) outbreaks.

Methods: Two outbreaks at the National Defense Medical College Hospital were analyzed: COVID-19 (9 individuals) and MDRA (22 cases). Transmission pathways were estimated via BadTrIP, an open-source Bayesian package for reconstructing transmission histories from genomic polymorphisms, with whole-genome sequencing performed on an Oxford Nanopore MinION. Input data included sample collection times, infection exposure periods, and nucleotide counts; SARS-CoV-2 genomes were obtained directly from nasopharyngeal swabs, whereas preserved bacterial isolates were used for MDRA.

Results: During the SARS-CoV-2 outbreak, one patient and eight staff members were infected within one week, and BadTrIP inferred transmission from the index patient to multiple healthcare workers, followed by onward spread mainly among staff, consistent with descriptive epidemiology. The inferred transmission tree highlighted staff providing intensive oral care to the index patient and identified a staff member who intermittently worked unmasked, aligning with known infection prevention considerations, although mask effectiveness could not be formally quantified. In the MDRA outbreak, the inferred transmission network showed several high-probability links between patients on different wards with limited apparent contact, indicating discrepancies with ward-based spread, and spreader analysis identified only longer hospitalized duration as significantly associated with spreader status.

Conclusions: BadTrIP is useful for transmission pathway estimation, although its accuracy depends on pathogen characteristics, outbreak duration, and sample properties. High concordance with clinical and epidemiological information was achieved for the short-term SARS-CoV-2 outbreak using directly sequenced clinical specimens, whereas the prolonged MDRA outbreak based on preserved isolates showed limited accuracy and epidemiologically difficult-to-explain links. BadTrIP therefore appears most informative for short- to intermediate-term outbreaks with sufficient genetic diversity and direct sampling, while transmission trees from long-term outbreaks using preserved isolates should be carefully interpreted as exploratory.