Evaluation of recombination detection methods for viral sequence analysis

Frederick Jaya⁰, Barbara Brito-Rodriguez⁰, Aaron Darling⁰
(0) University of Technology Sydney

Find me on Tues Nov 24th, 1:40-3pm AEDT in Remo, table 57

Abstract
To accurately infer the evolutionary history of viral genomes, the process of recombination needs to be accounted for and addressed appropriately. A vast choice of recombination detection methods have been developed over the past 20 years, but their ability to address the needs presented by high-throughput sequencing of viral data is unclear.

Here, we present the key considerations for selecting a suitable method for viral analyses. We assess five published methods used to detect recombination in nucleotide sequences - PhiPack (Profile), 3SEQ, GENECONV, UCHIME and gmos. The performance of methods were evaluated with analysis of within-host hepatitis C virus populations, simulated across a wide range of mutation and recombination rates. Scalability was assessed by recording the CPU time required to analyse datasets with n = 500, n = 1000 and n = 5000 sequences per alignment (1680 nt). In addition, empirical datasets of two bovine RNA viruses were analysed by each method and compared with simulation findings.

We find critical trade-offs between the methods, where the most scalable methods (PhiPack (Profile), UCHIME and gmos) may not be suitable for analysis of high coverage, within-host sequencing. Analysis of highly similar sequences (mean pairwise diversity < 1%) produced a high rate of positive detections in PhiPack (Profile), whereas 3SEQ and GENECONV are unable to process these. Overall, the five evaluated methods are inadequate for a rapid and reliable analysis of recombination in large viral datasets, presenting a severe unmet need for the development of scalable and accurate viral recombination detection methods.