Benchmark results.

The best value for each dataset is in bold. While Bowtie2+ (keeping reads mapping to the foreground reference) generally gives the lowest false-positive rate (FPR) and Bowtie2- (discarding reads mapping to the background reference) the highest sensitivity (SEN), RAMBO-K shows the best balance, providing high SEN and low FPR (F-Score) with the consistently lowest run-time. RAMBO-K outperforms other methods by the largest margin when the nearest known reference has a low identity to the sequenced genome, as in the Bat adenovirus dataset.a: The values for PhymmBL on the Wolbachia dataset were extrapolated based on the analysis of a subset of 5% of the reads.Benchmark results.