A binning tool to reconstruct viral haplotypes from assembled contigs

Jiao Chen; Jiayu Shang; Jianrong Wang; Yanni Sun

We applied VirBin to cluster contigs into 5 groups. Since the ground truth about the haplotype membership of each contig is known, we were able to evaluate the clustering results by calculating the precision and recall at the base level. The evaluation results are shown in Fig. 4 . The performance of clustering is worst for shortest contig set (denoted as 1000 along the Yaxis). With increasing contig lengths, the clustering performance becomes better for all three different abundance distributions. When the contigs are long, the clustering performance for haplotypes with different abundance distributions is comparable. The results were compared with MaxBin, which is a binning tool for metagenomic contigs based on tetranucleotide frequencies and reads coverage levels. MaxBin requires marker genes to identify seed contigs for binning. We were able to run the core clustering program of MaxBin by inputting both the number of haplotypes (i.e. 5) and the seed contigs manually. We randomly chose one contig from each haplotype as the seed contig and calculated the contigs' abundances by mapping reads to them using Bowtie2. Although the haplotype number was explicitly provided to MaxBin, empty clusters can be produced by MaxBin. The results from MaxBin are shown in Fig. 5 .