Table 1.

Comparison of two phasing methods, Falcon Unzip and our graph-based approach, at different PacBio coverage levels

Statistics	PacBio coverage	Graph-based approach	Falcon Unzip
Diploid assemblies quality
Average identity (%)	10×	99.50	—
	20×	99.61	—
	30×	99.80	99.4
Phasing error rate (%)	10×	2.5	—
	20×	1.5	—
	30×	0.7	3.8
Contiguity
N50 haplotig size (bp)	10×	40k	—
	20×	42k	—
	30×	43k	32k
Completeness
Haplotig size (Mbp)	10×	20.7	—
	20×	21.1	—
	30×	23.9	16.6
# Unphased contigs	10×	2	—
	20×	2	—
	30×	2	77

Note: For computing the ‘haplotig N50’, we only consider those portions of a contig for which two haplotypes are available, i.e. those regions where Falcon reports both a primary contig and an alternative haplotig. For ‘haplotig size’, we sum the length of contigs on both haplotypes (‘primary contigs’ plus ‘haplotigs’ in terms of Falcon’s output), so the target size is twice the genome size (24.3 Mbp in case of yeast).