Table 1.

Results of haplotype phasing with Nanopore long reads (NA12878)

		GraphMap		BLASR		LAST				LAST + LAST-TRAIN
						Manual (q = 1)		Manual (q = 2)		Training		Training+LAMA
Haplotype	Polymorphism	Count	Freq	Count	Freq	Count	Freq	Count	Freq	Count	Freq	Count	Freq
TT: CT	CYP2D6*4	207	11.9%	227	18.4%	340	20.1%	182	21.2%	327	27.3%	343	27.4%
TT: CC	(reference bias)	225	13.0%	329	26.6%	326	19.3%	164	19.1%	160	13.4%	134	10.7%
T−: CT		70	4.0%	31	2.5%	65	3.8%	36	4.2%	75	6.3%	78	6.2%
T−: CC	CYP2D6*3	226	13.0%	281	22.8%	232	13.7%	199	23.1%	199	16.6%	217	17.3%
Other		1006	58.1%	367	29.7%	726	43.1%	279	32.4%	436	36.4%	480	38.3%
Total		1734	100.0%	1235	100.0%	1689	100.0%	860	100.0%	1197	100.0%	1252	100.0%

In the first column, TX:CY indicates the phased haplotype where the 1st position (rs35742686) is ‘X’ (‘T’ in the reference genome) and the 2nd position (rs3892097) is ‘Y’ (‘C’ in the reference genome). See also Supplementary Table S14. The high frequency for TT:CC (the identical haplotype to the reference genome) is known as reference bias (Laver et al., 2016). The values for ‘BLASR’ were computed from the mapping results in Ammar et al. (2015), where BLASR was used for mapping Nanopore reads to the reference genome. The column ‘training + LAMA’ shows the results of probabilistic alignment (Hamada et al., 2011) using forward scores with the trained parameters by LAST-TRAIN. See Supplementary Materials S7 for the detailed command line options for every tool.