Table 1.

Genotyping accuracy of Thunder and HapSeq across 16 different experimental settings

Site type	Total site count	Thunder discordance rate (%)	HapSeq discordance rate (%)	Absolute discordance change (%)	Relative performance gain (%)
Total	935 880	0.86	0.60	0.26	30
Heterozygous sites	133 337	1.26	0.94	0.31	25
Homozygous sites	802 543	0.78	0.55	0.23	29
Homozygous reference sites	532 255	0.78	0.54	0.25	31
Homozygous alternative sites	270 288	0.75	0.55	0.20	27
Sites not covered by jumping reads	411 098	0.91	0.64	0.27	30
Sites covered by jumping reads from either left or right	353 538	0.85	0.57	0.28	33
Sites covered by jumping reads from both left and right	171 244	0.76	0.43	0.32	43
Sites with depth = 0	18 011	0.86	0.63	0.23	27
Sites with depth ≥ 4	535 763	0.67	0.50	0.18	26
Site with 0 < depth < 4	382 106	1.09	0.72	0.37	34
Sites with sequencing error	39 589	6.16	4.62	1.54	25
Sites with MAF = 0%	481 060	0.97	0.67	0.30	31
Sites with 0 < MAF ≤ 1%	35 380	1.10	0.80	0.31	28
Sites with 1% < MAF ≤ 5%	52 960	1.17	0.90	0.27	23
Sites with MAF >5%	366 480	0.79	0.58	0.21	26

Site counts are over all PPSs over all 16 experiments. Absolute discordance change is the difference between the genotypic discordance rate of Thunder and that of HapSeq. Relative performance gain is the fraction of absolute discordance change over the discordance rate of Thunder.