Impacts of low coverage depths and post-mortem DNA damage on variant calling: a simulation study

BMC Genomics

Table 2 Impact of independent variables on variant call accuracy and variant call completeness

Divergence		%GC	Read length	Damage	Coverage depth
Low	% indels called correctly	-0.0194	0.0153	0.2150	0.4723log(covdepth)*/-0.1687*
	% SNPs called correctly	-0.0757	0.0828*	-5.1790***	-0.2730log(covdepth)/0.0405
	% of total indels called	-0.0459	-0.1057***	-2.7754**	3.9708***
	% of total SNPs called	-0.0314	-0.0521	-1.6778*	5.5177***
High	% indels called correctly	0.0092	0.0396***	0.2563*	1.660log(covdepth)*/-0.3292*
	% SNPs called correctly	-0.1495***	0.0249***	-0.9437***	-0.2730***
	% of total indels called	-0.0282	-0.1159***	-1.8524***	2.8565***
	% of total SNPs called	-0.0867**	-0.0616**	-1.0032*	3.1498***

Slope coefficients resulting from multiple variable regression analysis, considering as dependent variables the percent of indels/SNPs correctly called (variant call accuracy) and the percent of total indels/SNPs called (variant call completeness), while GC content, read length, damage level and coverage depth were treated as independent variables. Values of 1, 2 and 3 were assigned to variables no-, low- and high-damage, respectively. Values not significant at Pr < 0.05 unless otherwise indicated; * signifies 0.01 < Pr < =0.05; ** signifies 0.001 < Pr < =0.01; *** signifies Pr < =0.001. For each row, the independent variable with the strongest significant effect is highlighted in bold. Results solely for best-fit linear models are shown except for “%indels called correctly” at low and high divergence and “% SNPs called correctly” at low divergence. For these cases, a slight improvement in correlation was seen using a logarithmic model, and slope coefficients for both models are shown as: log coefficient / linear coefficient.

ISSN: 1471-2164