Predictors of hDNA sample success
Allozyme supernatant replicates performed similarly to pellet replicates across metrics including depth of coverage, on-target mapping rates, and number of loci (Fig. 2A,C,D). Both replicates also recovered similar numbers of loci to the RADseq replicates (Fig. 2A), and a higher proportion of on-target reads (Fig. 2D). Overall, we inferred low rates of on-target mapping as expected with large amphibian genomes (Fig. 2D; Table S1), with a mean mapping rate of 12.77% (SD = 7.3) for supernatant replicates, 10.24% (SD = 0.8) for pellet, and a significantly lower rate of 4.31% (SD = 0.3) for RADseq (Tukey test adjusted p < 0.01). In the probe dataset the RADseq data had significantly higher coverage (Fig. S5).
By contrast, the formalin-fixed replicates had fewer loci on average (Fig. 2A), and significantly lower mean on-target mapping rates than the other replicates at 3.50% (SD = 1.1). Extraction yield was lower for formalin-fixed samples (Fig. 2A) and we found that at least ~200 ng of extracted DNA was needed to have ~50% of SNPs in the final matrix (Fig. 2A, S5A). Life stage was also qualitatively important, in that none of the larval specimens (n=3) gave high DNA yields from the formalin-fixed samples (Table S1) and the two we sequenced were high in exogenous DNA reads. As noted in the methods, larval specimens were stored in formalin for much longer periods of time than were metamorphosed specimens.