Telomere length assessments using whole-genome sequencing data
Telomere length estimates were significantly different across
bioinformatic approaches (H = 230.06, df = 2, p< 0.001, Figure 2). A post-hoc Dunn test between telomere
lengths estimated by Computel, K-Seek, and TRIP indicates significant
differences for all pairwise comparisons (p < 0.001).
On average, telomere length estimates using Computel (4,144 ± 796 bp)
were substantially lower than K-seek (35,487 ± 11,326 bp) and TRIP
(57,604 ± 17,393 bp). However, while the three bioinformatic approaches
produced different telomere length estimates on average, there was
substantial correlation between values estimated for individual
genotypes (Figure S3). Correlations between genotype measures were
lowest for Computel and K-seek (r = 0.86), but greater for
Computel and TRIP (r = 0.88), and K-seek and TRIP (r =
0.99). This suggests that despite variability in the average length
calculation there is a correlation between telomere length calculated
across approaches. This may suggest that differences in the parameters
included in the bioinformatics package, including the accounting for
genome coverage and minimum number of consecutive telomeric repeats
present in a read, may influence telomere length estimates on average,
but telomere length calculated for the same genotype across approaches
is largely correlated. Indeed, after correcting for genome coverage
(Equation 2) pairwise correlations between genotype measures increased
(Figure S4), but telomere length estimates were significantly different
across bioinformatic approaches (H = 246.59, df = 2, p< 0.001, Figure S5). Importantly, K-seek (1,423 ± 300.65 bp)
and TRIP (2,315± 451.67 bp) exhibit substantially shorter telomere
length estimates after genome coverage correction.