Data Analysis
We demultiplexed runs and conducted quality control with Torrent Suite version 4.4.3. We retained bases above the AQ20 confidence threshold. We trimmed sequences to 100 bp targeting the higher quality sequence at the 5’ end. We performed a de novo assembly and constructed a pseudo-reference using Geneious Prime v. 2022.1.1 (Dotmatics). We mapped individual sequences with BWA Galaxy Version 0.7.17.4 (Li and Durbin 2009, 2010). We used featureCounts Galaxy Version 1.6.4+galaxy1 (Liao et al. 2013) to determine read counts of fragments within 100 bp bins spanning the pseudo-reference. The 100 bp bins were used to count fragments among individuals ultimately to represent variation in DNA methylation among the CCGG sites screened. For a fragment to be sequenced, it had to have a non-methylated CCGG site. Counting matches to the bins across the pseudo-reference equates to variation in DNA methylation among the CCGG sites. As epiRADseq generates data with the zero read count result indicating DNA methylation, we used two approaches to control for sequencing coverage differences. First, we only analyzed individuals with 5,000 sequencing reads or higher. Second, we standardized all statistics by sequence read count at the individual sample level.
We used edgeR, Galaxy Version 3.24.1+galaxy1 (Robinson et al. 2010), to detect differently methylated regions (DMR), between the 0- and 8-hour samples, with a False Discovery Rate (FDR) of 0.05. We first compared all samples between 0- and 8-hour; we then repeated the comparison separately for individuals from native and introduced populations. We determined the number of DMRs in each comparison and identified DMR that were shared or unique to a particular comparison.
For every house sparrow, we calculated the change in DNA methylation between the 0- and 8-hour sample for all bins with significant differences as identified by the EdgeR analyses. We standardized each count for each bin by sequencing depth as (observed count for bin x / total read count) x 1,000. We compared methylation estimates among introduced and native birds using t-tests , f-tests , and Pearson’s correlations. Statistical tests used alpha = 0.05 and were corrected by the sequential Bonferroni method when appropriate (Rice 1989).