Whole genome sequencing data

The Illumina whole genome sequencing data of four pairs of tumor and matched normal cell line samples were downloaded from The Cancer Genome Atlas (TCGA) (29 (link)) or Illumina BaseSpace (BaseSpace.illumina.com">https://BaseSpace.illumina.com). All of the 8 samples are at higher than 49x coverage, except for HCC2218BL at 37x (Table 1). Two independent normal 30X samples were also available for HCC1143 and HCC1954. For the other two samples HCC1187 and HCC2218, we generated a second ‘normal’ by downsampling the available matched normal to 30X. We will refer to these second normal samples as ‘0% mixtures’ as we will use them as a negative control to investigate whether methods detect tumor DNA in mixtures without tumor DNA.
By mixing the different amount of reads from pure tumor samples together with the ‘0% mixture’ samples (i.e. the second normal sample), a series of mixtures samples were created at 30x coverage with 5%, 20%, 40%, 60%, 80% and 95% of tumor DNA. These mixture samples were obtained as BAM files from TCGA (29 (link)) for HCC1143 and HCC1954, from Illumina BaseSpace for HCC1187 and HCC2218 (Supplementary Data). We also created an extra 10% mixture for all samples.

TE-tracker software was used to interrogate available Illumina whole-genome sequencing data from 122 epiRILs for the presence of >2 shared transposition evens (STEs) within the epiQTLs intervals (Gilly et al., 2014 (link)). STEs were analysed for statistically significant linkage with resistance phenotypes (RIs), using the same linear regression model as described above for DMR linkage analysis.

Sixty-three F₁ offspring were reproduced from selfed R. breviuscula. Each F₁ plant was sequenced with ~3× Illumina whole-genome sequencing data. To genotype F₁ offspring, whole-genome sequencing Illumina sequences of each plant were first mapped to the rhyBreHap1 reference genome using bowtie2 (v.2.4.4) paired-end mode, and then SNPs were called by ‘bcftools mpileup’ and ‘bcftools call’ (v1.9) (with ––keep-alts, ––variants-only and ––multiallelic-caller flags enabled). Next, SNPs of each F₁ sample were input to TIGER^{91 (link)} for genotyping and to generate potential CO positions. In addition, RTIGER^{92 (link)} was used to identify the genotypes of chromosomal segments by utilizing the corrected markers that resulted from TIGER. Only COs that agreed using both tools were kept. The recombination landscape from F₁ COs (Supplementary Fig. 4) was plotted using the same strategy and sliding window, as illustrated for pollen nuclei.