Infinium human methylation 450 arrays

All the sequencing and clinical data associated with TNBC patients (n = 180) were obtained from The Cancer Genome Atlas (TCGA) hosted by the Genomic Data Commons (GDC). Patients (n = 180) were identified as triple negative breast cancer patients based on the analysis performed by Lehmann et al.^{21 (link)}. RNA-Seq and miRNA-Seq data were downloaded as raw counts. The CNV data from Affymetrix SNP 6.0 array was processed by GDC into output files containing segment mean values (transformed copy number values for each of the segmented genomic regions), which were used for this work. DNA methylation levels (as beta values) from Illumina Infinium Human Methylation 450 arrays (if available) were also obtained. (More information on GDC output files and pipelines can be found on the GDC website: https://docs.gdc.cancer.gov/.) Clinical metadata (Supplementary Table S1) were extracted from clinical XML files provided by GDC.
Purity estimates for the 180 TNBC tumors were obtained from Aran et al.^{55 (link)} and Li et al.^{30 (link)}. Patients were kept in the study if they had a consensus purity estimate (CPE) (from Aran et al.) or Clonal Heterogeneity Analysis Tool (CHAT) (from Li et al.) purity measurement of 60% or higher. The purity scores are presented in Supplementary Table S1.

Normalised ROR2 expression and methylation data of tumour and matched normal tissue were obtained from The Cancer Genome Atlas (http://cancergenome.nih.gov/) and analysed by Agilent microarrays and Illumina HiSeq 2000 RNA Sequencing. Methylation values were analysed using Illumina Infinium (HumanMethylation450) arrays and the beta-value average was obtained from methylation probes that fell within the 1958 bp ROR2 CpG island. Statistical significance of matched patient tumour and normal samples were carried out using paired t-test. Statistical significance of expression and methylation comparison of the entire cohort was evaluated using unpaired t-test. Statistical significance of expression in low and high methylation samples was evaluated using unpaired t-test. The results shown in these analyses are in whole or part based upon data generated by the TCGA Research Network; http://cancergenome.nih.gov/.

DNA methylation profiling of the six PDOX models was performed using Illumina Infinium Human Methylation450 arrays as previously described.^[27] Tumors were classified based on their DNA methylation profiles using the DKFZ brain tumor methylation classifier (http://www.molec ularneuropathology.org), which is based on a reference cohort of 2801 CNS tumors and normal tissues representing 82 distinct tumor and 9 distinct non‐neoplastic brain methylation classes.^[27]