Data for methylome sequencing: Enriching and Profiling Methylomes for Tumor Classification and Liquid Biopsies

We benchmarked and demonstrated the versatility of FLEXseq (Fragment Ligation EXclusive methylation sequencing) across different sample types: genomic DNA from the K562 (leukemia) cell line, DNA mix-in titrations of four immune cell types (B cells, T cells, monocytes, and neutrophils), DNA titrations of three cancer cell lines (breast invasive carcinoma [BRCA], colon adenocarcinoma [COAD], and glioblastoma [GBM]) mixed with those four immune cell mixtures separately, input titrations of cell-free (cf) DNA from one plasma sample and DNA from formalin-fixed paraffin-embedded (FFPE) tissues, cfDNA from 106 cerebrospinal fluids (CSF) and 42 other body fluids, and DNA from 37 FFPE tissues.

We sequenced all the samples mentioned above using FLEXseq. Paired-end reads were quality and length trimmed with cutadapt version 3.5, and all high-quality sequencing reads were then aligned to the hg38 reference genome using Bismark v0.23.0. We then filtered out reads with unmethylated cytosine in the non-CpG context with filter_non_conversion function. Next, we used the bismark_methylation_extractor function to extract the methylation calls (removing single-nucleotide polymorphisms [SNP]).

We also used the bam2pat function from wgbs_tools, to convert bam files into .pat files for deconvolution, keeping reads covering at least three CpG sites. The .pat files preserve fragment-level data and were de-identified by removing SNPs using the mask_pat function. 

We used CNVkit (v0.9.1) to analyze and visualize genome-wide copy numbers. Our inputs into CNVkit were Bismark/Bowtie 2 aligned BAM files that were deduplicated by Bismark based on end positions and fragment lengths. We then generated log2copy ratio plots for all body fluid and FFPE samples based on the pooled reference and visualized them across all bins using the DNAcopy R package.

We used t-distributed stochastic neighbor embedding (t-SNE) dimensionality reduction to visualize the embedding of our method cases using the TCGA and CNS tumor references. For the CNS tumor t-SNE plots, 2,801 references were used. We extracted 94 nontrivial components using the RSpectra R package and then employed the Rtsne R package with parameters of ‘pca=F, max_iter=2500, theta=0, perplexity=30’. For the TCGA t-SNE plots, 2,508 references were used with the same parameters above. Individual samples were incorporated separately into the references.