pycoMeth: A toolbox for differential methylation testing from Nanopore methylation calls

Abstract

Advances in Nanopore sequencing have opened up the possibility for the simultaneous analysis of genomic and epigenetic variation by way of base-calling and methylation-calling of the same long reads. Methylation analysis based on long read technologies requires a re-evaluation of data storage and analysis approaches previously developed for either CpG-methylation arrays or short-read bisulfite sequencing data. To address this, we here present a toolbox for the segmentation and differential methylation analysis of (haplotyped) methylation calls from Nanopore data. Additionally, we describe a storage format for read-level and reference-anchored methylation call data, which simultaneously allows for efficient storage and rapid data access.

Assembly of 43 diverse human Y chromosomes reveals extensive complexity and variation

“…After QC we leverage pycoMeth to de novo identify interesting methylation segments on chromosome Y. pycoMeth (version 2.2) 90 Meth_Seg is a Bayesian changepoint-detection algorithm that determines regions with consistent methylation rate from the read-level methylation predictions. Over the 139 QCed flowcells of the 41 samples, we find 2,861 segments that behave consistently in terms of methylation variation in a sample.…”

“…After QC we leverage pycoMeth to de novo identify interesting methylation segments on chromosome Y. pycoMeth (version 2.2) 90 Meth_Seg is a Bayesian changepoint-detection algorithm that determines regions with consistent methylation rate from the read-level methylation predictions.…”

Long-read sequencing of diagnosis and post-therapy medulloblastoma reveals complex rearrangement patterns and epigenetic signatures

“…We attempted to identify patterns of variation in DNA methylation by comparing methylation rates between primary tumor and relapse sample using PycoMeth 34 . We find that directly testing methylation rates of gene promoter regions (as defined in methods) yields poor power, with only 31 gene promoters called as differentially methylated (FDR <= 0.05, abs methylation rate difference > 0.5).…”

“…In order to find genomic regions with differential methylation between samples, we used the software package PycoMeth 34 . PycoMeth aggregates methylation likelihood ratios reported by Nanopolish over predefined regions, computes a read-level methylation rate from thresholded loglikelihood ratios (threshold 2.0) and then performs a Wilcoxon rank-sum test (for 2-sample comparison) or Kruskal Wallis test (for more than two samples) for methylation rates across samples.…”

Nanopore sequencing reveals methylation changes associated with obesity in circulating cell-free DNA from Göttingen Minipigs

“…A positive LLR indicated methylation and negative LLR indicated the absence of methylation [ 33 ]. Differential methylation analysis was performed by comparing the output from Nanopolish to that of pycoMeth (v. 0.4.7) [ 34 ]. Before comparison, two steps of quality control and data binning were performed in pycoMeth first, CpGs with their computed LLR were filtered so that only CpGs with a minimum of two cfDNA reads covering each CpG position were saved; second, the filtered CpGs were binned into genomic intervals of 1 kb spanning throughout the genome, only saving the intervals, which contained a minimum of 5 CpGs per interval.…”