Benchmarking nanopore methylation analysis by comparison to publicly available bisulphite datasets

‘Epigenetics’ refers to heritable alterations of DNA that do not change the nucleotide sequence. One of the most widespread epigenetic modifications is 5-methyl cytosine (5mC), which most frequently occurs in mammalian cells in CpG dinucleotides. CpG methylation can alter patterns of gene expression by suppressing transcription. Nanopore sequencing does not require amplification or strand synthesis, so during sequencing, modified bases pass through the pore and the signature of these bases is present in the raw signal. To benchmark nanopore methylation calling, we ran two replicates of NA12878 on a single PromethION (TM) Flow Cell to obtain 20x per sample. We ran both samples separately through megalodon to perform base calling, mapping to hg38 and calling methylation (Fig. 1a). Methylation calls were compared to two 50x bisulphite-seq datasets: from ENCODE (ENCFF835NTC) and from GEO (GSE103505). Nanopore read depth is more consistent across the genome than the bisulphite data (Fig. 1b) and not influenced by GC content of the genome (Fig. 1c). Furthermore, unlike bisulphite data almost all nanopore data can be mapped to the genome (Fig. 1d). Taken together this results in a higher percentage of successfully called CpGs (>94%) at moderate overall read depth (Fig. 1f). Nanopore 5mC calls correlate well with other methods of methylation calling (>0.9) and are highly reproducible (>0.95) (Fig. 1g). Analysis time for nanopore methylation data is considerably less than for bisulphite data (Fig. 1h).

To examine the methylation status of larger features, computed the average methylation frequency of all CpGs in CpG islands. We defined those CpG islands with >40% of their CpGs covered by at least 10 reads as high confidence (green). All others (blue) were excluded from further analysis. We found high agreement between 20x Oxford Nanopore Technologies and 50x bisulphite datasets (Fig. 2a), with a correlation >0.98. For perspective, comparison of two Oxford Nanopore Technologies runs showed the highest reproducibility (Fig. 2b), whereas two bisulphite runs showed more variability (Fig. 2c). Both 20x Oxford Nanopore Technologies replicates allow calling of >25,000 CpGs islands. The two 50x bisulphite samples allow calling of 15,000 and 18,000 CpG Islands respectively (Fig. 2d). The differences may be due to the repetitive, GC-rich nature of CpG islands, which is challenging to bisulphite sequencing.

Nanopore sequencing allows for easy, fast, and accurate haplotype-resolved methylation calling. We used Medaka to call SNPs and Whatshap to separate reads into the parental haplotypes. Next, we used megalodon to call methylation separately for each set of reads yielding methylation status for both haplotypes. Fig. 3 shows an example of a known imprinted region in the human genome. Read coverage is highly uniform allowing per-base methylation calling across the whole region for both haplotypes. Roughly two thirds of the CpG islands in these regions show a difference in methylation pattern between the two haplotypes. We found the differentially methylated region to contain six known promoters for BCAP31 and ABCD1. However, we were unable to verify this because the bisulphite calls are too variable in this region.