SPlinted Ligation Adapter Tagging (SPLAT), a novel library preparation method for whole genome bisulphite sequencing

Abstract

Sodium bisulphite treatment of DNA combined with next generation sequencing (NGS) is a powerful combination for the interrogation of genome-wide DNA methylation profiles. Library preparation for whole genome bisulphite sequencing (WGBS) is challenging due to side effects of the bisulphite treatment, which leads to extensive DNA damage. Recently, a new generation of methods for bisulphite sequencing library preparation have been devised. They are based on initial bisulphite treatment of the DNA, followed by adaptor tagging of single stranded DNA fragments, and enable WGBS using low quantities of input DNA. In this study, we present a novel approach for quick and cost effective WGBS library preparation that is based on splinted adaptor tagging (SPLAT) of bisulphite-converted single-stranded DNA. Moreover, we validate SPLAT against three commercially available WGBS library preparation techniques, two of which are based on bisulphite treatment prior to adaptor tagging and one is a conventional WGBS method.

Comparison of whole-genome bisulfite sequencing library preparation strategies identifies sources of biases affecting DNA methylation data

“…In order to increase relevance, comparability and robustness of our study, we have performed a retrospective cross-study and cross-species analysis combining datasets generated by our lab as well as other labs (225 libraries in total including non-BS control; see Table 2 and Additional file 1) [28][29][30][31][32][33][34][36][37][38][39][40][41][42][43][44]. To capture method differences over batch differences, each method is represented by at least two studies sourced by two different laboratories, where possible ( Table 2).…”

Data quality of Whole Genome Bisulfite Sequencing on Illumina platforms

“…Next, we were interested in how the amount of "usable" data (post alignment) obtained from a HiSeq X lane compares to a HiSeq2500 lane. Alignment rates of the WGBS data generated on the HiSeq X with HD.3.4.0 /RTA 2.7.7 were on par with previously generated WGBS data from HiSeq 2500 [9] and higher than those obtained from the previous HiSeq X software version (77-80 % as compared to 65-75 %) ( Table 1 and Supplementary Table 3). The levels of duplicate reads were 15-20 % for SPLAT and Accel-NGS Methyl-Seq, whilst the TSDM libraries had higher (34-40%) duplication rates.…”

“…Moreover, reduced Q-scores could potentially lead to lower alignment rates and data loss as less data might pass quality filtering. For bisulfite sequencing on HiSeq 2500 (v4 chemistry), the guanine Q-score issue was mitigated from software update HCS v2.2.38/RTA 1.18.61 and forward so that WGBS data generated with this system is now generally of high quality [9] and can such be used for comparison with data generated on the HiSeq X system.…”

“…General metrics for all of the HiSeq X sequencing runs is found in Table 1. For comparison we included data generated from libraries from the same DNA samples sequenced on the HiSeq 2500 system (HCS v2.2.38/RTA 1.18.61) [9].…”

“…SPLAT libraries prepared from various human sample sources (REH, NA10860, NA11992), each with differing levels of global methylation, were equally affected ( Supplementary Table 1). TSDM libraries display a typical and pronounced GC bias (26.1-26.8% GC for NA10860), meaning that GC regions are overrepresented in this library type, whereas SPLAT libraries have a more even representation of the genome (22.5-25.1% GC for NA10860) [9]. Therefore, it appears that the guanine Q-scores assigned by RTA: 2.7.1 and RTA: 2.7.5 appear to improve with increasing GC content of the library (Supplementary Figure 2).…”