Abstract
'Tagmentation' approaches to bisulphite sequencing use a transposase to simultaneously make double-stranded breaks and ligate adaptors to the resulting fragments, allowing for higher throughput with less starting material. However, it has also been noted that certain tagmentation protocols have an unusually high number unmethylated cytosines that are not converted to thymine. Here we describe this phenomenon in detail, and find that results are consistent with single strand nicks by the transposase, followed by strand displacement of part or all of the DNA fragment, leading to erroneous incorporation of methylated cytosines. Nevertheless we show that these errors can be accounted for in downstream analysis and need not impede biological conclusions. We provide a Python package to allow users to implement this framework. Ultimately the additional effort of accounting for errors must be traded off against the scalability of the protocol in planning experiments.