TAPS libraries

TAPS (TET-assisted pyridine borane sequencing) is supported via --meth=taps. This page is the short version; the mechanics live on the existing pages and are linked from here rather than repeated.

TL;DR

# Index once — the SAME index serves both chemistries.
bwa-mem3 index --meth ref.fa

# Align TAPS reads. Note the '=' — see the warning below.
bwa-mem3 mem --meth=taps -t 16 ref.fa R1.fq.gz R2.fq.gz | samtools sort -o out.bam

Do not align TAPS data with a bare --meth. It will succeed, produce well-placed reads, and emit XM:Z methylation calls that are completely inverted — percent methylation comes out as 1 − truth, with no warning. Measured against simulated ground truth: Pearson r = −0.956 with --meth versus +0.956 with --meth=taps.

Use --meth=taps, not --meth taps. The argument is optional, and getopt_long only accepts optional arguments in the = form. A separate word is parsed as a positional — i.e. as your first FASTQ.

Why one flag is enough

Bisulfite/EM-seq converts the unmethylated cytosines; TAPS converts the methylated ones (5mC/5hmC). But both produce the same base change as far as the aligner can see — ref C → read T on the top strand, ref G → read A on the bottom. So seeding, the converted index, and the scoring matrices are shared, and TAPS reads align correctly with the existing machinery.

The chemistries differ only in which cytosines converted, which inverts the meaning of the observed base when calling methylation. --meth=taps therefore changes the XM:Z call polarity and nothing else. CpG/CHG/CHH context classification is a property of the reference, so it is identical either way. See SAM tags.

What --meth=taps changes

The scoring default differs because TAPS conversions are far sparser — 3.2 % of cytosines versus 94.6 % for em-seq on a matched simulation — so the collapsed 3-letter alphabet costs specificity it can no longer repay. neutral goes one step further than genomic: it scores the conversion cell as 0 (tolerated but not rewarded) rather than a full match, so a sparse TAPS conversion no longer over-credits a spurious C→T alignment. Measured placement on simulated TAPS against full hg38, across three methylation loads:

neutral reaches essentially the unconverted ceiling — a robust +0.24–0.28 pp over genomic — with identical MAPQ calibration (60+ bin 99.99 % either way) and NM (mean 1.966 vs 1.969), so the gain is not bought with over-confidence or inflated edit distance. An explicit --meth-scoring always overrides the default.

Downstream methylation calling

Most extractors assume bisulfite polarity. MethylDackel reports the fraction of reads showing C at a reference C, which under TAPS is the unmethylated fraction — so TAPS methylation is 100 − MethylDackel_rate. Either apply that inversion explicitly, or use a TAPS-aware caller such as Rastair (what nf-core/methylseq uses for its TAPS path).

XM:Z from --meth=taps is already in the correct polarity, so tools that consume XM:Z directly need no adjustment.

Library directionality

--meth assumes a directional library (R1 → CT, R2 → GA). Standard TAPS protocols are directional; verified on real data (SRR25526768), where the XG:Z split is 49.8 % CT / 49.9 % GA. Undirectional libraries are not supported.

Validation status

• Simulated TAPS with ground truth: r = +0.956, RMSE 0.117 (em-seq control: +0.954, 0.119) over ~600 k CpG sites.
• Real human TAPS (SRR25526768, HAP1): --meth=taps and --meth produce exact complement call sets on identical alignments.
• Not yet validated: absolute call accuracy on real data (no truth set), x86 SIMD tiers (results are arm64/NEON), a methylation-rate sweep, and undirectional libraries.

Full experiment, including the controls and what was not tested: reports/2026-07-20-taps-alignment-experiment-results.md.

See also: Flags · SAM tags: XR, XG, XM · Conversion details · Overview