Choosing the chemistry

FiberHMM needs to know two things about your data: the enzyme that marked accessible DNA and, for Hia5, the sequencing platform. Together they pick the bundled model, the observation mode and every chemistry-dependent default.

--enzyme

Your assay	--enzyme	What FiberHMM reads
Fiber-seq with the m6A methyltransferase Hia5	hia5	m6A calls in MM/ML
DAF-seq with the double-stranded deaminase DddB	dddb	C→T / G→A conversions
DAF-seq with the single-strand-preferring deaminase DddA	ddda	C→T / G→A conversions

fiberhmm-pipeline, fiberhmm-call, fiberhmm-apply, fiberhmm-recall-tfs/-recall-nucs and fiberhmm-posteriors take --enzyme. Only these three presets are accepted. fiberhmm-qc also takes --enzyme (default auto, read from the header). For anything else, supply a model file with -m/--model (see Training a model).

--seq

--seq pacbio|nanopore matters for Hia5 only:

• PacBio reports m6A on both strands of the molecule (A calls on the read strand as A+a, and A calls on the opposite strand as T-a).
• Nanopore reports m6A only on the sequenced strand (A+a).

The two platforms therefore have different models (hia5_pacbio.json, hia5_nanopore.json), different observation modes (pacbio-fiber, nanopore-fiber) and different defaults (see below).

For DddA and DddB the deamination read-out does not depend on the platform: both use the daf mode on every platform, and --seq only sets the platform recorded in the output header. Without --seq, fiberhmm-call takes it from the header (@RG PL, aligner or basecaller @PG) and otherwise records pacbio for DddA and nanopore for DddB; fiberhmm-pipeline records nanopore for DAF reads it aligns that name no platform.

Automatic detection of --seq

When --seq is omitted for Hia5, fiberhmm-pipeline, fiberhmm-call, fiberhmm-apply, fiberhmm-posteriors and fiberhmm-recall-tfs/-recall-nucs look at the input, strongest evidence first:

1. a FIBERHMM-CHEMISTRY declaration in the header (a BAM FiberHMM already called);
2. the MM specifications of the first 200 records: T-a present means PacBio, A+a alone means Nanopore;
3. @RG PL values and @PG program names (for example pbmm2, ccs, jasmine for PacBio; dorado, guppy, minimap2 -x map-ont for Nanopore).

The choice is printed:

NOTE: --seq not given; using --seq nanopore (detected from MM specs of 200 read(s) (A+a only, no T-a)).

If the sources disagree, or more than 10% of the inspected reads show the other MM pattern, the run stops and asks for --seq:

error: fiberhmm-call: cannot infer the sequencing platform for --enzyme hia5: MM specs are mixed: ... Pass --seq pacbio or --seq nanopore.

Input that cannot be inspected (stdin) falls back to PacBio with a warning, so pass --seq when you stream:

WARNING: --seq not given for --enzyme hia5 and the platform could not be detected (stdin (not inspected)); assuming PacBio (--seq pacbio). Pass --seq nanopore for Nanopore data.

For Hia5, an explicit --seq that the reads contradict is refused, because the other platform's model changes the calls:

error: fiberhmm-call: --seq pacbio was given, but the input looks like nanopore (MM specs of 200 read(s) (A+a only, no T-a)). Calling nanopore reads with the pacbio model changes the calls (for example, Nanopore reads called as PacBio give ~100x more TF calls). Pass --seq nanopore, omit --seq to detect it, or add --force-seq to use --seq pacbio anyway.

Hia5 runs on reads whose first records carry no m6A calls at all (MM A+a or T-a) are refused too, with a hint that the data may be DAF-seq. --force-seq (in fiberhmm-call, -apply, -posteriors and -recall-tfs/-recall-nucs) runs anyway in both cases. For DddA and DddB a --seq that disagrees with the header only warns.

What the chemistry changes

Setting	Hia5 PacBio	Hia5 Nanopore	DddB	DddA
Model	hia5_pacbio.json	hia5_nanopore.json	dddb_nanopore.json	ddda_nuc.json (HMM), ddda_TF.json (TF recall)
Observation mode	pacbio-fiber	nanopore-fiber	daf	daf
ML threshold in fiberhmm-call	128	248	128 (MM/ML dU only)	128 (MM/ML dU only)
Nucleosome recall	accessible-cut split	accessible-cut split	accessible-cut split	phase-aware radial
CpG mask (first-pass HMM and recall; --use-m5c)	off	off	off	on
Adjacent-target thinning	—	—	off	runs ≥ 2, keep-one
Duplicate marking, SNP screen	—	—	on (file input)	on (file input)

The full table, including the thresholds of the other tools, is in Chemistries and platforms.

Re-calling a BAM FiberHMM already called

fiberhmm-call, fiberhmm-apply and fiberhmm-recall-tfs/-recall-nucs write a FIBERHMM-CHEMISTRY line into the output header. Tools that read such a BAM take the chemistry from it:

• fiberhmm-recall-tfs/-recall-nucs fill a missing --seq from it;
• fiberhmm-extract and fiberhmm-qc use it for their ML threshold and QC reference;
• fiberhmm-consensus and fiberhmm-transfer use it to pick the emission model.

A run whose chemistry contradicts the declaration stops with a one-line fix:

error: fiberhmm-recall-tfs: the input BAM declares chemistry [assay=daf enzyme=dddb platform=nanopore mode=daf] but this run would declare [assay=fiber-seq enzyme=hia5 platform=nanopore mode=nanopore-fiber]. Pass --enzyme/--seq matching the input (with --model for a custom table), or --replace-chemistry to re-declare the output deliberately.

A custom --model given without --enzyme inherits the declared enzyme and platform when its observation mode matches, and then uses that enzyme's defaults:

NOTE: fiberhmm-recall-tfs: custom model on an input declaring [assay=daf enzyme=dddb platform=nanopore mode=daf]; using the defaults of --enzyme dddb --seq nanopore with the given model file(s).

--replace-chemistry drops the input's declaration and writes this run's (for a custom model, enzyme=custom;platform=unknown). Use it only when you are deliberately re-calling with a different chemistry.