Consensus sequence
TL;DR: The consensus sequence panel calls a per-position consensus from the reads in an alignments track, letting the reads rather than the reference define the sequence over a region of interest. Download the result as FASTA, or as a VCF of the positions that differ from the reference (also openable directly as a variant track).
Opening the panel
Two ways in, differing only in which region the panel opens on:
- From the track menu — Launch → Consensus sequence (visible region) on the alignments track. The panel opens on what the view is showing, and its Region field is editable, so this is a starting point rather than the call.
- From a selection — click and drag across the region of interest, then Launch → Consensus sequence in the menu that appears, then the alignments track to call from. That last step is there even with a single track open (which track the reads come from decides the answer, so it is always named), and the dialog title repeats the name.
Regions are limited to 500kb. Reads are much heavier to fetch than reference sequence, so a larger region is refused — the panel says so and waits for a smaller one typed into the field.
How a position is called
Every read covering a position votes for the base it carries there, or for a gap if the read has a deletion. Votes are weighted, so a read carrying an ambiguity code splits its vote across the bases that code covers, and a read carrying N contributes a small amount to all four bases.
The base with the most votes wins, provided it clears the minimum call fraction; otherwise the position is N (reads were present but did not agree enough).
Deletions are treated asymmetrically. A winning gap is always called, even below the call fraction, because agreement that a base is absent is still evidence of absence. A winning base below the call fraction is N.
Settings
The panel opens showing just the called sequence. Tick Show options to adjust how it was called; every setting, including whether the options are showing, is remembered for the next consensus you run.
- Min read depth - positions covered by fewer reads than this are N, regardless of how well the reads agree. Defaults to 1.
- Min call fraction - the fraction of the weighted vote the call must account for. Defaults to 0.75.
- Report disagreeing positions as IUPAC ambiguity codes - see below. Off by default.
- Min het fraction - appears once ambiguity codes are on, since it does nothing otherwise. Defaults to 0.5.
- Include insertions supported by the reads - with this off, the output stays in the reference coordinate frame.
- Exclude secondary alignments - on by default, and the one flag this dialog sets for itself: unchecking it includes secondary reads even on a track whose own filter drops them. Every other flag is inherited from the track's Filter by settings, which by default exclude unmapped, QC-fail and duplicate reads — so a track you have re-filtered feeds that same filter to the consensus. Reads stored without a sequence are always skipped, having no bases to vote with.
Ambiguity codes
By default a position where the reads disagree is reported as N, which loses the distinction between "no coverage" and "two alleles at similar depth". Turning on IUPAC ambiguity codes reports such positions as the standard code for the bases involved, so a clean heterozygous A/G site reads as R instead of N.
A base joins the call when its support is at least the minimum het fraction of the winning base's. Lowering that value folds in weaker alleles and produces more ambiguity codes.
Positions where a base and a deletion are both well supported have no IUPAC code, since the notation only covers bases. These are written as a lowercase base, following the same convention samtools uses.
A genuine four-way split is also written N, because that is what IUPAC spells for all four bases. A tetraploid site with four real alleles and a position with no usable signal are indistinguishable in this notation.
Exporting
The called sequence is shown in the panel and can be copied to the clipboard or downloaded as FASTA.
Positions where the consensus differs from the reference are also collected as variants, downloadable as VCF or openable as a variant track in the current session. The button shows how many were found.
The variant output only contains definite calls. N positions are not variants, and with ambiguity codes on, positions that resolved to a code are omitted too, since a VCF alternate allele must be a definite sequence. A region can therefore produce fewer variants with ambiguity codes on than off.
A deletion running from the very first position of the selection is also left out. A VCF deletion record is anchored on the reference base in front of it, and that base is outside the selection, so the deletion cannot be written without misstating which bases it covers. It is still reflected in the FASTA, which omits the deleted bases. Extend the selection to the left to get the record.
Relationship to samtools
The calling matches samtools consensus -a -m simple at the same settings, and
the ambiguity mode corresponds to its --ambig option, with one deliberate
difference: samtools folds in at most one runner-up allele, which is enough for
a diploid heterozygous site but drops a real third or fourth allele in a
polyploid, pooled, or mixed sample. Here every allele clearing the het fraction
joins the call, so such a site can report as V, H, D, or B rather than being
truncated to two alleles.
This is quality-independent calling: it counts and weights reads without using base quality scores or a genotype likelihood model. It answers "what do these reads say is here"; for genotypes and confidence scores, use a dedicated caller.