Skip to content

Signal and sequence

Plots for a value or a symbol at every position: read depth, a statistic in windows, methylation by strand, the reference bases, sequence logos and per-base model attribution.

How to choose

Decide first what zero means for your values: the floor they rise from, as for depth; a line they fall either side of, as for a statistic in windows; or nothing at all, as for bases and letters.

Your question Plot Build it with
How much is there at each base, counting up from zero? Coverage CoverageTrack, --coverage
Which side of its baseline does each window fall on? Windowed statistic WindowTrack, --windows
How methylated is each site, strand by strand? Methylation by strand MethylationTrack, --methylation
What are the reference bases here? Reference sequence SequenceTrack, --sequence
What does a motif look like across aligned sequences? Sequence logo LogoTrack, --logo
Which bases did a model rely on, and in which direction? Per-base attribution DynseqTrack, --dynseq with --with-sequence

Plots

  • A read depth profile over the rpoB locus with a dropout, above a reference band, the rpoB gene model and variant calls

    Coverage One value per base drawn up from a floor of zero, each pixel column showing its maximum, or its minimum when you are hunting dropouts.

  • pN/pS on a log2 scale and GC skew in windows along forty kilobases, each drawn either side of its own baseline, with the two sides of each line in different colours

    Windowed statistic A statistic per window drawn either side of the line it is read against, such as pN/pS, GC skew or Tajima's D.

  • Dam methylation at GATC sites around the E. coli origin of replication, forward-strand calls above the line and reverse-strand calls below, with the reverse strand close to zero inside oriC

    Methylation by strand The methylated fraction at each site with each strand in its own lane, calls from too few reads dropped and counted, and the rest faded by depth.

  • Sixty bases of the rpoB locus at base resolution: a depth profile, the reference drawn as coloured letters, and variant calls

    Reference sequence The bases as letters when there is room, coloured blocks when there is not, and a prompt to zoom in once they are too thin to draw.

  • One eight-column DNA motif drawn three ways: as probabilities, as information in bits, and as enrichment above the line with depletion below it

    Sequence logo A motif from aligned sequences or a weight matrix, scored as probability, information or enrichment against a background, with depletion hanging below the line.

  • Four columns scored five ways against a background, as log odds, KL divergence, difference, ratio and odds ratio, each putting the emphasis on a different column

    Logo scores compared Seven scores to choose from, and the choice decides which column reads loudest, not only how the logo looks.

  • The same motif proportions from five, fifty and five hundred sequences, each drawn raw and shrunk: the raw logos look alike, while the shrunk ones grow with the number of sequences

    Stabilised logo Each column shrunk towards the background, more when it rests on fewer sequences, so five sequences do not look as certain as five hundred.

  • A sequence logo whose symbols are three-letter amino acid codes

    Logo of any alphabet Symbols are strings, so three-letter residue codes and k-mers stack the way bases do.

  • Junction arcs over an RNA depth profile with per-base model attribution beneath, and a close-up of a promoter motif where each base is a letter sized by its score, above or below the line

    Per-base attribution One signed score per base drawn as the base itself: tall where the model relied on it, hanging below the line where it pulled the prediction down.