Metaplot over a feature set

Identical to its reference in the validationdeepTools’ computeMatrix and plotProfile (Ramírez et al. 2016). doi:10.1093/nar/gkw257

In the app

Analysis panel, under 1D tasks: Metaplot over a feature set

Command

chromosomix metaplot

Reads

Tracks and interval files, with or without a contact map

Also known as

profile, aggregate, average, TSS, deeptools

Metaplot over a feature set averages one or more tracks in a window about every feature of a set, such as ATAC-seq about every transcription start site, into a profile with a band, over a heatmap of every feature.

Running it

Load one or more tracks and a feature set, such as a BED, narrowPeak or GTF (Tracks and features). Open Metaplot over a feature set on the 1D tasks card of the Analysis panel, tick the tracks and choose the features; no map is needed. The drawer shows a line per track, and per group when you rank the set into groups, each over its heatmap.

The same run from a shell:

chromosomix metaplot [<map>] --tracks <tracks.bw> --features <features.bed> [--<parameter> <value> …] --out <dir>

Parameters

mode chooses the window. reference point is a fixed flank either side of each feature’s start, centre or end, under anchor on, in columns resolution base pairs wide. scaled region resamples every feature’s body to body columns columns between its flanks, so features of every length share one axis. Leave respect strand on for anything stranded, or an asymmetric signal cancels itself out.

rank by sorts the features by a numeric column of the set, and then draws one line, the top N, the bottom N, or K groups of equal size: the check that the strongest features look like something and the weakest like nothing. z-score tracks first puts tracks in different units on one axis.

Parameter	Takes	Default	What it does
General
tracks--tracks <file,…>	files: a signal track	—	—
features--features <file>	a file: a feature set	—	—
mode--mode <value>	reference or scaled	reference	reference — a fixed flank around one point; scaled — every feature body resampled to one width.
anchor on--at <value>	start, centre or end	start	—
flank--flank <n>	a number, at least 0, in bp	2000	Read this far either side of the anchor.
resolution--step <n>	a number, at least 1, in bp	50	One column of the plot per this many base pairs.
body columns--body <n>	a whole number, at least 10	100	Read only under other settings of this task.
respect strand--stranded / --no-stranded	on or off	on	A minus-strand feature is read right to left, or an asymmetric signal cancels itself.
Ranking
rank by--rankBy <column>	a numeric column of the set --features names; optional	not used	—
then--subset <value>	all, top, bottom or groups	all	all — rank only; still one average; top — top N; bottom — bottom N; groups — one line per equal-sized group. Read only under other settings of this task.
N--count <n>	a whole number, at least 1	500	Read only under other settings of this task.
groups--groups <n>	a whole number, 2 to 10	4	Read only under other settings of this task.
Summary
z-score tracks first--zscore / --no-zscore	on or off	off	Centre and scale each track by its own mean and SD over this run’s windows, so tracks in different units share one axis.
average--average <value>	mean, median or trimmed	mean	—
band--band <value>	bootstrap, sd or none	bootstrap	bootstrap — bootstrap CI; sd — SD.
bootstrap draws--iterations <n>	a whole number, 20 to 2000	200	—
seed--seed <n>	a seed, a whole number	1	The same seed gives the same band twice.
per-feature heatmap--heatmap / --no-heatmap	on or off	on	The mean of a bimodal set is a line that describes nothing.

What it writes

• curve: a curve: TSV
• matrix: a matrix, a pileup or a heatmap: TSV

On the command line every file goes into --out as <file stem>.metaplot.<output>.<ext>, and each TSV opens with # lines naming the build, the file and every parameter that produced it.

Reading the result

The profile

Each column is the average, over the features that reach it, of the track’s mean over the column’s bases; a feature with no value there is left out of that column, never counted as zero. The x axis is the distance from the anchor in base pairs. In a scaled window it is marked Start and End where every feature’s body begins and ends; between them, the share of the way through the body (25%, 50%, 75%, or tenths); outside them, the distance from the nearer edge, -1 kb before the start and +1 kb after the end, as deepTools labels the same plot. A wider figure or a window wheeled into one part of the axis takes finer steps, and the pointer reads positions the same way (37.4%, -150 bp). The heatmap under it marks the flanks’ ends, Start, 50% and End where they fall.

The band is by default a 95 % bootstrap interval over the features: how sure the average is. SD draws the spread of the features instead, how alike they are. A median or a trimmed mean, of the middle 80 %, keeps a few extreme features from carrying the line.

The heatmap

One row per feature, strongest first by its mean over every column of every track, in the same order in every heatmap of the run. Colours run from zero to the 99th percentile, symmetric about zero for a track that goes negative, so one extreme feature does not wash out the rest; a feature no track covers sits at the bottom.

Read it before the line: a line that averages a strong block of rows and an empty one describes neither, and ranking the set into groups shows both.

What to watch for

• In a scaled window, a feature shorter than the number of body columns, or whose window is clipped away at a sequence end, is dropped; a note says how many.
• Near the end of a sequence a reference-point window is clipped: the columns past the end are empty.
• Every feature is a row, so a gene whose GTF transcripts share a start counts once per transcript.

Compared with deepTools

The columns are computeMatrix‘s: in the validation, the profile and every heatmap row are deepTools’, within the two tools’ rounding. The one difference is where the rows no track covers go: last here, first in deepTools.