hicFindTADs (version 3.7.2+galaxy0)

Matrix to compute on:

Compute from scratch or use precomputed data:

Minimum window length (in bp) to be considered to the left and to the right of each Hi-C bin.:

This number should be at least 3 times as large as the bin size of the Hi-C matrix.

Maximum window length (in bp) to be considered to the left and to the right of each Hi-C bin.:

This number should around 6-10 times as large as the bin size of the Hi-C matrix.

Step size when moving from minDepth to maxDepth:

The step size grows exponentially as maxDeph + (step * int(x)**1.5) for x in [0, 1, ...] until it reaches maxDepth. For example, selecting step=10,000, minDepth=20,000 and maxDepth=150,000 will compute TAD-scores for window sizes: 20,000, 30,000, 40,000, 70,000 and 100,000

Multiple Testing Corrections:

q-value:

Minimum threshold of the difference between the TAD-separation score of a putative boundary and the mean of the TAD-sep. score of surrounding bins.:

The delta value reduces spurious boundaries that are shallow, which usually occur at the center of large TADs when the TAD-sep. score is flat. Higher delta threshold values produce more conservative boundary estimations. By default, multiple delta thresholds are saved for the following delta values: 0.001, 0.01, 0.03, 0.05, 0.1. Other single or multiple values can be given.

Minimum distance between boundaries (in bp).:

This parameter can be used to reduce spurious boundaries caused by noise.

List of chromosomes to be included in the correlations

List of chromosomes to be included in the correlation 0

Calculate Topologic Associated Domains

Toplogical domains (TADs) are large mainly self-interacting domains. Chromatin interactions occur with higher frequency within a TAD as between TADs. More information.

Usage

This tool must be used on unmerged matrices (restiction enzyme resolution) produced by hicBuildMatrix and corrected by hicCorrectMatrix.

Computation details

hicFindTADs computes the TAD regions in two steps: in a first step it computes a TAD-separation score based on a z-score matrix for all bins. The z-score is defined as:

“The absolute value of z represents the distance between the raw score and the population mean in units of the standard deviation. z is negative when the raw score is below the mean, positive when above.” [Source].

Source of image

In our case the distribution describes the counts per bin of a genomic distance. In a second step the local minima of the TAD-separation score is evaluated with respect to the surrounding bins to assign a p-value. Two multiple testing corrections can be applied to filter the results: Bonferroni or the false discovery rate.

Output

hicFindTADs produces multiple outputs:

TAD boundaries positions as a BED file and TAD separation score.
TAD boundaries positions with delta, p-value and TAD separation score as GFF.
TAD domains as a BED file.
TAD seperation score as bigwig (bw), bedgraph and numpy array (npz) format. These files can be used to plot the so-called TAD insulation score or TAD separation score along the genome or at specific regions. This score is much more reliable across samples than the number of TADs or the TADs width that can vary depending on the sequencing depth because of the lack of information at certain bins, and depending on the parameters used with this tool.
Matrix with multi-scale TAD scores as a bed-matrix (bm) file that can be plotted inside pyGenomeTracks to nicely display TAD insulation score alongside Hi-C heatmap and other datasets.
Z-score matrix in h5 format that is useful to quickly test the --thresholdComparisons, --delta and --correctForMultipleTesting parameters by using the --TAD_sep_score_prefix option pointing to this zscore_matrix.h5 file (will be added in a future update).

Usage hints

It is mandatory to test multiple parameters of TAD calling with hicFindTADs before making conclusions about the number of TADs in a given sample or before comparing TAD calling between multiple conditions. In order to compare numerous TAD calling parameters at once, it is recommended to use pyGenomeTracks, below you can find a plot where multiple TAD calling parameters are displayed for Drosophila melanogaster embryos:

/repository/static/images/dbfd8373baf7c017/hicFindTADs_TAD_calling_comparison.png

We can see that the fourth set of hicFindTADs parameters with a threshold of 0.001 gives the best results in terms of TAD calling compared to the corrected Hi-C counts distribution and compared to the enrichment of H3K36me3, which is known to be enriched at TAD boundaries in Drosophila melanogaster.

For more information about HiCExplorer please consider our documentation on readthedocs.io