Change matrix resolution
hicMergeMatrixBins is used to decrease the resolution of a matrix. With this tool, you can for example create out of a 100 kb contact matrix a 1000 kb one:
Number of bins to merge = 10
100 kb * 10 = 1000 kb = 1 Mb
Depending on the downstream analyses to perform on a Hi-C matrix generated with HiCExplorer, one might need different bin resolutions. For example using hicPlotMatrix to display chromatin interactions of a whole chromosome will not produce any meaningful vizualisation if it is performed on a matrix at restriction sites resolution (unmerged). Furthermore, the higher the resolution of a matrix, the more detailed it is, which can make it difficult to interpret, especially if the read depth of the Hi-C data is not high enough. hicMergeMatrixBins address these issues by merging a given number of adjacent bins to reduce Hi-C matrices resolution.
Usage
To limit the loss of information, it is mandatory to perform hicMergeMatrixBins on matrices prior to any correction and any other bin merging (direct output from hicBuildMatrix). After bin merging, hicCorrectMatrix must be used for downstream analyses requiring corrected matrices.
Output
hicMergeMatrixBins outputs a Hi-C matrix with reduced resolution.
Below, we will develop the example of a Hi-C matrix in Drosophila melanogaster that we want to display at the whole X-chromosome scale and at the scale of a 1Mb region of the X chromosome. To do this, we performed two different bin merging using hicMergeMatrixBins on an uncorrected matrix built at the restiction sites resolution using hicBuildMatrix.
Starting from a matrix with bins of a median length of 529bp (restriction enzyme resolution, here DpnII), running hicMergeMatrixBins with a number of bins to merge of 3 produced a matrix with bins of a median length of 1661bp, while hicMergeMatrixBins with a number of bins to merge of 50 produced a matrix with bins of a median length of 29798bp.
After the correction of these three matrices using hicCorrectMatrix, we plotted them using hicPlotMatrix at the scale of the whole X-chromosome and at the scale of the X:2000000-3000000 region to see the effect of bin merging on the interactions visualization.
- Effect of bins merging at the scale of a chromosome:
When observed altogether, the plots above show that the merging of bins by 50 is the most adequate way to plot interactions for a whole chromosome in Drosophila melanogaster when starting from a matrix with bins of a median length of 529bp.
- Effect of bins merging at the scale of a specific region:
When observed altogether, the plots above show that the merging of bins by 3 is the most adequate way to plot interactions for a region of 1Mb in Drosophila melanogaster when starting from a matrix with bins of a median length of 529bp.