5. Obtain the same metrics for muliple samples

5.1 Calculation

h1d multisamples IS ./test_data/multisample1/metadata.txt \
		50000 chr21 -o multisamples_metrics

The output would be multisamples_metrics.csv:

chr start end Control1 Control2 Treat1 Treat2 ...
... ... ... ... ... ... ... ...
chr21 23900000 23950000 0.600162 0.658727 0.734661 0.692226 ...
chr21 23950000 24000000 0.610387 0.674535 0.721764 0.70977 ...
... ... ... ... ... ... ... ...

5.2 Parameters

$ h1d multisamples
usage: h1d multisamples [-h] [--datatype DATATYPE]
                        [--samplelist SAMPLELIST] [--labels LABELS]
                        [-p PARAMETER] [-o OUTNAME] [--gt GT] [--corr]
                        [--heat] [--line] [--anova] [--discrete] [-s START] [-e END]
                        type data resolution chromosome

Required parameters:

  • type, type of 1D metrics could be one of {IS,CI,DI,SS,DLR,PC1,IES,IAS,IF}, as described

  • data, a tsv (Tab-separated) file contain name and paths for all samples (below). Used file could be contact matrix or raw .hic as introduced here

Control1 ./test_data/multisample1/Control1.chr21.matrix.gz
Control1 ./test_data/multisample1/Control2.chr21.matrix.gz
Treat1 ./test_data/multisample1/Treat1.chr21.matrix.gz
Treat2 ./test_data/multisample1/Treat2.chr21.matrix.gz
Treat3 ./test_data/multisample1/Treat3.chr21.matrix.gz
Treat4 ./test_data/multisample1/Treat4.chr21.matrix.gz

  • resolution, resolution (50000, i.e.) of given contact matrix, or choosed resolution for analyzing .hic file.

  • chromosome, selected chromosome to be analyzed.

Optional parameters:

  • --datatype, type of input data: “matrix” (default) or “rawhic”.

  • -p, parameters for one-sample metrics.

  • -o, output name, default: multisamples_metrics

  • --gt, genome table file when using raw .hic data.

5.3 Visualization

  • --corr, Plot correlation for all samples.

  • --heat, Plot raw heatmap for all samples.

  • --line, Plot line chart for all samples.

  • -s, start site of plot when using --corr and --line.

  • -e, end site of plot when using --corr and --line.

5.3.1 Correlation map of all samples

h1d multisamples IS ./test_data/multisample1/metadata.txt 50000 chr21 \
	-o multisamples_metrics --corr
RTDimport

5.3.2 Line chart or multiple samples:

h1d multisamples IS ./test_data/multisample1/metadata.txt 50000 chr21 -o multisamples_metrics \
	--line -s 24500000 -e 34500000
RTDimport

5.3.3 Heatmap or multiple samples:

h1d multisamples IS ./test_data/multisample1/metadata.txt 50000 chr21 -o multisamples_metrics \
	--heat -s 24500000 -e 34500000
RTDimport

5.3.4 Discrete heatmap

Some metrics such as PC1, can not be quantitatively compared, thus we convert it to discrete value to draw a heatmap:

h1d multisamples PC1 ./test_data/multisample2/metadata.txt 50000 chr19 -o multisamples2_metrics \
	--discrete -s 22500000 -e 32500000 -p test_data/multisample2/mm10_geneDensity50000.txt
RTDimport

5.4 Statistically comparison

As we mentioned in the manuscript, most of (except PC1 and DI) 1D metrics could be quantitatively compared. Considering most of them exhibited the unimodal distribution without obvious skew, we use ANOVA-like test to statistically compare multiple Hi-C samples. Each bin and its surrounding bins (default=2) are considered to to run such test. The obtained p-values is then adjusted via Benjamini/Hochberg method, to get the qvalue.

To run this, please specify a genomic region and type:

h1d multisamples IS ./test_data/multisample1/metadata.txt 50000 chr21 -o ptest --anova -s 24500000 -e 34500000

, where the test.txt is described above. The out put will be:

chr start end pvalue qvalue
... ... ... ... ...
chr21 24650000 24700000 1.4207106516019808e-05 5.25394882479223e-05
chr21 24700000 24750000 0.000600635194484068 0.0014715562264859666
chr21 24750000 24800000 0.00011666149851486222 0.0003412784135658656
... ... ... ... ...