Code for finding putative enhancers using Hi-C data

python htad-chain.py <config file>

For running the example use python htad-chain.py examples/himr90.chr20.conf from the repo directory.

• res resolution of the Hi-C file, in bases (40000)
• win interaction distance cutoff in bases, usually 2000000
• chrname chromosome name, used for genes, size and output. should use 'chr1'..'chrX'
• chrsize path to bed file of chromsome lengths (examples/hg19.size.bed)
• output_prefix output prefix for this conf file (hIMR90)
• output_dir path in which to store the output files (examples/output)
• input_matrix path to input Hi-C matrix for the chromosome (see format specifications below)
• genes_file path to bed file describing genes (examples/hg19.genes.bed)

For a functioning example config file consult examples/himr90.chr20.conf

The code was used on a Linux machine. It has scripts in matlab, python and perl, so the minimal requirements would be -

• Matlab
• python2.7
• Perl
• Unix tools - cut, sed, pushd, popd (typically installed by default)

• htad-chain.py Main command line interface for PSYCHIC
• matlab/ Main matlab files
• domaincall_software/ Slightly adapted files from Hi-C DomainCaller
• insulation/ Domain caller from Crane et al. and additional scripts Insulation Score DomainCaller
• examples/ Example files, contains config, Hi-C matrix, chromsome sizes and gene list

input_matrix should be in a csv or tab-delimitered file, specifying the Hi-C data (for a given chomromsome). The first column could be either empty or contain the names of each genomic segment (in a fixed stride, as specified in the res variable). The rest of the matrix should be symmetric, with each cell (< i,j >) containing the number of contacts between the matching segments in the chromosome. These data could be previously normalized to account for various Hi-C biases, and is assumed to be symmetric. See example under examples/hIMR90.chr20.matrix.txt.

The program outputs multiple intermediate files, and final enhancers files.

• .enh_p.bed bed file of over-represented pairs with FDR value < p, each line is of the format [chr start end], [gene, distance to enhancer, FDR, p-value, expected # of interactions, observed # of interactions]
• .model.estimated.params.bed the power-law parameters for the model, each line represents a TAD / merge / Sky Specifically, fields include [ chr start end ], class { TAD / merge / Sky }, parameters for first (or only) power-law segment [ slope, intersect, RMSE ], same for second power-law segment (or NaN), transition point between two models, in log2(bp).

Temporary files

• .enh_rand.bed set of random interactions with promoters, used as control for near-promoter enrichment values
• .domains, domains.txt the domains found be the specified TAD-calling algorithm (DI/Insulation Score)
• .domains.new.txt refined domains found by the program
• .7col,.DI,.HMM used by Directionality Index domain caller
• .prob.{bg,tad}.matrix.txt,supersum.txt the probabilstic parameters, based on the given domains
• .model.estimated.matrix.txt the model, estimated using the data, in tab deilimeted format
• .llr.txt the log-likelihood ratio of observed counts and the model
• .hierarchy.bed the constructed hierarchy of the domains
• .mdump* files are the outputs of executed matlab functions, for debugging.
• .fixed_matrix the input matrix converted to the desired format

The prefix specified in the configuration file is prepended to the mentioned names output_prefix and chrname