Fit-Hi-C (or FitHiC) was initially developed by Ferhat Ay, Timothy Bailey, and William Noble January 19th, 2014. It is currently maintained and updated by Ferhat Ay (ferhatay@lji.org) and Arya Kaul (akaul@lji.org) at the Ay Lab in the La Jolla Institute for Allergy and Immunology.
The current version is named as FitHiC2 (or FitHiC 2.0) due to the addition of many new features compared to FitHiC, like:
-
finding inter-chromosomal significant interactions,
-
applying a merging filter algorithm to filter out putative bystander interactions and keep only the direct CIS chromosomal interactions,
-
Reporting the expected contact count between interacting pairs of bins, along with the raw (observed) contact count, to assess the enrichment of observed from the expected contact count.
Please use the Google Group for discussions/bug reports/analysis questions. Sending an email to fithic@googlegroups.com will also post directly to the Group.
Fit-Hi-C may be installed through one of three ways.
- Bioconda
- Github
- Pip
Out of all of the following, we recommend installing through bioconda to automatically install all dependencies.
If this is your first time using the conda distribution system, we recommend using Miniconda as your preferred conda distribution system. This is chosen because it is the most lightweight out of all of Anaconda's distribution system; however, you're welcome to use any one you would like. More information on each may be found here.
Once your conda distribution platform has been set up, you need to add the bioconda channel to access the bioinformatics recipes hosted there. If you're unfamiliar with bioconda, I highly recommend you check out the wonderful work they've done ( link here!). To set up the bioconda channel, run the following:
conda config --add channels defaults
conda config --add channels conda-forge
conda config --add channels bioconda
Afterwards, simply run:
conda install fithic
This command will automatically install a command-line executable version of Fit-Hi-C along with all of its dependencies. Please run:
fithic -V
and ensure that the version number matches the version number in the Anaconda Cloud badge at the top of this README.
Install git and then run:
git clone https://github.com/ay-lab/fithic.git
You will need the following dependencies installed to run Fit-Hi-C:
- Python 3.+
- Numpy 1.14.+
- Scipy 1.1.+
- Scikit-learn 0.19.+
- SortedContainers 2.0.+
- Matplotlib 2.2.+
This will create a direct clone of Fit-Hi-C in your working directory with the name fithic. You may now run Fit-Hi-Cv.2 by calling the fithic.py file in the fithic/ directory:
python fithic.py --ARGUMENTS
Cloning into the repository does not automatically install the command line version of Fit-Hi-C, and if you desire that functionality follow the other installation instructions.
Ensure that you have Python3 successfully installed on your computer.
Then run:
pip install fithic
After this is done, run fithic --help to ensure all necessary dependencies have been installed. Some users report that this automatically installs all dependencies, while some say it does not. If fithic does not work properly, install the necessary dependencies using pip. i.e.
pip install DEPENDENCY
A good part of any software installation is being able to run tests on the correct installation of it.
Run the command:
svn export https://github.com/ay-lab/fithic/trunk/fithic/tests/
(If you receive an error, ensure that you have svn installed correctly.) This command will generate a tests folder in your working directory. Going into that and running ./run_tests-cli.sh will automatically run Fit-Hi-C on a variety of data and if everything was installed correctly you should see a final message that Fit-Hi-C executed correctly!
Simply navigate into the repo and run ./fithic/tests/run_tests-git.sh. If everything is working fine, you will see a final message that Fit-Hi-C executed correctly!
Congratulations! If you have gotten to this point, then you have a working, fully installed version of Fit-Hi-C running on your computer. Good on you! But that was the easy part, now comes the difficult question.
How do I correctly use Fit-Hi-C to analyze my desired Hi-C dataset?
Correctly answering this question requires navigating through several basic understandings:
- What exactly is Hi-C data?
- What does Fit-Hi-C tell me about this Hi-C data?
- Why is what Fit-Hi-C tells me important?
If you feel utterly comfortable with the answers to these three questions, then feel free to skip to the next section. If you are unclear about the answer to any of the above, then read on!
While a beautiful, Latex type-set, easy-to-understand, and comprehensive document is being created, read this !
At the beginning of this README, I stated:
Fit-Hi-C is a tool for assigning statistical confidence estimates to chromosomal contact maps produced by genome architecture assays.
The phrase 'chromosomal contact maps produced by genome architecture assays' may be faithfully reduced to 'Hi-C data.' Applying that change yields:
Fit-Hi-C is a tool for assigning statistical confidence estimates to Hi-C data.
Much less scary! From the above sentence, the only real phrase that could be misinterpreted is 'assigning statistical confidence estimates.' What does that mean? Well to find out you should read the paper Dr. Ay wrote (found here)!
Fit-Hi-C tells you what contacts are significant. This is incredibly important because not all of the contacts seen in your Hi-C data are truly unexpected interactions. By assigning statistical confidence to each interaction, you will be able to determine which interactions are the most important and consequently, which ones warrant further investigation.
The -f argument is used to pass in a full path to what we deem a 'fragments file,' Each line will have 5 entries. The second and fifth fields can be any integer as they are not needed in most cases. The first field is the chromosome name or number, the third field is the coordinate of the midpoint of the fragment on that chromosome, the fourth field is the total number of observed mid-range reads (contact counts) that involve the specified fragment. The fields can be separated by space or tab. All possible fragments need to be listed in this file.
One example file would look like below (excluding the header which is not a part of input):
| chr | extraField | fragmentMid | marginalizedContactCount | mappable? (0/1) |
|---|---|---|---|---|
| 1 | 0 | 15000 | 234 | 1 |
| 1 | 0 | 25000 | 0 | 0 |
| ... | ... | ... | ... | ... |
Note: the file should be gzipped before providing as an input parameter
The interactions file contains a list of mid-range contacts between the fragments/windows/meta-fragments listed in the first file above. Each fragment will be represented by its chromosome and midpoint coordinate. Each line will have 5 fields. The first two will represent first fragment, the following two will represent the second and the fifth field will correspond to number of contacts between these two fragments. The fields can be separated by space or tab. Only the fragment pairs with non-zero contact counts are listed in this file.
One example file would look like below (excluding the header which is not a part of input):
| chr1 | fragmentMid1 | chr2 | fragmentMid2 | contactCount |
|---|---|---|---|---|
| 1 | 15000 | 1 | 35000 | 23 |
| 1 | 15000 | 1 | 55000 | 12 |
| ... | ... | ... | ... | ... |
Note: the file should be gzipped before providing as an input parameter
A full path to an output directory of your choice. If it is not already created, it creates if for you.
Numerical value indicating resolution of fixed-size dataset being analyzed. If non-fixed size data being studied, set -r 0.
Accepts - a fullpath to a bias file generated by ICE or Knight-Ruiz normalization for Fit-Hi-C with the following format:
| chr | midpoint | bias |
|---|---|---|
| 1 | 20000 | 1.061 |
| ... | ... | ... |
Default - None
Description - Bias files help Fit-Hi-C accurately generate statistical significance estimates. If you have it, use it!
Note: the file should be gzipped before providing as an input parameter
Accepts - Number of spline passes.
Default - 1
Description - Increasing it beyond 2 is unlikely to affect Fit-Hi-C's output significantly. If you don't understand what spline fit means then you have not read the paper!
Accepts - integer representing number of equal occupancy bins you would like Fit-Hi-C to bin your data with
Default - 100
Description - used for spline fitting
Accepts - integer representing the minimum number of hits per locus that has to exist to call it mappable
Default - 1
Description - Increasing it leads to more stringent requirements for treating an interaction as reasonable. Decreasing it leads to less stringent requirements. If you have extremely high resolution data, it may help to bump this up.
Accepts - String representing prepending information for output files
Default - fithic
Description - Name of the library that is to be analyzed.
Accepts - Integer representing upper bound for the intrachromosomal interactions to be considered in base pairs.
Default - -1 (no limit)
Description - Highly recommended to bound the intrachromosomal interactions being considered.
Accepts - Integer representing lower bound for the intrachromosomal interactions to be considered in base pairs.
Default - -1 (no limit)
Description - Highly recommended to bound the intrachromosomal interactions being considered.
Accepts - no argument
Default - None (no plots)
Description - Use if plots of spline fitting are desired. Unfortunately, different matplotlib versions are unstable in different ways. If you're getting an error, I suggest trying to run Fit-Hi-C without this option and see if that helps.
Accepts - 'interOnly', 'intraOnly', 'All'
Default - intraOnly
Description -
interOnly is used if you would only like to analyze interchromosomal interactions.
intraOnly is used if youd would only like to analyze intrachromosomal interactions.
All is used if you would like to analyze inter and intrachromosomal interactions.
While you may now be thinking, "Why would I ever not choose 'All'? More analysis is better!" It is not this simple. Since you are adding significantly more interactions when you analyze interchromosomal and intrachromosomal interactions in tandem, qvalues will be depressed across the board. In addition, few to no datasets are at a high enough resolution to find significanct interchromosomal interactions.
Accepts - float value of lower bound for bias values
Default - 0.5
Description - bias values below this number will be discarded
Accepts - float value of upper bound for bias values
Default - 2
Description - bias values above this number will be discarded
Accepts - No arguments
Default - None
Description - Prints version number. Check to make sure this is the latest version based on version.log file here
Accepts - No arguments
Default - None
Description - prints help message with all options
Each step of Fit-Hi-C, the number of which is user-defined through the -p flag, generates two output files. For step N and library name prefix denoted by ${PREFIX} the two output files will have the following names:
- ${PREFIX}.fithic_passN.txt
- ${PREFIX}.spline_passN.significances.txt.gz
The first file will report the results of equal occupancy binning in five fields. An example of which is shown below:
| avgGenomicDist | contactProb | stdErr | numLocusPairs | CCtotal |
|---|---|---|---|---|
| 20077 | 2.38e-05 | 2.11e-06 | 210 | 19574 |
| 20228 | 1.88e-05 | 1.44e-06 | 268 | 19662 |
| ... | ... | ... | .. | ... |
The second file will have the exact same lines as in the input file that contains the list of mid-range contacts. This input file had 5 fields as described above. The output from each step will append the following columns to these fields:
-
p-value: p-value of the corresponding interaction, as computed by the binomial distribution model employed in FitHiC.
-
q-value: q-value or FDR obtained by applying Benjamini-Hochberg correction to the p-values.
-
bias1: Bias value of the first interacting fragment.
-
bias2: Bias value of the second interacting fragment.
-
ExpCC: Expected contact count of the current interaction, computed using the raw contact count, spline fit probability of the raw contact count (with respect to the loop distance), and the given bias values. Enrichment of the raw (observed) contact count with respect to the expected contact count is reflected in the q-value.
| chr1 | fragmentMid1 | chr2 | fragmentMid2 | contactCount | p-value | q-value | bias1 | bias2 | ExpCC |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 15000 | 1 | 35000 | 23 | 1.000000e+00 | 1.000000e+00 | 1 | 1.2 | 22 |
| 1 | 15000 | 1 | 55000 | 12 | 2.544592e-02 | 1.202603e-01 | 1.1 | 0.9 | 4 |
| ... | ... | ... | ... | ... | .. | .. | .. | .. | .. |
These utilities are provided as part of Fit-Hi-C (/fithic/utils/) to aid in certain common pre-processing/post-processing steps. They are as follows:
- HiCKRy.py (Pre-processing. Generates --bias calculation)
- HiCPro2FitHiC.py (Pre-processing. Generates --interactions, --fragments, and --bias inputs)
- createFitHiCFragments-fixedsize.py (Pre-processing. Generates --fragments input)
- createFitHiCFragments-nonfixedsize.sh (Pre-processing. Generates --fragments input)
- validPairs2FitHiC-fixedSize.sh (Pre-processing. Generates --interactions input)
- createFitHiCContacts-hic.sh (Pre-processing. Generates --interactions input from .hic output)
- visualize-UCSC.sh (Post-processing. Visualizes Fit-Hi-C interactions on the UCSC Genome Browser)
- createFitHiCHTMLout.sh (Post-processing. Generates HTML page describing Fit-Hi-C run)
- merge-filter.sh (Post-processing. Filters Fit-Hi-C interactions and merges nearby ones using FANCY GRAAAAAAAAAAAPH magic)
- merge-filter-parallelized.sh (Post-processing. Filters Fit-Hi-C interactions and merges nearby ones using FANCY GRAAAAAAAAAAAPH magic + parallelizes per chr)
Regardless of the implementation, we strongly recommend the use of a normalization method in order to have meaningful results for further analysis. The only way for Fit-Hi-C to utilize data from Hi-C normalization is through the bias files. As long as the bias value are scaled to have an average of 1 and high values represent loci with higher overall raw counts, Fit-Hi-C will be able to use them in significance assignment.
HiCKRy is an in-house version of Hi-C contact map normalization using the Knight-Ruiz algorithm for fast matrix balancing. It takes three arguments:
-i,--interactions Path to the interactions file to generate bias values. Required.
-f, --fragments Path to the interactions file to generate bias values. Required.
-o, --output Full path to output the generated bias file to. Required.
-x, --percentOfSparseToRemove Percent of sparse low contact count loci to remove. The default value is 0.05.
It then outputs a bias file in the format of Fit-Hi-C's -t input option.
NOTE: one should always check the output of the bias file for correctness. If one finds a large number of -1 values or exceedingly small values then it is likely that the KR algorithm failed to converge. To rectify this we recommend rerunning HiCKRy with a different value in -x. This is similar to the approach employed by the Aiden lab in Juicebox see here. In the future, we might make this step automatically rerun until it converges. Until then, please check output!
HiC-Pro is a common Hi-C mapping tool used to extract information from the raw reads after the Hi-C assay is run. The following script enables the generation of Fit-Hi-C input directly from HiC-Pro's output.
It takes the following arguments:
-i MATRIX, --matrix MATRIX Input matrix file with raw contact frequencies. Required.
-b BED, --bed BED BED file with bins coordinates. Required.
-s BIAS, --bias BIAS The bias file provided after IC normalization.
-o OUTPUT, --output OUTPUT Output path.
-r RESOLUTION, --resolution RESOLUTION Resolution of the matrix.
The output is the contact maps and fragments file in the format of Fit-Hi-C.
Generates the fragments file if using a fixed-size resolution with your Hi-C data.
The script takes the following arguments:
--chrLens Path to a file describing chromosome lengths of the model organism. Required.
--resolution Resolution of dataset being studied. Required.
--outFile Full path to the output file desired.
Output is a fragments file in the format of Fit-Hi-C.
A bash script to generate the fragments file if the Hi-C data is RE-digested. Note, order of arguments is critical.
bash createFitHiCFragments-nonfixedsize.sh [outputFile] [RE] [fastaReferenceGenome]
[outputFile] A desired output file path. Required.
[RE] Either the name of the restriction enzyme used, or the cutting position using “^”. For example, A^AGCTT for HindIII. Required.
[fastaReferenceGenome] A reference genome in fasta format. Required.
A bash script to generate the contact maps input for Fit-Hi-C from a valid pairs file. Note, order of arguments is critical.
bash validPairs2FitHiC-fixedSize.sh [resolution] [libraryName] [validPairsFile]
[resolution] The resolution of the dataset being studied. Required.
[libraryName] The prefix of the file generated. Required.
[validPairsFile] A textfile containing the validPairs. Required.
A bash script to convert Fit-Hi-C output into visualization input for UCSC's Genome Browser in 'interact' format.
DESCRIPTION:
bash visualize-UCSC.sh [inputFile] [outputFile] [QvalThresh]
[inputFile] Input Fit-Hi-C file to visualize
[outputFile] Output file for UCSC to visualize
[QvalThresh] Q-value threshold to filter Fit-Hi-C interactions at
A bash script to generate an HTML report of the Fit-Hi-C run. Note, works best if Fit-Hi-C was run with --visual option.
bash createFitHiCHTMLout.sh [Library Name] [No. of passes] [Fit-Hi-C output folder]
[Library Name] The library name (-l option) used during Fit-Hi-C’s run
[No. of passes] The number of spline passes conducted by the Fit-Hi-C run
[Fit-Hi-C output folder] Path to the output folder for that Fit-Hi-C run (-o option)
A bash script to create Fit-Hi-C contacts from .hic files.
bash createFitHiCContacts-hic.sh [Juicer's dump command] [chr1] [chr2] [Output file name]
[Juicer's dump command] Full path to the output of Juicer's dump command
[chr1] Chromosome 1 of the argument used in Juicer's dump command
[chr2] Chromosome 2 of the argument used in Juicer's dump command
[Output file name] Name of output file
A bash script to merge nearby significant interactions and filter Fit-Hi-C output
bash merge-filter.sh [inputFile] [resolution] [outputDirectory] [fdr]
[inputFile] Input file of Fit-Hi-C interactions
[resolution] Resolution used
[outputFile] Output file to dump output to
[fdr] False Discovery rate to use when subsetting interactions
[utilities] Full path to utilities folder (folder where CombineNearbyInteraction.py is)
A bash script to parallelize merge nearby significant interactions and filter Fit-Hi-C output. Note - you will have to modify the actual script contents to assume your cluster config/ organism.
bash merge-filter-parallelized.sh [inputFile] [resolution] [outputDirectory] [fdr]
[inputFile] Input file of Fit-Hi-C interactions
[resolution] Resolution used
[outputDirectory] Directory to dump output to
[fdr] False Discovery rate to use when subsetting interactions
[utilities] Full path to utilities folder (folder where CombineNearbyInteraction.py is)
If Fit-Hi-C was used in your analysis, please issue the following citations:
-
Arya Kaul, Sourya Bhattacharyya & Ferhat Ay 2020. "Identifying statistically significant chromatin contacts from Hi-C data with FitHiC2." Nature Protocols. 15:991-1012, 2020. doi: 10.1038/s41596-019-0273-0.
-
Ferhat Ay, Timothy L. Bailey, William S. Noble. 2014. "Statistical confidence estimation for Hi-C data reveals regulatory chromatin contacts." Genome Research. 24(6):999-1011, 2014. doi: 10.1101/gr.160374.113.
Copyright (c), 2012, University of Washington
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.