MMCT-Loop can only run on UNIX or Mac operating systems.
If you are familiar with conda, you can run these commands to start MMCT-Loop by creating a new conda environment:
conda create -n MMCT-Loop python=3.7
conda activate MMCT-Loop
conda install matplotlib=3.5.3
conda install seaborn=0.12.0
conda install joblib=1.1.1
conda install -c conda-forge hnswlib=0.6.2
conda install scipy=1.7.3
Or any other Python(>=3.7) environment that meet the following python package version requirements is ok.
matplotlib >= 3.5.3
seaborn >= 0.12.0
joblib >= 1.1.1
hnswlib >= 0.6.2
scipy >= 1.7.3
All the running steps of MMCT-Loop are written in run_MMCT-Loop.sh.
The mandatory parameters of this file are as follows, just configured and run bash run_MMCT-Loop.sh.
Initially, we have configured the run_MMCT-Loop.sh file to provide loop predictions on the example data.
-out_prefix The prefix of the output file name.
-in_valid_bedpe Processed valid pairs in '.bedpe' or '.bedpe.gz' format, multiple files use ',' split.
-cpu The number of CPU used to run the job, default is 1, set -1 to use all cpus available.
(set as the number of chomes is recommanded)
-M The maximum number of adjacency edges will be established for each point in building HNSW network
stage. Default is 32.
-k Query k approximate nearest neighbors for each point in HNSW query stage. Default is 4.
-ef The maximum length of the array of cached nearest neighbor relationships in HNSW query stage.
Default is 500.
-bs The maximum steps from each start point to construct base graph before clustering. Default is 5.
-minPts The minimum count of PET points per cluster required after clustering. Default is 4.
-dis_arr A list of required anchor distances between the two candidate loops to be compressed.
Default is '(-100000 0 100000 1000000)'. (Negative distance indicate the required length of overlap)
-merge_pets The file name of PETs used to merge candidate loops.
This function support users to calculate loops ES, FDR, Poisson_p-value for each candidate loop in bedpe format.
Configure follow parameters and just run bash run_estimate.sh.
-in_valid_bedpe Processed valid pairs in '.bedpe' format.
-candidate_loops_bedpe Candidate loops file(in 'bedpe' format) that need to be estimated.
Both input valid pairs (PETs) and output loops are of '.bedpe' format, which contains at least six of the following columns:
chr1 start1 end1 chr2 start2 end2
while the output of test model with '.loop' format in MMCT-Loop evaluation part contains following nine columns:
chr1 start1 end1 chr2 start2 end2 ES FDR Poisson_p-value
| Filename | Function |
|---|---|
| ./scripts/bedpe_merge.cpp | Merge two loop files in '.bedpe' format according to the distance or overlap between the two loops. |
| ./scripts/loop2csv.py | Filter out loops whose significance is 1 (with high reliability) in the '.loop' format file and convert it to a table in csv format. |
| ./scripts/merge_I_S.cpp | Filter the effective loops in inter- and intra-ligation loops, extract the main features and merge them. |
| ./src/plot.py | Plot the estimated fragment size and the finger print for quality control. |
| ./src/model.py | Calculation of ES, FDR and Poisson_p-value. |
| ./src/test_I.py ./src/test_S.py |
Calculate the eigenvalues of the respective loops for clusters of inter- and self-ligation loop types. |