Genome-wide association studies (GWASs) is an effective strategy to identify susceptibility loci for human complex diseases. However, missing heritability is still a big problem. Here we developed a pipeline named interaction analyses within chromatin regulatory circuitry (IACRC), to identify genetic variants impacting complex traits. IACRC would automatically selected chromatin regulatory circuits regions with Hi-C datasets, enhancer data, and super enhancer regions. SNP × SNP interaction analyses were then performed in regions within chromatin regulatory circuits.
This software is distributed under the terms of GPL 2.0
https://github.com/studentyaoshi/IACRC
Shan-Shan Dong, Shi Yao, Yi-Xiao Chen, Yu-Jie Zhang, Hui-Min Niu, Yan Guo, Tie-Lin Yang
Key Laboratory of Biomedical Information Engineering of Ministry of Education, School of Life Science and Technology, Xi'an Jiaotong University, Xi'an, Shaanxi Province, 710049, P. R. China
yangtielin@mail.xjtu.edu.cn
Shan-Shan Dong and Shi Yao
You can contact dongss@xjtu.edu.cn or studentyaoshi@stu.xjtu.edu.cn when you have any questions, suggestions, comments, etc. Please describe in details, and attach your command line and log messages if possible.
R;
Python2.7;
Bedtools;
PLINK;
GCTA if you would like to use principal components as covariates;
and METAL if you would like to do meta analysis.
cd /your/local/path/pipeline
There are five original files you need to prepare before starting calculation, according to the disease you are studing. These files including gene information, Hi-C pairs, enhancer annotation, gene-enhancer information and gene-super enhancer information. The description and data format are as follows.
Gene information contains four columns divided by tab including the gene name, chromosome, the start and end position of gene without header:
OR4F5 chr1 69091 70008
AL627309.1 chr1 134901 139379
OR4F29 chr1 367640 368634
OR4F16 chr1 621059 622053
AL669831.1 chr1 738532 739137
...
Hi-C pairs contains six columns divided by tab including the chromosome, the start position and the end position of the two regions without header. We downloaded the autosomal interactions in 9 cell lines (NHEK, K562, Islet, IMR90, HUVEC, HMEC, HELA, GM12878, and A549) in /your/local/path/files/hic/ from 4DGenome and some recent papers, chromatin looping, 3D Genome Architecture, long-range promoter, cis-regulatory landscape, Genome Architecture Links long-range interactions, and Hi-C from A549.
chr1 1050000 1060000 chr1 1180000 1190000
chr1 1585000 1590000 chr1 1645000 1650000
chr1 1710000 1715000 chr1 1835000 1840000
chr1 2120000 2130000 chr1 2310000 2320000
chr1 2130000 2135000 chr1 2515000 2520000
...
Enhancer annotation is .bed file contains the chromatin states downloaded from Roadmap. The format of .bed can be found here.
In our paper, we used the chromatin states of GM12878 and '7_Enh' annotation in the forth column as input.
We already downloaded and converted the enhancers in 9 cell lines (NHEK, K562, Islet, IMR90, HUVEC, HMEC, HELA, GM12878, and A549) according to Hi-C pairs in /your/local/path/files/HMM/.
chr10 174200 174400 7_Enh 0 . 174200 174400 255,255,0
chr10 175600 176000 7_Enh 0 . 175600 176000 255,255,0
chr10 187400 188400 7_Enh 0 . 187400 188400 255,255,0
chr10 228600 229000 7_Enh 0 . 228600 229000 255,255,0
chr10 264600 265200 7_Enh 0 . 264600 265200 255,255,0
...
Gene-enhancer contains two columns divided by tab including the gene name and the enhancer region. We used enhancer-gene interactions predicted by PreSTIGE and the data can be downloaded here. We downloaded and converted the gene-enhancer interactions in six cell lines (GM12878, K562, HUVEC, HMEK, NHEK, HELA) for convenience in /your/local/path/files/gene_enhancer/.
SLC25A28 chr10:101314742-101317123
SLC25A28 chr10:101359480-101360496
SLC25A28 chr10:101375165-101375877
SLC25A28 chr10:101387579-101391927
COX15 chr10:101409861-101410384
...
Gene-super enhancer contains three columns divided by tab including chromsome, the start and the end position of the super enhancer. We used the super enhancer data from this paper. We downloaded and converted the super enhancer data in 86 tissues/cell lines in /your/local/path/files/super_enhancer/ which the detail of the tissues/cell lines can be found in /your/local/path/files/super_enhancer/86-Cell.anno.
chr1 68148186 68237688
chr2 37850717 37940881
chr3 16099720 16182650
chr4 7804602 7914384
chr5 172284905 172383585
...
These five files need to be moved to /your/local/path/original/ after preparation and then you need to change the file names in /your/local/path/original/files accordingly. For example, the following file indicates that the five files are gene_information, GM12878_HIC.bed, E116_enhancer.bed, GM12878_gene_enhancer.txt, obesity.SE.bed.
cat /your/local/path/original/files
Original files:
GENE_POSITION gene.test
HIC hic.test
HMM_ENHANCER HMM_enhancer.test
GENE_ENHANCER enhancer.test
SUPER_ENHANCER SE.test
You need to set the absolute path of the software accordingly in /your/local/path/original/files before calculating.
Path of software:
R_path /opt/software/R-3.3.2/bin/R
Python_path /opt/software/python/bin/python
Bedtools_path /opt/software/bedtools2-2.25.0/bin/bedtools
Plink_path /opt/software/plink_1.90/plink
We have made the analyses being able to run on multiple CPUs. Specify the number of threads (1-20) in /your/local/path/original/files which the program will be running on multiple CPUs.
Default value is 5.
CPU number:
Thread 5
sh getcircuit.sh
You now get a folder named /your/local/path/result which contains different types of chromatin regulatory circuitry according to the paper.
Please wait until this job finishes. This step can be long. So, it is recommend to use smaller files to test this software first.
sh getsnppairs.sh name genotype
- Name: a character of the name of your data.
- Genotype: the absolute path of the binary genotype data, including .bed .bim .fam, the format can be found here.
sh cat_snppairs.sh name
You now get a file named /your/local/path/result/name/name.pairs.uniq which contains the SNPs pairs in circuitry.
There are some parameters you need to change in /your/local/path/original/files before calculate, which are IMR_threshold, CR_threshold, MAF_threshold, HWE_threshold and LD_threshold. For example:
Threshold:
IMR_threshold 0.05
CR_threshold 0.95
MAF_threshold 0.05
HWE_threshold 0.001
LD_threshold 0.5
-
IMR_threshold: a number range from 0 to 1, indicates the threshold of individual missing rate that need to be excluded from this calculation. Default value is 0.05, which means the individuals with IMR over 0.05 will be excluded.
-
CR_threshold: a number range from 0 to 1, indicates the threshold of call rate of SNPs that need to be excluded from this calculation. Default value is 0.95, which means the SNPs with CR less than 0.95 will be excluded.
-
MAF_threshold: a number range from 0 to 0.5, indicates the threshold of MAF of SNPs that need to be excluded from this calculation. Default value is 0.05, which means the SNPs with MAF less than 0.05 will be excluded.
-
HWE_threshold: a number range from 0 to 0.05, indicates the threshold of HWE p-value of SNPs that need to be excluded from this calculation. Default value is 0.001, which means the SNPs with HWE p-value less than 0.001 will be excluded.
-
LD_threshold: a number range from 0 to 1, indicates the threshold of r2 of SNP pairs that need to be excluded from this calculation. Default value is 0.5, which means the SNP pair in LD with each other (r2 < LD_threshold) will be excluded.
- Note: IMR, Individual missing rate; CR, Call rate; MAF, Minor allele frequency; HWE, Hardy-Weinberg equilibrium; LD, Linkage disequilibrium.
sh snp_purning.sh name genotype
- Name: a character of the name of your data.
- Genotype: the absolute path of the binary genotype data, including .bed .bim .fam, the format can be found here.
You now get a file named /your/local/path/result/name/name.pairs.purning which contains the SNPs pairs in circuitry after SNP pruning.
Briefly, our pipeline makes a model based on allele dosage for each SNP through R, which fits a linear regression model for continuous phenotypes or logistic regression model for categorical phenotypes in the following equation:
Y ~ β + β1*SNP1 + β2*SNP2 + β3*SNP1*SNP2 + β4*Cov1 + ... + βn*Covn + e
sh calculate.sh name genotype.keep phenotype
-
Name: a character of the name of your data.
-
Genotype: the absolute path of the binary genotype data, including .bed .bim .fam, the format can be found here.
-
Phenotype: the absolute path of the phenotype data with covariates and header divided by tab:
- Note: the
PHENOis 1/0 rather than 2/1 in phenotype for categorical phenotypes.
- Note: the
IID PHENO COV1 COV2 COV3 COV4 COV5 COV6 COV7 COV8
4806 0 2 67 0.0184 0.0019 -0.0006 -0.0112 0.0018 0
4807 0 1 67 0.0203 -0.0003 0.0062 0.0053 0.0004 1
4808 1 1 59 0.0222 -0.0078 0.0012 -0.0180 -0.0038 0
4809 1 1 67 0.0169 -0.0027 -0.0099 -0.0146 0.0001 1
4810 0 2 62 0.0116 -0.0061 -0.0180 0.00076 -0.0042 1
sh getresult.sh name
You now get the results of IACRC in /your/local/path/result/name/name.allepi after this step.
Please wait until this job finishes. This step can be long. So, it is recommend to use smaller files to test this software first.