LAMPLINK - detection of statistically significant epistatic interactions
The LAMPLINK can detect statistically significant epistatic interactions of two or more SNPs from GWAS data.
This software can be used in the same way as the widely used GWAS analysis software PLINK, but LAMPLINK has the additional options for the detection of epistatic interactions with LAMP, which is a multiple testing procedure for combinatorial effects discovery.
You can apply LAMPLINK to an analysis pipeline with PLINK simply by replacing plink
with lamplink
and adding the --lamp
option.
Installation
Linux binary of LAMPLINK and its C++ source code are available.
- Linux binary (v1.12) (Updated on May 19, 2017)
- Source code on GitHub
The LAMPLINK binary should be placed in the current working directory or the directory in your command path. Then, typing lamplink
or ./lamplink
will run LAMPLINK.
Demonstration
We show an example of LAMPLINK using a sample GWAS dataset in example/ (example.zip).
- lamplink_sample.map: Annotation file for each SNP locus
- lamplink_sample.ped: Sample information file such as phenotype and genotype
- phenotype_sample.txt: An example file for quantitative trait
The following command finds all of the significant combinations with Fisher’s exact test. The significance level was set as 0.05.
$ ./lamplink --file ./example/lamplink_sample --lamp --model-dom --sglev 0.05 --upper 0.5 --out example --fisher
When LAMPLINK completes successfully, the result is output in example.lamplink and example.lamp.
example.lamp:
COMBID Raw_P Adjusted_P COMB
COMB1 0.006993 0.034965 rs7817762,rs2631899,rs1841195
This file indicates that one combination of SNPs (rs7817762, rs2631899, and rs1841195) was significantly associated with the phenotype after the multiple testing correction. The raw and adjusted P-values are 0.006993 and 0.034965, respectively.
example.lamplink:
CHR SNP A1 A2 TEST AFF UNAFF P OR COMB1
8 rs7817762 2 1 DOM 5/2 1/7 0.0405594 17.5 1
8 rs2631899 2 1 DOM 5/2 2/6 0.131935 7.5 1
8 rs1841195 2 1 DOM 5/2 1/7 0.0405594 17.5 1
8 rs1564125 2 1 DOM 2/5 4/4 0.608392 0.4 0
This file reports detailed information about each of the SNPs. Column P shows the P-value of the SNP. If the value in COMB1 is 1, the SNP is a member of COMB1 in example.lamp.
The output file formats are described in Output file format.
For the quantitive trait, Mann-Whitney U test can be used by the following command.
$ ./lamplink --file ./example/lamplink_sample --pheno ./example/phenotype_sample.txt --lamp --model-dom --sglev 0.05 --upper 0.5 --out example_utest --utest --alternative "greater"
Usage
LAMPLINK contains the following user-selectable functions.
- Main function: Detect significant epistatic interactions (
--lamp
option) - Post-processing: Eliminate SNP combinations that show linkage disequilibrium (LD) (
--lamp-ld-remove
option) - All functions implemented in PLINK v1.07 (list of PLINK options).
Detect significant epistatic interactions
The following options are available in LAMPLINK to enumerate significant epistatic interactions.
Option | Description |
---|---|
--lamp | Find combinatorial elements using LAMP |
--file (or --bfile) <filename> | Input filename without the extension |
--out <filename> | Output filename (default “lamplink.***”) |
--model-dom (or –model-rec) | dom: minor allele is dominant, rec: minor allele is recessive |
--fisher | Use Fisher’s exact test as the statistical significance test (default is Chi-square test) |
--utest | Use Mann-Whitney U test as the statistical significance test (default is Chi-square test) |
--alternative | Select which alternative hypothesis is used from "greater", "less" or "two.sided" (default is two.sided) |
--ci <value> | Output confidence interval for CMH odds ratios |
--sglev <value> | Set statistical significance level used in LAMP (default is 0.05) |
--upper <value> | Set maximum MAF value (default is 0.1) |
Input file format
Input files should be set with the --file
or --bfile
option.
For inputs using --file filename
, filename.ped and filename.map files should exist.
For inputs using --bfile
, filename.bed, filename.bim, and filename.fam files should exist.
For detailed descriptions of the required formats, please refer to Basic usage / data formats (PLINK web page).
Output file format
Results are output in two files.
- lamplink.lamp: result file of LAMP
- lamplink.lamplink: integrated file of the results of LAMP and PLINK
The filename can be changed using the --out
option.
lamplink.lamp
The lamplink.lamp file contains the following columns:
Column | Description |
---|---|
COMBID | Combination ID corresponding to COMBID in lamplink.lamplink |
Raw_P | P-value |
Adjusted_P | Adjusted P-value |
COMB | SNPs that are members of the combination |
lamplink.lamplink
The lamplink.lamplink file contains the following columns independent of the statistical significance test:
Column | Description |
---|---|
CHR | Chromosome number |
SNP | SNP name |
A1 | Minor allele name (calculated from the input ped file) |
A2 | Major allele name |
Test | DOM: minor allele is dominant, REC: minor allele is recessive (This output depends on the input option) |
AFF | Frequency of this allele in cases |
UNAFF | Frequency of this allele in controls |
P | P-value for this SNP |
OR | Estimated odds ratio (for A1, i.e., A2 is the reference) |
COMBx | 1: This SNP is a member of COMBx, 0: This SNP is absent from COMBx, x: Combination ID that corresponds to COMID in lamplink.lamp |
The lamplink.lamplink file contains the following additional columns when the Chi-square test is used:
Column | Description |
---|---|
CHSQ | Chi-square score |
DF | Degree of freedom |
The lamplink.lamplink file contains the following additional columns when the --ci
option is used:
Column | Description |
---|---|
Lx | Lower bound of x% confidence interval for odds ratio |
Ux | Upper bound of x% confidence interval for odds ratio |
Post-processing
The following options can be used to eliminate redundant SNP combinations.
Option | Description |
---|---|
--lamp-ld-remove | Eliminate SNP combinations in LD |
--file (or --bfile) <filename> | Input filename without the extension |
--comb <filename> | Result filenames set with the --lamp option (“***.lamplink” and “***.lamp”) without the extension |
--out <filename> | Output filename (default is “lamplink.***”) |
--lamp-r2 |
Set the threshold for the r-squared value (default is 0.8) |
The commands used to detect significant combinations and to eliminate redundant SNP combinations are shown using the example files example/lamplink_sample.map and example/lamplink_sample.ped.
$ ./lamplink --file ./example/lamplink_sample --lamp --model-dom --sglev 0.3 --upper 0.5 --out example2 --fisher
$ ./lamplink --file ./example/lamplink_sample --comb example2 --lamp-ld-remove --out example2_elimld --lamp-r2 0.5
When the results in example2.lamp and example2_elimld.lamp were compared, COMB1 was eliminated from example2_elimld.lamp.
example2.lamp:
COMBID Raw_P Adjusted_P COMB
COMB1 0.006993 0.034965 rs7817762,rs2631899,rs1841195
COMB2 0.040559 0.2028 rs7817762
COMB3 0.040559 0.2028 rs1841195
example2_elimld.lamp:
COMBID Raw_P Adjusted_P COMB
COMB2 0.040559 0.2028 rs7817762
COMB3 0.040559 0.2028 rs1841195
FAQ
- Q. LAMPLINK is still running after a few days.
- A. Please test --model-rec and/or set --upper to smaller value.
- Q. Can I run LAMPLINK on Mac or Windows?
- A. We have a plan to distribute Mac and Windows binary. Wait a sec!
Compile
LAMPLINK source code is available from Github. LAMPLINK can be compiled using the following command on Linux. C++ and Boostlibrary should be installed in advance.
$ git clone https://github.com/a-terada/lamplink.git
$ cd lamplink
$ ./install.sh
License
LAMPLINK is distributed under the GNU General Public License, Version 2. Please see GNU General Public License, version 2 for more details.
Contributors
LAMPLINK was developed through discussions with Aika Terada, Jun Sese, and Koji Tsuda.
Contact
Please contact us at the address below if you find a bug or need further information.
- lamp_staff(AT)googlegroups.com
Reference
- LAMPLINK
- Terada, A., Yamada, R. , Tsuda, K., & Sese, J. (2016). LAMPLINK: detection of statistically significant SNP combinations from GWAS data. Bioinformatics, 32(22): 3513-3515.
- LAMP
- Terada, A., Okada-Hatakeyama, M., Tsuda, K., & Sese, J. (2013). Statistical significance of combinatorial regulations. Proc. Natl. Acad. Sci. USA, 110(32), 12996–13001.
- software
- PLINK
- Purcell, S., Neale, B., Todd-Brown, K., et al. (2007). PLINK: a tool set for whole-genome association and population-based linkage analyses. American Journal of Human Genetics, 81(3), 559–575.
- software