Overview

DENTIST (Detecting Errors iN analyses of summary staTISTics) is a quality control (QC) tool for summary-level data from genome-wide association studies (GWASs). It leverages the difference between the observed GWAS test-statistic of a variant and its predicted value (using the neighbouring variants and linkage equilibrium (LD) data from a reference panel) to remove problematic variants. It can detect genotyping/imputation errors in either the original GWAS or the LD reference samples, allelic errors (i.e., the effect alleles of the variants are mislabelled) in the GWAS summary data, as well as heterogeneity between the GWAS and LD reference samples. As shown in our paper, DENTIST can significantly improve the performance of the summary-data based conditional and jointly association analysis (COJO; Yang et al. 2012 Nat Genet) especially for rare variants. It can also improve the performance of LD score regression (LDSC; Builk-Sulivant et al. 2015 Nat Genet) and SMR-HEIDI (Zhu et al. 2015 Nat Genet). We believe that DENTIST can in principle work for all analyses that use GWAS summary data, although this has not been tested extensively. Apart from its application to summary data based analyses, DENTIST can also be used in complementary to the conventional QCs in GWAS with individual-level data.

Software License: GPLv3

Credits

The method is developed by Wenhan Chen, Zhihong Zhu and Jian Yang in the Yang Lab at The University of Queensland. The software is programmed and maintained by Wenhan Chen.

Questions and Help Requests

If you meet any bugs or questions, please send an email to Wenhan Chen or Jian Yang.

Citations

Chen, W., Wu, Y., Zheng, Z. et al. Improved analyses of GWAS summary statistics by reducing data heterogeneity and errors. Nat Commun 12, 7117 (2021). https://doi.org/10.1038/s41467-021-27438-7

Install

For compiling yourself, I have not made a user-friendly MAKEFILE yet. For now, you can type

git clone https://github.com/Yves-CHEN/DENTIST
cd DENTIST
make

and address the dependencies yourself on MKL, BOOST and EIGEN Libraries.

Downloads

Pre-compiled Executable Files

The executable file below is compiled with "-static" and tested on 64-bit Linux distributions on the x86_64 CPU platform, DENTIST_1.3.0.0 DENTIST_1.2.0.0 DENTIST_1.1.0.0 DENTIST 0.9.2.1.

To download, you can run

wget -O DENTIST_1.3.0.0.gz https://www.dropbox.com/scl/fi/1yna9rizfy6s0ixek0cca/DENTIST.1.3.0.0.gz?rlkey=0ak0cdsev152jteibceit4mje&dl=0 wget -O DENTIST_1.2.0.0.gz https://www.dropbox.com/s/cq5mmfonocvdiwh/DENTIST.1.2.0.0.gz?dl=0

Basic usage

To run DENTIST with essential parameters,

DENTIST --gwas-summary summary_data --bfile ref --out prefix

To specify the number of CPUs,

DENTIST --gwas-summary summary_data --bfile ref --out prefix --thread-num 10

To run DENTIST at a targeted region specified by the rsID of variant (e.g. rs1234),

DENTIST --gwas-summary summary_data --bfile ref --out prefix --thread-num 10 --target rs1234

To run DENTIST for allelic frequencies

DENTIST --gwas-summary summary_data --bfile ref --out prefix --freq

Input and output

--bfile \<PLINK bed file prefix>

Reads individual-level genotype data in PLINK bed format, e.g. test.fam, test.bim and test.bed.

--gwas-summary \<summary statistics in GCTA-COJO format>

Reads GWAS summary data in in GCTA-COJO format (see below for an example).

SNP A1 A2 freq beta se p N
rs131538 A G 0.05 0.007 0.02 0.7 6000
rs140378 C G 0.05 0.007 0.02 0.7 6000
...

--out \<the output file prefix> Specifies the prefix of the output files including *.DENTIST.full.txt, *.DENTIST.short.txt, *.DENTIST.ignored.txt *.frq.

The .DENTIST.full.txt contains the statistics for all the tested variants in the following format, e.g.

head -3 \<prefix>.DENTIST.full.txt

rsID        chisq   -log10(P)  ifDup
rs131538    0.012   0.91       0
rs140378    14.4    2          0
...

As shown above, the first column contains the rsID for each variant, followed by the DENTIST test statistic (follows a distribution with 1 degree of freedom under the null), and the corresponding in the 3rd column. The 4th column is an indicator of whether the variant is in strong LD (|r| > 0.99) with any other variants, 0 for none and 1 for at least one.

The DENTIST.short.txt contains the rsIDs for variants that cannot pass DENTIST QC and are suggested for removal.

The DENTIST.ignored.txt contains variants that are ignored through QC for reasons including 1) inconsistently found in summmary-data and PLINK files; 2) MAF threshold; 3) MAF difference threshold between MAFs from summary and reference data; 4) alleles do not match between summary and reference data. e.g. head -3 \<prefix>.DENTIST.ignored.txt

rs141578542     notFoundInGWAS
chr22:16051722:D        notFoundInGWAS
rs139918843     notFoundInGWAS

The \<prefix>.frq shows the allelic frequencies of the A1 allele specific by --freq flag, e.g.

chr     RS_ID   BP      A1      A2      Freq_A1 n       nMissingGeno
1       1       1       2       1       0.04879 7703    693
1       2       2       2       1       0.05588 7703    742
1       3       3       2       1       0.04507 7703    758

In example above, the Freq_A1 is the allelic frequency, n is for the sample size of the cohort, nMissingGeno is for the number of individuals which has NA genotype and excluded for the calculation.

Cautions

• Do not perform cutoff on GWAS p-value.
• Make sure HWE QC is performed before running DENTIST.
• DENTIST will skip over low SNP density regions, any 2Mb-region with < 1000 SNPs.

Parameters