plink(1) whole genome SNP analysis


PLINK v1.90b3.36 64-bit (16 Apr 2016) (C) 2005-2016 Shaun Purcell, Christopher Chang GNU General Public License v3

In the command line flag definitions that follow,

* [square brackets] denote a required parameter, where the text between the
brackets describes its nature.
* <angle brackets> denote an optional modifier (or if '|' is present, a set
of mutually exclusive optional modifiers).
Use the EXACT text in the
definition, e.g. '--dummy acgt'.
* There's one exception to the angle brackets/exact text rule: when an angle
bracket term ends with '=[value]', '[value]' designates a variable parameter.
* {curly braces} denote an optional parameter, where the text between the
braces describes its nature.
* An ellipsis (...) indicates that you may enter multiple parameters of the
specified type.
plink [input flag(s)...] {command flag(s)...} {other flag(s)...} plink --help {flag name(s)...}

Most PLINK runs require exactly one main input fileset. The following flags are available for defining its form and location:

--bfile {prefix} : Specify .bed + .bim + .fam prefix (default 'plink').
--bed [filename] : Specify full name of .bed file.
--bim [filename] : Specify full name of .bim file.
--fam [filename] : Specify full name of .fam file.
: With --file/--tfile/--lfile/--vcf/--bcf/--data/--23file,
don't delete autogenerated binary fileset at end of run.
--file {prefix}
: Specify .ped + .map filename prefix (default 'plink').
--ped [filename] : Specify full name of .ped file.
--map [filename] : Specify full name of .map file.
: .fam/.ped file does not contain column 1 (family ID).
: .fam/.ped file does not contain columns 3-4 (parents).
: .fam/.ped file does not contain column 5 (sex).
: .fam/.ped file does not contain column 6 (phenotype).
--tfile {prefix} : Specify .tped + .tfam filename prefix (default 'plink').
--tped [fname]
: Specify full name of .tped file.
--tfam [fname]
: Specify full name of .tfam file.
--lfile {prefix} : Specify .lgen + .map + .fam (long-format fileset) prefix.
--lgen [fname]
: Specify full name of .lgen file.
--reference [fn] : Specify default allele file accompanying .lgen input.
: When used with --lfile/--lgen + --reference, specifies
that the .lgen file contains reference allele counts.
--vcf [filename] : Specify full name of .vcf or .vcf.gz file.
--bcf [filename] : Specify full name of BCF2 file.
--data {prefix}
: Specify Oxford .gen + .sample prefix (default 'plink').
--gen [filename] : Specify full name of .gen or .gen.gz file.
--bgen [f] <snpid-chr> : Specify full name of .bgen file.
--sample [fname] : Specify full name of .sample file.
--23file [fname] {FID} {IID} {sex} {pheno} {pat. ID} {mat. ID} :
Specify 23andMe input file.
--grm-gz {prfx}
: Specify .grm.gz + (GCTA rel. matrix) prefix.
--grm-bin {prfx} : Specify .grm.bin + .grm.N.bin + (GCTA triangular
binary relationship matrix) filename prefix.
--dummy [sample ct] [SNP ct] {missing geno freq} {missing pheno freq}
<acgt | 1234 | 12> <scalar-pheno>
This generates a fake input dataset with the specified number of samples and SNPs. By default, the missing genotype and phenotype frequencies are zero, and genotypes are As and Bs (change the latter with 'acgt'/'1234'/'12'). The 'scalar-pheno' modifier causes a normally distributed scalar phenotype to be generated instead of a binary one.
--simulate [simulation parameter file] <tags | haps> <acgt | 1234 | 12>
--simulate-qt [simulation parameter file] <tags | haps> <acgt | 1234 | 12>
--simulate generates a fake input dataset with disease-associated SNPs,
while --simulate-qt generates a dataset with quantitative trait loci.

Output files have names of the form 'plink.{extension}' by default. You can change the 'plink' prefix with

--out [prefix]
: Specify prefix for output files.

Most runs also require at least one of the following commands:

Create a new binary fileset.
Unlike the automatic text-to-binary
converters (which only heed chromosome filters), this supports all of PLINK's filtering flags.
Variants of --make-bed which only write a new .bim or .fam file.
Can be
used with only .bim/.fam input. USE THESE CAUTIOUSLY. It is very easy to desynchronize your binary genotype data and your .bim/.fam indexes if you use these commands improperly. If you have any doubt, stick with --make-bed.
--recode <01 | 12> <23 | A{-transpose} | AD | beagle{-nomap} | bimbam{-1chr}
| compound-genotypes | fastphase{-1chr} | HV{-1chr} | lgen{-ref} |
list | oxford | rlist | structure | transpose | vcf | vcf-fid | vcf-iid> <tab | tabx | spacex | bgz | gen-gz> <include-alt> <omit-nonmale-y>
Create a new text fileset with all filters applied.
By default, the
fileset consists of a .ped and a .map file, readable with --file. * The '12' modifier causes A1 (usually minor) alleles to be coded as '1'
and A2 alleles to be coded as '2', while '01' maps A1 -> 0 and A2 -> 1.
* The '23' modifier causes a 23andMe-formatted file to be generated.
can only be used on a single sample's data (--keep may be handy).
* The 'AD' modifier causes an sample-major additive (0/1/2) + dominant
(het = 1, otherwise 0) component file, suitable for loading from R, to be generated. If you don't want the dominant component, use 'A' instead. If you need uncounted alleles to be named in the header line, add the 'include-alt' modifier.
* The 'A-transpose' modifier causes a variant-major additive component file
to be generated.
* The 'beagle' modifier causes unphased per-autosome .dat and .map files,
readable by early BEAGLE versions, to be generated, while 'beagle-nomap' generates a single .beagle.dat file.
* The 'bimbam' modifier causes a BIMBAM-formatted fileset to be generated.
If your input data only contains one chromosome, you can use 'bimbam-1chr' instead to write a two-column .pos.txt file.
* The 'compound-genotypes' modifier removes the space between pairs of
allele codes for the same variant when generating a .ped + .map fileset.
* The 'fastphase' modifier causes per-chromosome fastPHASE files to be
If your input data only contains one chromosome, you can use
'fastphase-1chr' instead to exclude the chromosome number from the file extension.
* The 'HV' modifier causes a Haploview-format .ped + .info fileset to be
generated per chromosome.
'HV-1chr' is analogous to 'fastphase-1chr'.
* The 'lgen' modifier causes a long-format fileset (loadable with --lfile)
to be generated, while 'lgen-ref' generates a (usually) smaller long-format fileset loadable with --lfile + --reference.
* The 'list' modifier creates a genotype-based list, while 'rlist' creates
a rare-genotype fileset.
With these formats, the 'omit-nonmale-y'
modifier causes nonmale genotypes to be omitted on the Y chromosome.
* 'oxford' causes an Oxford-format .gen + .sample fileset to be generated.
If you also include the 'gen-gz' modifier, the .gen file is gzipped.
* The 'structure' modifier causes a Structure-format file to be generated. * 'transpose' creates a transposed text fileset (loadable with --tfile). * 'vcf', 'vcf-fid', and 'vcf-iid' result in production of a VCFv4.2 file.
'vcf-fid' and 'vcf-iid' cause family IDs or within-family IDs respectively to be used for the sample IDs in the last header row, while 'vcf' merges both IDs and puts an underscore between them. If the 'bgz' modifier is added, the VCF file is block-gzipped. The A2 allele is saved as the reference and normally flagged as not based on a real reference genome ('PR' INFO field value). When it is important for reference alleles to be correct, you'll also want to include --a2-allele and --real-ref-alleles in your command.
* The 'tab' modifier makes the output mostly tab-delimited instead of
mostly space-delimited.
'tabx' and 'spacex' force all tabs and all
spaces, respectively.
--flip-scan <verbose>
(alias: --flipscan)
LD-based scan for case/control strand inconsistency.
If a --covar file is loaded, --make-bed/--make-just-fam and --recode automatically generate an updated version (with all filters applied). However, if you do not wish to simultaneously generate a new genotype file, you can use --write-covar to just produce a pruned covariate file.
--write-cluster <omit-unassigned>
If clusters are specified with --within/--family, this generates a new cluster file (with all filters applied). The 'omit-unassigned' modifier causes unclustered samples to be omitted from the file; otherwise their cluster is 'NA'.
If sets have been defined, --write-set dumps 'END'-terminated set membership lists to {output prefix}.set, while --set-table writes a variant-by-set membership table to {output prefix}.set.table.
--merge [.ped filename] [.map filename]
--merge [text fileset prefix]
--bmerge [.bed filename] [.bim filename] [.fam filename]
--bmerge [binary fileset prefix]
Merge the given fileset with the initially loaded fileset, writing the result to {output prefix}.bed + .bim + .fam. (It is no longer necessary to simultaneously specify --make-bed.)
--merge-list [filename]
Merge all filesets named in the text file with the reference fileset, if one was specified. (However, this can also be used *without* a reference; in that case, the newly created fileset is then treated as the reference by most other PLINK operations.) The text file is interpreted as follows: * If a line contains only one name, it is assumed to be the prefix for a
binary fileset.
* If a line contains exactly two names, they are assumed to be the full
filenames for a text fileset (.ped first, then .map).
* If a line contains exactly three names, they are assumed to be the full
filenames for a binary fileset (.bed, then .bim, then .fam).
--write-snplist writes a .snplist file listing the names of all variants
which pass the filters and inclusion thresholds you've specified, while --list-23-indels writes the subset with 23andMe-style indel calls (D/I allele codes).
--list-duplicate-vars <require-same-ref> <ids-only> <suppress-first>
--list-duplicate-vars writes a .dupvar file describing all groups of
variants with matching positions and allele codes. * By default, A1/A2 allele assignments are ignored; use 'require-same-ref'
to override this.
* Normally, the report contains position and allele codes.
To remove them
(and produce a file directly usable with e.g. --extract/--exclude), use 'ids-only'. Note that this command will fail in 'ids-only' mode if any of the reported IDs are not unique.
* 'suppress-first' causes the first variant ID in each group to be omitted
from the report.
--freq <counts | case-control> <gz>
--freqx <gz>
--freq generates a basic allele frequency (or count, if the 'counts'
modifier is present) report.
This can be combined with --within/--family
to produce a cluster-stratified allele frequency/count report instead, or the 'case-control' modifier to report case and control allele frequencies separately. --freqx generates a more detailed genotype count report, designed for use with --read-freq.
--missing <gz>
Generate sample- and variant-based missing data reports.
If clusters are
defined, the variant-based report is cluster-stratified.
'gz' causes the
output files to be gzipped.
Check for association between missing calls and flanking haplotypes.
--hardy <midp> <gz>
Generate a Hardy-Weinberg exact test p-value report.
(This does NOT
simultaneously filter on the p-value any more; use --hwe for that.)
the 'midp' modifier, the test applies the mid-p adjustment described in Graffelman J, Moreno V (2013) The mid p-value in exact tests for Hardy-Weinberg Equilibrium.
--mendel <summaries-only>
Generate a Mendel error report.
The 'summaries-only' modifier causes the
.mendel file (listing every single error) to be skipped.
--het <small-sample> <gz>
Estimate inbreeding coefficients.
--het reports method-of-moments
estimates, while --ibc calculates all three values described in Yang J, Lee SH, Goddard ME and Visscher PM (2011) GCTA: A Tool for Genome-wide Complex Trait Analysis. (That paper also describes the relationship matrix computation we reimplement.) * These functions require decent MAF estimates. If there are very few
samples in your immediate fileset, --read-freq is practically mandatory since imputed MAFs are wildly inaccurate in that case.
* They also assume the marker set is in approximate linkage equilibrium. * By default, --het omits the n/(n-1) multiplier in Nei's expected
homozygosity formula.
The 'small-sample' modifier causes it to be
included, while forcing --het to use MAFs imputed from founders in the immediate dataset.
--check-sex {female max F} {male min F}
--check-sex ycount {female max F} {male min F} {female max Y obs}
{male min Y obs}
--check-sex y-only {female max Y obs} {male min Y obs}
--impute-sex {female max F} {male min F}
--impute-sex ycount {female max F} {male min F} {female max Y obs}
{male min Y obs}
--impute-sex y-only {female max Y obs} {male min Y obs}
--check-sex normally compares sex assignments in the input dataset with
those imputed from X chromosome inbreeding coefficients. * Make sure that the X chromosome pseudo-autosomal region has been split
off (with e.g. --split-x) before using this.
* You also need decent MAF estimates (so, with very few samples in your
immediate fileset, use --read-freq), and your marker set should be in approximate linkage equilibrium.
* By default, F estimates smaller than 0.2 yield female calls, and values
larger than 0.8 yield male calls.
If you pass numeric parameter(s) to
--check-sex, the first two control these thresholds.
There are now two modes which consider Y chromosome data. * In 'ycount' mode, gender is still imputed from the X chromosome, but
female calls are downgraded to ambiguous whenever more than 0 nonmissing Y genotypes are present, and male calls are downgraded when fewer than 0 are present. (Note that these are counts, not rates.) These thresholds are controllable with --check-sex ycount's optional 3rd and 4th numeric parameters.
* In 'y-only' mode, gender is imputed from nonmissing Y genotype counts.
The male minimum threshold defaults to 1 instead of zero in this case.
--impute-sex changes sex assignments to the imputed values, and is
otherwise identical to --check-sex.
It must be used with
--fst <case-control>
(alias: --Fst)
Estimate Wright's Fst for each autosomal diploid variant using the method introduced in Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure, given a set of subpopulations defined via --within. Raw and weighted global means are also reported. * If you're interested in the global means, it is usually best to perform
this calculation on a marker set in approximate linkage equilibrium.
* If you have only two subpopulations, you can represent them with
case/control status and use the 'case-control' modifier.
--indep [window size]<kb> [step size (variant ct)] [VIF threshold]
--indep-pairwise [window size]<kb> [step size (variant ct)] [r^2 threshold]
--indep-pairphase [window size]<kb> [step size (variant ct)] [r^2 threshold]
Generate a list of markers in approximate linkage equilibrium.
With the
'kb' modifier, the window size is in kilobase instead of variant count units. (Pre-'kb' space is optional, i.e. '--indep-pairwise 500 kb 5 0.5' and '--indep-pairwise 500kb 5 0.5' have the same effect.) Note that you need to rerun PLINK using --extract or --exclude on the file to apply the list to another computation.
--r <square | square0 | triangle | inter-chr> <gz | bin | bin4> <spaces>
<in-phase> <dprime> <with-freqs> <yes-really>
--r2 <square | square0 | triangle | inter-chr> <gz | bin | bin4> <spaces>
<in-phase> <dprime> <with-freqs> <yes-really>
LD statistic reports.
--r yields raw inter-variant correlations, while
--r2 reports their squares.
You can request results for all pairs in
matrix format (if you specify 'bin' or one of the shape modifiers), all pairs in table format ('inter-chr'), or a limited window in table format (default). * The 'gz' modifier causes the output text file to be gzipped. * 'bin' causes the output matrix to be written in double-precision binary
format, while 'bin4' specifics single-precision binary.
The matrix is
square if no shape is explicitly specified.
* By default, text matrices are tab-delimited; 'spaces' switches this. * 'in-phase' adds a column with in-phase allele pairs to table-formatted
(This cannot be used with very long allele codes.)
* 'dprime' adds Lewontin's D-prime statistic to table-formatted reports,
and forces both r/r^2 and D-prime to be based on the maximum likelihood solution to the cubic equation discussed in Gaunt T, Rodriguez S, Day I (2007) Cubic exact solutions for the estimation of pairwise haplotype frequencies.
* 'with-freqs' adds MAF columns to table-formatted reports. * Since the resulting file can easily be huge, you're required to add the
'yes-really' modifier when requesting an unfiltered, non-distributed all pairs computation on more than 400k variants.
* These computations can be subdivided with --parallel (even when the
'square' modifier is active).
--ld [variant ID] [variant ID] <hwe-midp>
This displays haplotype frequencies, r^2, and D' for a single pair of variants. When there are multiple biologically possible solutions to the haplotype frequency cubic equation, all are displayed (instead of just the maximum likelihood solution identified by --r/--r2), along with HWE exact test statistics.
--show-tags [filename]
--show-tags all
* If a file is specified, list all variants which tag at least one variant
named in the file.
(This will normally be a superset of the original
list, since a variant is considered to tag itself here.)
* If 'all' mode is specified, for each variant, each *other* variant which
tags it is reported.
--blocks <no-pheno-req> <no-small-max-span>
Estimate haplotype blocks, via Haploview's interpretation of the block definition suggested by Gabriel S et al. (2002) The Structure of Haplotype Blocks in the Human Genome. * Normally, samples with missing phenotypes are not considered by this
computation; the 'no-pheno-req' modifier lifts this restriction.
* Normally, size-2 blocks may not span more than 20kb, and size-3 blocks
are limited to 30kb.
The 'no-small-max-span' modifier removes these
The .blocks file is valid input for PLINK 1.07's --hap command.
the --hap... family of flags has not been reimplemented in PLINK 1.9 due to poor phasing accuracy relative to other software; for now, we recommend using BEAGLE instead of PLINK for case/control haplotype association analysis. (You can use '--recode beagle' to export data to BEAGLE 3.3.) We apologize for the inconvenience, and plan to develop variants of the --hap... flags which handle pre-phased data effectively.
--distance <square | square0 | triangle> <gz | bin | bin4> <ibs> <1-ibs>
<allele-ct> <flat-missing>
Write a lower-triangular tab-delimited table of (weighted) genomic distances in allele count units to {output prefix}.dist, and a list of the corresponding sample IDs to {output prefix} The first row of the .dist file contains a single {genome 1-genome 2} distance, the second row has the {genome 1-genome 3} and {genome 2-genome 3} distances in that order, etc. * It is usually best to perform this calculation on a marker set in
approximate linkage equilibrium.
* If the 'square' or 'square0' modifier is present, a square matrix is
written instead; 'square0' fills the upper right triangle with zeroes.
* If the 'gz' modifier is present, a compressed .dist.gz file is written
instead of a plain text file.
* If the 'bin' modifier is present, a binary (square) matrix of
double-precision floating point values, suitable for loading from R, is instead written to {output prefix}.dist.bin. ('bin4' specifies single-precision numbers instead.) This can be combined with 'square0' if you still want the upper right zeroed out, or 'triangle' if you don't want to pad the upper right at all.
* If the 'ibs' modifier is present, an identity-by-state matrix is written
to {output prefix}.mibs.
'1-ibs' causes distances expressed as genomic
proportions (i.e. 1 - IBS) to be written to {output prefix}.mdist. Combine with 'allele-ct' if you want to generate the usual .dist file as well.
* By default, distance rescaling in the presence of missing genotype calls
is sensitive to allele count distributions: if variant A contributes, on average, twice as much to other pairwise distances as variant B, a missing call at variant A will result in twice as large of a missingness correction. To turn this off (because e.g. your missing calls are highly nonrandom), use the 'flat-missing' modifier.
* The computation can be subdivided with --parallel.
These deprecated commands are equivalent to '--distance 1-ibs flat-missing square' and '--distance ibs flat-missing square', respectively, except that they generate space- instead of tab-delimited text matrices.
--make-rel <square | square0 | triangle> <gz | bin | bin4>
<cov | ibc2 | ibc3>
Write a lower-triangular variance-standardized realized relationship matrix to {output prefix}.rel, and corresponding IDs to {output prefix} * It is usually best to perform this calculation on a marker set in
approximate linkage equilibrium.
* 'square', 'square0', 'triangle', 'gz', 'bin', and 'bin4' act as they do
on --distance.
* The 'cov' modifier removes the variance standardization step, causing a
covariance matrix to be calculated instead.
* By default, the diagonal elements in the relationship matrix are based on
--ibc's Fhat1; use the 'ibc2' or 'ibc3' modifiers to base them on Fhat2
or Fhat3 instead.
* The computation can be subdivided with --parallel.
--make-grm-gz <no-gz> <cov | ibc2 | ibc3>
--make-grm-bin <cov | ibc2 | ibc3>
--make-grm-gz writes the relationships in GCTA's original gzipped list
format, which describes one pair per line, while --make-grm-bin writes them in GCTA 1.1+'s single-precision triangular binary format. Note that these formats explicitly report the number of valid observations (where neither sample has a missing call) for each pair, which is useful input for some scripts. These computations can be subdivided with --parallel.
--rel-cutoff {val}
(alias: --grm-cutoff)
Exclude one member of each pair of samples with relatedness greater than the given cutoff value (default 0.025). If no later operation will cause the list of remaining samples to be written to disk, this will save it to {output prefix} Note that maximizing the remaining sample size is equivalent to the NP-hard maximum independent set problem, so we use a greedy algorithm instead of guaranteeing optimality. (Use the --make-rel and --keep/--remove flags if you want to try to do better.)
--ibs-test {permutation count}
--groupdist {iters} {d}
Given case/control phenotype data, these commands consider three subsets of the distance matrix: pairs of affected samples, affected-unaffected pairs, and pairs of unaffected samples. Each of these subsets has a distribution of pairwise genomic distances; --ibs-test uses permutation to estimate p-values re: which types of pairs are most similar, while --groupdist focuses on the differences between the centers of these distributions and estimates standard errors via delete-d jackknife.
--regress-distance {iters} {d}
Linear regression of pairwise genomic distances on pairwise average phenotypes and vice versa, using delete-d jackknife for standard errors. A scalar phenotype is required. * With less than two parameters, d is set to {number of people}^0.6 rounded
With no parameters, 100k iterations are run.
--regress-rel {iters} {d}
Linear regression of pairwise genomic relationships on pairwise average phenotypes, and vice versa. Defaults for iters and d are the same as for --regress-distance.
--genome <gz> <rel-check> <full> <unbounded> <nudge>
Generate an identity-by-descent report. * It is usually best to perform this calculation on a marker set in
approximate linkage equilibrium.
* The 'rel-check' modifier excludes pairs of samples with different FIDs
from the final report.
* 'full' adds raw pairwise comparison data to the report. * The P(IBD=0/1/2) estimator employed by this command sometimes yields
numbers outside the range [0,1]; by default, these are clipped.
'unbounded' modifier turns off this clipping.
* Then, when PI_HAT^2 < P(IBD=2), 'nudge' adjusts the final P(IBD=0/1/2)
estimates to a theoretically possible configuration.
* The computation can be subdivided with --parallel.
--homozyg <group | group-verbose> <consensus-match> <extend>
--homozyg-snp [min var count]
--homozyg-kb [min length]
--homozyg-density [max inverse density (kb/var)]
--homozyg-gap [max internal gap kb length]
--homozyg-het [max hets]
--homozyg-window-snp [scanning window size]
--homozyg-window-het [max hets in scanning window hit]
--homozyg-window-missing [max missing calls in scanning window hit]
--homozyg-window-threshold [min scanning window hit rate]
These commands request a set of run-of-homozygosity reports, and allow you to customize how they are generated. * If you're satisfied with all the default settings described below, just
use --homozyg with no modifiers.
Otherwise, --homozyg lets you change a
few binary settings: * 'group{-verbose}' adds a report on pools of overlapping runs of
(Automatically set when --homozyg-match is present.)
* With 'group{-verbose}', 'consensus-match' causes pairwise segmental
matches to be called based on the variants in the pool's consensus segment, rather than the variants in the pairwise intersection.
* Due to how the scanning window algorithm works, it is possible for a
reported ROH to be adjacent to a few homozygous variants.
The 'extend'
modifier causes them to be included in the reported ROH if that wouldn't cause a violation of the --homozyg-density bound.
* By default, segment bp lengths are calculated as [end bp position] -
[start bp position] + 1.
Therefore, reports normally differ slightly
from PLINK 1.07, which does not add 1 at the end.
For testing
purposes, you can use the 'subtract-1-from-lengths' modifier to apply the old formula.
* By default, only runs of homozygosity containing at least 100 variants,
and of total length >= 1000 kilobases, are noted.
You can change these
minimums with --homozyg-snp and --homozyg-kb, respectively.
* By default, a ROH must have at least one variant per 50 kb on average;
change this bound with --homozyg-density.
* By default, if two consecutive variants are more than 1000 kb apart, they
cannot be in the same ROH; change this bound with --homozyg-gap.
* By default, a ROH can contain an unlimited number of heterozygous calls;
you can impose a limit with --homozyg-het.
* By default, the scanning window contains 50 variants; change this with
* By default, a scanning window hit can contain at most 1 heterozygous
call and 5 missing calls; change these limits with --homozyg-window-het and --homozyg-window-missing, respectively.
* By default, for a variant to be eligible for inclusion in a ROH, the hit
rate of all scanning windows containing the variant must be at least 0.05; change this threshold with --homozyg-window-threshold.
--cluster <cc> <group-avg | old-tiebreaks> <missing> <only2>
Cluster samples using a pairwise similarity statistic (normally IBS). * The 'cc' modifier forces every cluster to have at least one case and one
* The 'group-avg' modifier causes clusters to be joined based on average
instead of minimum pairwise similarity.
* The 'missing' modifier causes clustering to be based on
identity-by-missingness instead of identity-by-state, and writes a space-delimited identity-by-missingness matrix to disk.
* The 'only2' modifier causes only a .cluster2 file (which is valid input
for --within) to be written; otherwise 2 other files will be produced.
* By default, IBS ties are not broken in the same manner as PLINK 1.07, so
final cluster solutions tend to differ.
This is generally harmless.
However, to simplify testing, you can use the 'old-tiebreaks' modifier to force emulation of the old algorithm.
--pca {count} <header> <tabs> <var-wts>
Calculates a variance-standardized relationship matrix (use --make-rel/--make-grm-gz/--make-grm-bin to dump it), and extracts the top 20 principal components. * It is usually best to perform this calculation on a marker set in
approximate linkage equilibrium.
* You can change the number of PCs by passing a numeric parameter. * The 'header' modifier adds a header line to the .eigenvec output file.
(For compatibility with the GCTA flag of the same name, the default is no header line.)
* The 'tabs' modifier causes the .eigenvec file(s) to be tab-delimited. * The 'var-wts' modifier requests an additional .eigenvec.var file with PCs
expressed as variant weights instead of sample weights.
--neighbour [n1] [n2]
(alias: --neighbor)
Report IBS distances from each sample to their n1th- to n2th-nearest neighbors, associated Z-scores, and the identities of those neighbors. Useful for outlier detection.
--assoc <perm | mperm=[value]> <perm-count> <fisher | fisher-midp> <counts>
--assoc <perm | mperm=[value]> <perm-count> <qt-means> <lin> <set-test>
--model <perm | mperm=[value]> <perm-count>
<fisher | fisher-midp | trend-only> <set-test> <dom | rec | gen | trend>
Basic association analysis report. Given a case/control phenotype, --assoc performs a 1df chi-square allelic test, while --model performs 4 other tests as well (1df dominant gene action, 1df recessive gene action, 2df genotypic, Cochran-Armitage trend). * With 'fisher'/'fisher-midp', Fisher's exact test is used to generate
'fisher-midp' also applies Lancaster's mid-p adjustment.
* 'perm' causes an adaptive permutation test to be performed. * 'mperm=[value]' causes a max(T) permutation test with the specified
number of replications to be performed.
* 'perm-count' causes the permutation test report to include counts instead
of frequencies.
* 'counts' causes --assoc to report allele counts instead of frequencies. * 'set-test' tests the significance of variant sets. Requires permutation;
can be customized with --set-p/--set-r2/--set-max.
* 'dom', 'rec', 'gen', and 'trend' force the corresponding test to be used
as the basis for --model permutation.
(By default, the most significant
result among the allelic, dominant, and recessive tests is used.)
* 'trend-only' causes only the trend test to be performed. Given a quantitative phenotype, --assoc normally performs a Wald test. * In this case, the 'qt-means' modifier causes trait means and standard
deviations stratified by genotype to be reported as well.
* 'lin' causes the Lin statistic to be computed, and makes it the basis for
multiple-testing corrections and permutation tests.
Several other flags (most notably, --aperm) can be used to customize the permutation test.
--mh <perm | mperm=[value]> <perm-count> <set-test>
(alias: --cmh)
--bd <perm | perm-bd | mperm=[value]> <perm-count> <set-test>
Given a case/control phenotype and a set of clusters, --mh computes 2x2xK Cochran-Mantel-Haenszel statistics for each variant, while --bd also performs the Breslow-Day test for odds ratio homogeneity. Permutation and variant set testing based on the CMH (default) or Breslow-Day (when 'perm-bd' is present) statistic are supported. The following similar analyses are also available: * --mh2 swaps the roles of case/control status and cluster membership,
performing a phenotype-stratified IxJxK Cochran-Mantel-Haenszel test on association between cluster assignments and genotypes.
* --homog executes an alternative to the Breslow-Day test, based on
partitioning of the chi-square statistic.
--gxe {covariate index}
Given both a quantitative phenotype and a case/control covariate loaded with --covar defining two groups, --gxe compares the regression coefficient derived from considering only members of one group to the regression coefficient derived from considering only members of the other. By default, the first covariate in the --covar file defines the groups; use e.g. '--gxe 3' to base them on the third covariate instead.
--linear <perm | mperm=[value]> <perm-count> <set-test>
<genotypic | hethom | dominant | recessive | no-snp> <hide-covar> <sex | no-x-sex> <interaction> <beta> <standard-beta> <intercept>
--logistic <perm | mperm=[value]> <perm-count> <set-test>
<genotypic | hethom | dominant | recessive | no-snp> <hide-covar> <sex | no-x-sex> <interaction> <beta> <intercept>
Multi-covariate association analysis on a quantitative (--linear) or case/control (--logistic) phenotype. Normally used with --covar. * 'perm' normally causes an adaptive permutation test to be performed on
the main effect, while 'mperm=[value]' starts a max(T) permutation test.
* 'perm-count' causes the permutation test report to include counts instead
of frequencies.
* 'set-test' tests the significance of variant sets.
Requires permutation;
can be customized with --set-p/--set-r2/--set-max.
* The 'genotypic' modifier adds an additive effect/dominance deviation 2df
joint test (0/1/2 and 0/1/0 coding), while 'hethom' uses 0/0/1 and 0/1/0 coding instead. If permutation is also requested, these modifiers cause permutation to be based on the joint test.
* 'dominant' and 'recessive' specify a model assuming full dominance or
recessiveness, respectively, for the A1 allele.
* 'no-snp' causes regression to be performed only on the phenotype and the
covariates, without reference to genomic data.
If permutation is also
requested, results are reported for all covariates.
* 'hide-covar' removes covariate-specific lines from the report. * By default, sex (male = 1, female = 0) is automatically added as a
covariate on X chromosome variants, and nowhere else.
The 'sex' modifier
causes it to be added everywhere, while 'no-x-sex' excludes it.
* 'interaction' adds genotype x covariate interactions to the model.
cannot be used with the usual permutation tests; use --tests to define the permutation test statistic instead.
* 'intercept' causes intercepts to be included in the main report. * For logistic regressions, the 'beta' modifier causes regression
coefficients instead of odds ratios to be reported.
* With --linear, the 'standard-beta' modifier standardizes the phenotype
and all predictors to zero mean and unit variance before regression.
--dosage [allele dosage file] <noheader> <skip0=[i]> <skip1=[j]> <skip2=[k]>
<dose1> <format=[m]> <Zout> <occur | standard-beta> <sex> <case-control-freqs>
--dosage [list file] list <sepheader | noheader> <skip0=[i]> <skip1=[j]>
<skip2=[k]> <dose1> <format=[m]> <Zout> <occur | standard-beta> <sex> <case-control-freqs>
Process (possibly gzipped) text files with variant-major allelic dosage data. This cannot be used with a regular input fileset; instead, you must *only* specify a .fam and possibly a .map file, and you can't specify any other commands. * PLINK 2.0 will have first-class support for genotype probabilities. An
equivalent data import flag will be provided then, and --dosage will be retired.
* By default, --dosage assumes that only one allelic dosage file should be
To specify multiple files,
1. create a master list with one entry per line.
There are normally two
supported formats for this list: just a filename per line, or variant batch numbers in the first column and filenames in the second.
2. Provide the name of that list as the first --dosage parameter. 3. Add the 'list' modifier.
* By default, --dosage assumes the allelic dosage file(s) contain a header
line, which has 'SNP' in column i+1, 'A1' in column i+j+2, 'A2' in column i+j+3, and sample FID/IIDs starting from column i+j+k+4. (i/j/k are normally zero, but can be changed with 'skip0', 'skip1', and 'skip2' respectively.) If such a header line is not present, * when all samples appear in the same order as they do in the .fam file,
you can use the 'noheader' modiifer.
* Otherwise, use the 'sepheader' modifier, and append sample ID filenames
to your 'list' file entries.
* The 'format' modifier lets you specify the number of values used to
represent each dosage.
'format=1' normally indicates a single 0..2 A1
expected count; 'dose1' modifies this to a 0..1 frequency.
(the default) indicates a 0..1 homozygous A1 likelihood followed by a 0..1 het likelihood, while 'format=3' indicates 0..1 hom A1, 0..1 het, 0..1 hom A2.
* 'Zout' causes the output file to be gzipped. * Normally, an association analysis is performed. 'standard-beta' and
'sex' behave as they are supposed to with --linear/--logistic. 'case-control-freqs' causes case and control allele frequencies to be reported separately.
* There are three alternate modes which cause the association analysis to
be skipped. * 'occur' requests a simple variant occurrence report. * --write-dosage causes a simple merged file matching the 'format'
specification (not including 'dose1') to be generated.
* --score applies a linear scoring system to the dosages.
--lasso [h2 estimate] {min lambda} <report-zeroes>
Estimate variant effect sizes via LASSO regression.
You must provide an
additive heritability estimate to calibrate the regression. Note that this method may require a very large sample size (e.g. hundreds of thousands) to be effective on complex polygenic traits.
--test-missing <perm | mperm=[value]> <perm-count> <midp>
Check for association between missingness and case/control status, using Fisher's exact test. The 'midp' modifier causes Lancaster's mid-p adjustment to be applied.
--make-perm-pheno [ct]
Generate phenotype permutations and write them to disk, without invoking an association test.
--tdt <exact | exact-midp | poo> <perm | mperm=[value]> <perm-count>
<parentdt1 | parentdt2 | pat | mat> <set-test>
Report transmission disequilibrium test statistics, given case/control phenotypes and pedigree information. * A Mendel error check is performed before the main tests; offending
genotypes are treated as missing by this analysis.
* By default, the basic TDT p-value is based on a chi-square test unless
you request the exact binomial test with 'exact' or 'exact-midp'.
* 'perm'/'mperm=[value]' requests a family-based adaptive or max(T)
permutation test.
By default, the permutation test statistic is the
basic TDT p-value; 'parentdt1'/'parentdt2' cause parenTDT or combined test p-values, respectively, to be considered instead.
* 'set-test' tests the significance of variant sets.
This cannot be used
with exact tests for now.
The 'poo' modifier causes a parent-of-origin analysis to be performed instead, with transmissions from heterozygous fathers and heterozygous mothers considered separately. * The parent-of-origin analysis does not currently support exact tests. * By default, the permutation test statistic is the absolute
parent-of-origin test Z score; 'pat'/'mat' cause paternal or maternal TDT chi-square statistics, respectively, to be considered instead.
--qfam <perm | mperm=[value]> <perm-count> <emp-se>
--qfam-parents <perm | mperm=[value]> <perm-count> <emp-se>
--qfam-between <perm | mperm=[value]> <perm-count> <emp-se>
--qfam-total <perm | mperm=[value]> <perm-count> <emp-se>
QFAM family-based association test for quantitative traits. * A Mendel error check is performed before the main tests; offending
genotypes are treated as missing by this analysis.
* This procedure requires permutation.
'perm' and 'perm-count' have the
usual meanings.
However, 'mperm=[value]' just specifies a fixed number
of permutations; the method does not support a proper max(T) test.
* The 'emp-se' modifier adds BETA and EMP_SE (empirical standard error for
beta) fields to the .perm output file.
--annotate [PLINK report] <attrib=[file]> <ranges=[file]> <filter=[file]>
<snps=[file]> <NA | prune> <block> <subset=[file]> <minimal> <distance>
Add annotations to a variant-based PLINK report.
This requires an
annotation source: * 'attrib=[file]' specifies a (possibly gzipped) attribute file. * 'ranges=[file]' specifies a gene/range list file. (Both source types can be specified simultaneously.) The following options are also supported: * 'filter=[file]' causes only variants within one of the ranges in the file
to be included in the new report.
* 'snps=[file]' causes only variants named in the file to be included in
the new report.
* The 'NA' modifier causes unannotated variants to have 'NA' instead of '.'
in the new report's ANNOT column, while the 'prune' modifier excludes them entirely.
* The 'block' modifier replaces the single ANNOT column with a 0/1-coded
column for each possible annotation.
* With 'ranges',
* 'subset=[file]' causes only intervals named in the subset file to be
loaded from the ranges file.
* interval annotations normally come with a parenthesized signed distance
to the interval boundary (0 if the variant is located inside the interval; this is always true without --border). They can be excluded with the 'minimal' modifier.
* the 'distance' modifier adds 'DIST' and 'SGN' columns describing signed
distance to the nearest interval.
* When --pfilter is present, high p-values are filtered out.
--clump [PLINK report filename(s)...]
Process association analysis report(s) with 'SNP' and p-value columns, organizing results by LD-based clumps. Multiple filenames can be separated by spaces or commas.
--gene-report [PLINK report] [gene range file]
Generate a gene-based report from a variant-based report. * When --pfilter is present, high p-values are filtered out. * When --extract (without 'range') is present, only variants named in the
--extract file are considered.
--meta-analysis [PLINK report filenames...]
--meta-analysis [PLINK report filenames...] + <logscale | qt>
<no-map | no-allele> <study> <report-all> <weighted-z>
Perform a meta-analysis on several variant-based reports with 'SNP' and 'SE' fields. * Normally, an 'OR' odds ratio field must also be present in each input
With 'logscale', 'BETA' log-odds values/regression coefficients
are expected instead, but the generated report will still contain odds ratio estimates. With 'qt', both input and output values are regression betas.
* 'CHR', 'BP', and 'A1' fields are also normally required.
'no-map' causes
them to all be ignored, while 'no-allele' causes just 'A1' to be ignored.
* If 'A2' fields are present, and neither 'no-map' nor 'no-allele' was
specified, A1/A2 allele flips are handled properly.
Otherwise, A1
mismatches are thrown out.
* 'study' causes study-specific effect estimates to be collated in the
meta-analysis report.
* 'report-all' causes variants present in only a single input file to be
included in the meta-analysis report.
* 'weighted-z' requests weighted Z-score-based p-values (as computed by the
Abecasis Lab's METAL software) in addition to the usual inverse variance-based analysis. This requires P and effective sample size fields.
* When --extract (without 'range') is present, only variants named in the
--extract file are considered.
* Unless 'no-map' is specified, chromosome filters are also respected.
--fast-epistasis <boost | joint-effects | no-ueki> <case-only>
<set-by-set | set-by-all> <nop>
--epistasis <set-by-set | set-by-all>
Scan for epistatic interactions.
--fast-epistasis inspects 3x3 joint
genotype count tables and only applies to case/control phenotypes, while --epistasis performs linear or logistic regression. * By default, --fast-epistasis uses the PLINK 1.07 allele-based test. Two
newer tests are now supported: 'boost' invokes the likelihood ratio test introduced by Wan X et al. (2010) BOOST: A Fast Approach to Detecting Gene-Gene Interactions in Genome-wide Case-Control Studies, while 'joint-effects' applies the joint effects test introduced in Ueki M, Cordell HJ (2012) Improved statistics for genome-wide interaction analysis.
* The original --fast-epistasis test normally applies the variance and
empty cell corrections suggested by Ueki and Cordell's paper.
To disable
them, use the 'no-ueki' modifier.
* 'case-only' requests a case-only instead of a case/control test. * By default, all pairs of variants across the entire genome are tested.
To just test pairs of variants within a single set, add the 'set-by-set' modifier and load exactly one set with --set/--make-set; with exactly two sets loaded, all variants in one set are tested against all variants in the other. 'set-by-all' tests all variants in one set against the entire genome instead.
* 'nop' strips p-values from the main report. * These computations can be subdivided with --parallel; however...
--epistasis-summary-merge [common file prefix] [ct]
When a --{fast-}epistasis job is subdivided with --parallel, the main report can be assembled at the end by applying Unix 'cat' in the usual manner, but the .summary.1, .summary.2, ... files may require a specialized merge. --epistasis-summary-merge takes care of the latter.
--twolocus [variant ID] [variant ID]
Two-locus joint genotype count report.
--score [filename] {i} {j} {k} <header> <sum | no-sum>
<no-mean-imputation | center> <include-cnt> <double-dosage>
Apply a linear scoring system to each sample. The input file should have one line per scored variant. Variant IDs are read from column #i, allele codes are read from column #j, and scores are read from column #k, where i defaults to 1, j defaults to i+1, and k defaults to j+1. * The 'header' modifier causes the first nonempty line of the input file to
be ignored; otherwise, --score assumes there is no header line.
* By default, final scores are averages of the valid per-variant scores.
The 'sum' modifier causes sums to be reported instead.
(This cannot be
used with 'no-mean-imputation'.
And for backward compatibility, 'sum' is
automatically on with dosage data unless 'no-sum' is specified.)
* By default, copies of the unnamed allele contribute zero to score, while
missing genotypes contribute an amount proportional to the loaded (via --read-freq) or imputed allele frequency. To throw out missing observations instead (decreasing the denominator in the final average when this happens), use the 'no-mean-imputation' modifier.
* Alternatively, you can use the 'center' modifier to shift all scores to
mean zero.
* This command can be used with dosage data.
By default, the 'CNT' column
is omitted from the output file in this case; use 'include-cnt' to keep it. Also, note that scores are multiplied by 0..1 dosages, not 0..2 diploid allele counts, unless the 'double-dosage' modifier is present.
--write-var-ranges [block ct]
Divide the set of variants into equal-size blocks.
(Can be used with
--snps to split a job across multiple machines.)

The following other flags are supported. (Order of operations is described at .)
--script [fname] : Include command-line options from file.
--rerun {log}
: Rerun commands in log (default 'plink.log').
: Display only version number before exiting.
: Suppress output to console.
: Reserved for interoperation with gPLINK.
--missing-genotype [char] : Set missing genotype code (normally '0').
: Set both FIDs and IIDs to the VCF/BCF sample ID.
--const-fid {ID}
: Set all FIDs to the given constant (default '0').
--id-delim {d}
: Parse sample IDs as [FID][d][IID] (default delim '_').
--vcf-idspace-to [c] : Convert spaces in sample IDs to the given character.
--biallelic-only <strict> <list> : Skip VCF variants with 2+ alt. alleles.
--vcf-min-qual [val]
: Skip VCF variants with low/missing QUAL.
--vcf-filter {exception(s)...}
: Skip variants which have FILTER failures.
: Skip variants with no GT field.
--vcf-min-gq [val]
: No-call a genotype when GQ is below the
given threshold.
--vcf-min-gp [val]
: No-call a genotype when 0-1 scaled GP is below the given threshold.
--vcf-half-call [m]
: Specify how '0/.' and similar VCF GT values should be handled. The following four modes are supported: * 'error'/'e' (default) errors out and reports line #. * 'haploid'/'h' treats them as haploid calls. * 'missing'/'m' treats them as missing. * 'reference'/'r' treats the missing value as 0.
--oxford-single-chr [chr nm] : Specify single-chromosome .gen file with
ignorable first column.
--oxford-pheno-name [col nm] : Import named phenotype from the .sample file.
--hard-call-threshold [val]
: When an Oxford-format fileset is loaded, calls
--hard-call-threshold random
with uncertainty level greater than 0.1 are normally treated as missing. You can adjust this threshold by providing a numeric parameter, or randomize all calls with 'random'.
--missing-code {string list} : Comma-delimited list of missing phenotype
(alias: --missing_code)
values for Oxford-format filesets (def. 'NA').
--simulate-ncases [num]
: Set --simulate case count (default 1000).
--simulate-ncontrols [n]
: Set --simulate control count (default 1000).
--simulate-prevalence [p] : Set --simulate disease prevalence (default 0.01).
--simulate-n [num]
: Set --simulate-qt sample count (default 1000).
--simulate-label [prefix] : Set --simulate{-qt} FID/IID name prefix.
--simulate-missing [freq] : Set --simulate{-qt} missing genotype frequency.
--allow-extra-chr <0>
: Permit unrecognized chromosome codes. The '0'
(alias: --aec)
modifier causes them to be treated as if they had been set to zero.
--chr-set [autosome ct] <no-x> <no-y> <no-xy> <no-mt> :
Specify a nonhuman chromosome set.
The first parameter sets the number of
diploid autosome pairs if positive, or haploid chromosomes if negative. Given diploid autosomes, the remaining modifiers indicate the absence of the named non-autosomal chromosomes.
--cow/--dog/--horse/--mouse/--rice/--sheep : Shortcuts for those species.
--autosome-num [value]
: Alias for '--chr-set [value] no-y no-xy no-mt'.
--cm-map [fname pattern] {chr} : Use SHAPEIT-format recombination maps to set
centimorgan positions. To process more than one chromosome, include a '@' in the first parameter where the chrom. number belongs, e.g. 'genetic_map_chr@_combined_b37.txt'.
: Zero out centimorgan positions.
--pheno [fname]
: Load phenotype data from the specified file, instead of using the values in the main input fileset.
: For basic association tests, loop through all phenotypes in --pheno file.
--mpheno [n]
: Load phenotype from column (n+2) in --pheno file.
--pheno-name [c] : If --pheno file has a header row, use column with the
given name.
: When the main input fileset contains an phenotype value for a sample, but the --pheno file does not, use the original value instead of treating the phenotype as missing.
--missing-phenotype [v] : Set missing phenotype value (normally -9).
: Expect case/control phenotypes to be coded as
0 = control, 1 = case, instead of the usual 0 = missing, 1 = control, 2 = case.
--make-pheno [fn] [val] : Define a new case/control phenotype.
If the val parameter is '*', all samples listed in the given file are cases, and everyone else is a control. (Note that, in some shells, it is necessary to surround the * with quotes.) Otherwise, all samples with third column entry equal to the val parameter are cases, and all other samples mentioned in the file are controls.
--tail-pheno [Lt] {Hbt} : Downcode a scalar phenotype to a case/control
phenotype. All samples with phenotype values greater than Hbt are cases, and all with values less than or equal to Lt are controls. If Hbt is unspecified, it is equal to Lt; otherwise, in-between phenotype values are set to missing.
--covar [filename] <keep-pheno-on-missing-cov> : Specify covariate file.
--covar-name [...]
: Specify covariate(s) in --covar file by name.
Separate multiple names with spaces or commas, and use dashes to designate ranges.
--covar-number [...]
: Specify covariate(s) in --covar file by index.
: Exclude constant covariates.
--within [f] <keep-NA>
: Specify initial cluster assignments.
--mwithin [n]
: Load cluster assignments from column n+2.
: Create a cluster for each family ID.
--loop-assoc [f] <keep-NA>
: Run specified case/control association commands once for each cluster in the file, using cluster membership as the phenotype.
--set [filename]
: Load sets from a .set file.
--set-names [name(s)...]
: Load only sets named on the command line. Use spaces to separate multiple names.
--subset [filename]
: Load only sets named in the given text file.
--set-collapse-all [set name] : Merge all sets.
: Invert all sets. (Names gain 'C_' prefixes.)
--make-set-complement-all [s] : --set-collapse-all + inversion.
--make-set [filename]
: Define sets from a list of named bp ranges.
--make-set-border [kbs]
: Stretch regions in --make-set file.
: Define sets from groups instead of sets in --make-set file.
--keep [filename]
: Exclude all samples not named in the file.
--remove [filename]
: Exclude all samples named in the file.
--keep-fam [filename] : Exclude all families not named in the file.
--remove-fam [fname]
: Exclude all families named in the file.
--extract <range> [f] : Exclude all variants not named in the file.
--exclude <range> [f] : Exclude all variants named in the file.
--keep-clusters [filename]
: These can be used individually or in
--keep-cluster-names [name(s)...]
combination to define a list of clusters to keep; all samples not in a cluster in that list are then excluded. Use spaces to separate cluster names for --keep-cluster-names.
--remove-clusters [filename]
: Exclude all clusters named in the file.
--remove-cluster-names [name(s)...] : Exclude the named clusters.
--gene [sets...] : Exclude variants not in a set named on the command line.
(Separate multiple set names with spaces.)
: Exclude variants which aren't a member of any set. (PLINK
1.07 automatically did this under some circumstances.)
--attrib [f] {att lst} : Given a file assigning attributes to variants, and a
--attrib-indiv [f] {a}
comma-delimited list (with no whitespace) of attribute names, remove variants/samples which are either missing from the file or don't have any of the listed attributes. If some attribute names in the list are preceded by '-', they are treated as 'negative match conditions' instead: variants with at least one negative match attribute are removed. The first character in the list cannot be a '-', due to how command-line parsing works; add a comma in front to get around this.
--chr [chrs...]
: Exclude all variants not on the given chromosome(s). Valid choices for humans are 0 (unplaced), 1-22, X, Y, XY, and MT. Separate multiple chromosomes with spaces and/or commas, and use a dash (no adjacent spaces permitted) to denote a range, e.g. '--chr 1-4, 22, xy'.
--not-chr [...]
: Reverse of --chr (exclude variants on listed chromosomes).
: Exclude all non-autosomal variants.
: Exclude all non-autosomal variants, except those with chromosome code XY (pseudo-autosomal region of X).
--snps-only <no-DI> : Exclude variants with multi-character allele codes.
--from [var ID]
: Use ID(s) to specify a variant range to load. When used
[var ID] together, both variants must be on the same chromosome.
[var ID] : Specify a single variant to load.
--exclude-snp [] : Specify a single variant to exclude.
[kbs] : With --snp or --exclude-snp, loads/excludes all variants
within half the specified kb distance of the named one.
--from-bp [pos]
: Use physical position(s) to define a variant range to
[pos] load. --from-kb/--to-kb/--from-mb/--to-mb allow decimal
--from-kb [pos]
values. You must also specify a single chromosome (using
[pos] e.g. --chr) when using these flags.
--from-mb [pos]
--snps [var IDs...]
: Use IDs to specify variant range(s) to load or
--exclude-snps [...]
exclude. E.g. '--snps rs1111-rs2222, rs3333, rs4444'.
--thin [p]
: Randomly remove variants, retaining each with prob. p.
--thin-count [n] : Randomly remove variants until n of them remain.
--bp-space [bps] : Remove variants so that each pair is no closer than the
given bp distance. (Equivalent to VCFtools --thin.)
--thin-indiv [p]
: Randomly remove samples, retaining with prob. p.
--thin-indiv-count [n]
: Randomly remove samples until n of them remain.
--filter [f] [val(s)...] : Exclude all samples without a 3rd column entry in
the given file matching one of the given space-separated value(s).
--mfilter [n]
: Match against (n+2)th column instead.
--geno {val}
: Exclude variants with missing call frequencies greater than a threshold (default 0.1). (Note that the default threshold is only applied if --geno is invoked without a parameter; when --geno is not invoked, no per-variant missing call frequency ceiling is enforced at all. Other inclusion/exclusion default thresholds work the same way.)
--mind {val}
: Exclude samples with missing call frequencies greater than a threshold (default 0.1).
--oblig-missing [f1] [f2] : Specify blocks of missing genotype calls for
--geno/--mind to ignore. The first file should have variant IDs in the first column and block IDs in the second, while the second file should have FIDs in the first column, IIDs in the second, and block IDs in the third.
: Remove samples with missing phenotypes.
--maf {freq}
: Exclude variants with minor allele frequency lower than a threshold (default 0.01).
--max-maf [freq]
: Exclude variants with MAF greater than the threshold.
--mac [ct]
: Exclude variants with minor allele count lower than the
(alias: --min-ac)
given threshold.
--max-mac [ct]
: Exclude variants with minor allele count greater than
(alias: --max-ac)
the given threshold.
: Rule of succession MAF estimation (used in EIGENSOFT). Given j observations of one allele and k >= j observations of the other, infer a MAF of (j+1) / (j+k+2), rather than the default j / (j+k).
--read-freq [fn] : Estimate MAFs and heterozygote frequencies from the given
--freq{x} report, instead of the input fileset.
--hwe [p] <midp> <include-nonctrl> : Exclude variants with Hardy-Weinberg
equilibrium exact test p-values below a threshold.
--me [t] [v] <var-first> : Filter out trios and variants with Mendel error
rates exceeding the given thresholds.
--me-exclude-one {ratio} : Make --me exclude only one sample per trio.
--qual-scores [f] {qcol} {IDcol} {skip} : Filter out variants with
out-of-range quality scores. Default range is now [0, \infty ).
--qual-threshold [min qual score]
: Set --qual-scores range floor.
--qual-max-threshold [max qual score]
: Set --qual-scores range ceiling.
: Do not treat ambiguous-sex samples as having missing phenotypes in analysis commands. (Automatic ,/w/ --no-sex.)
: Force ambiguous-sex phenotypes to missing on --make-bed/--make-just-fam/--recode/--write-covar.
: Include only cases in the current analysis.
: Include only controls.
: Include only males.
: Include only females.
: Include only founders.
--filter-nonfounders : Include only nonfounders.
: Include nonfounders in allele freq/HWE calculations.
--make-founders <require-2-missing> <first> : Clear parental IDs for those
with 1+ missing parent(s).
--recode-allele [fn] : With --recode A/A-transpose/AD, count alleles named in
the file (otherwise A1 alleles are always counted).
--output-chr [MT code] : Set chromosome coding scheme in output files by
providing the desired human mitochondrial code. (Options are '26', 'M', 'MT', '0M', 'chr26', 'chrM', and 'chrMT'.)
--output-missing-genotype [ch] : Set the code used to represent missing
genotypes in output files (normally the --missing-genotype value).
--output-missing-phenotype [s] : Set the string used to represent missing
phenotypes in output files (normally the --missing-phenotype value).
--zero-cluster [f] : In combination with --within/--family, set blocks of
genotype calls to missing. The input file should have variant IDs in the first column and cluster IDs in the second. This must now be used with --make-bed and no other output commands.
--set-hh-missing : Cause --make-bed and --recode to set heterozygous haploid
genotypes to missing.
--split-x [bp1] [bp2] <no-fail> : Changes chromosome code of all X chromosome
--split-x [build] <no-fail>
variants with bp position <= bp1 or >= bp2 to XY. The following build codes are supported as shorthand: * 'b36'/'hg18' = NCBI 36, 2709521/154584237 * 'b37'/'hg19' = GRCh37, 2699520/154931044 * 'b38'/'hg38' = GRCh38, 2781479/155701383 By default, PLINK errors out when no variants would be affected by --split-x; the 'no-fail' modifier (useful in scripts) overrides this.
--merge-x <no-fail>
: Merge XY chromosome back with X.
: Cause --make-bed to set Mendel errors to missing.
--fill-missing-a2 : Cause --make-bed to replace all missing calls with
homozygous A2 calls.
--set-missing-var-ids [t]
: Given a template string with a '@' where the chromosome code should go and '#' where the bp coordinate belongs, --set-missing-var-ids assigns chromosome-and-bp-based IDs to unnamed variants. You may also use '$1' and '$2' to refer to allele names in the template string, and in fact this becomes essential when multiple variants share the same coordinate.
--new-id-max-allele-len [n] : Specify maximum number of leading characters
from allele names to include in new variant IDs (default 23).
--missing-var-code [string] : Change unnamed variant code (default '.').
[f] {chrcol} {IDcol} {skip} : Update variant chromosome codes.
[f] {cmcol} {IDcol} {skip} : Update centimorgan positions.
[f] {bpcol} {IDcol} {skip} : Update variant bp positions.
--update-name [f] {newcol} {oldcol} {skip} : Update variant IDs.
--update-alleles [fname] : Update variant allele codes.
--allele1234 <multichar> : Interpret/recode A/C/G/T alleles as 1/2/3/4.
With 'multichar', converts all A/C/G/Ts in allele names to 1/2/3/4s.
--alleleACGT <multichar> : Reverse of --allele1234.
--update-ids [f]
: Update sample IDs.
--update-parents [f] : Update parental IDs.
--update-sex [f] {n} : Update sexes.
Sex (1 or M = male, 2 or F = female, 0 = missing) is loaded from column n+2 (default n is 1).
--flip [filename]
: Flip alleles (A<->T, C<->G) for SNP IDs in the file.
--flip-subset [fn]
: Only apply --flip to samples in --flip-subset file.
--flip-scan-window [ct+1] : Set --flip-scan max variant ct dist. (def. 10).
--flip-scan-window-kb [x] : Set --flip-scan max kb distance (default 1000).
--flip-scan-threshold [x] : Set --flip-scan min correlation (default 0.5).
: Keep the allele order defined in the .bim file,
instead of forcing A2 to be the major allele. --real-ref-alleles also removes 'PR' from the INFO values emitted by --recode vcf{-fid/-iid}.
--a1-allele [f] {a1col} {IDcol} {skip} : Force alleles in the file to A1.
--a2-allele [filename] {a2col} {IDcol} {skip} :
Force alleles in the file to A2.
("--a2-allele [VCF filename] 4 3 '#'",
which scrapes reference allele assignments from a VCF file, is especially useful.)
--indiv-sort [m] {f} : Specify FID/IID sort order.
The following four modes are supported: * 'none'/'0' keeps samples in the order they were
Default for non-merge operations.
* 'natural'/'n' invokes 'natural sort', e.g.
'id2' < 'ID3' < 'id10'. Default when merging.
* 'ascii'/'a' sorts in ASCII order, e.g.
'ID3' < 'id10' < 'id2'.
* 'file'/'f' uses the order in the given file (named
in the second parameter).
For now, only --merge/--bmerge/--merge-list and
--make-bed/--make-just-fam respect this flag.
--with-phenotype <no-parents> <no-sex | female-2> : Include more sample info
in new .cov file.
--dummy-coding {N} <no-round> : Split categorical variables (n categories,
2 < n <= N, default N is 49) into n-1 binary dummy variables when writing covariate file.
--merge-mode [n]
: Adjust --{b}merge/--merge-list behavior based on a numeric code. 1 (default) = ignore missing calls, otherwise difference
-> missing
2 = only overwrite originally missing calls 3 = only overwrite when nonmissing in new file 4/5 = never overwrite and always overwrite, respectively 6 = report all mismatching calls without merging 7 = report mismatching nonmissing calls without merging
: With --merge/--bmerge/--merge-list, merge variants with different names but identical positions. (Exception: same-position chromosome code 0 variants aren't merged.)
: Make Mendel error checks consider samples with only one parent in the dataset.
: Make Mendel error checks consider (great-)grandparental genotypes when parental genotype data is missing.
--ld-window [ct+1] : Set --r/--r2 max variant ct pairwise distance (usu. 10).
--ld-window-kb [x] : Set --r/--r2 max kb pairwise distance (usually 1000).
--ld-window-r2 [x] : Set threshold for --r2 report inclusion (usually 0.2).
--ld-snp [var ID]
: Set first variant in all --r/--r2 pairs.
--ld-snps [vID...] : Restrict first --r/--r2 variant to the given ranges.
--ld-snp-list [f]
: Restrict first --r/--r2 var. to those named in the file.
: Generate the 'all' mode report when using --show-tags in
file mode.
--tag-kb [kbs]
: Set --show-tags max tag kb distance (default 250).
--tag-r2 [val]
: Set --show-tags min tag r-squared (default 0.8)
: Use two-column --show-tags (file mode) I/O format.
--ld-xchr [code]
: Set Xchr model for --indep{-pairwise}, --r/--r2, --flip-scan, and --show-tags. 1 (default) = males coded 0/1, females 0/1/2 (A1 dosage) 2 = males coded 0/2 3 = males coded 0/2, but females given double weighting
--blocks-max-kb [kbs]
: Set --blocks maximum haploblock span (def. 200).
--blocks-min-maf [cutoff]
: Adjust --blocks MAF minimum (default 0.05).
--blocks-strong-lowci [x]
: Set --blocks 'strong LD' CI thresholds (defaults
--blocks-strong-highci [x]
0.70 and 0.98).
--blocks-recomb-highci [x] : Set 'recombination' CI threshold (default 0.90).
--blocks-inform-frac [x]
: Force haploblock [strong LD pairs]:[total
informative pairs] ratios to be larger than this value (default 0.95).
--distance-wts exp=[x]
: When computing genomic distances, assign each variant a weight of (2q(1-q))^{-x}, where q is the loaded or inferred MAF.
--read-dists [dist file] {id file} : Load a triangular binary distance matrix
instead of recalculating from scratch.
--ppc-gap [val]
: Minimum number of base pairs, in thousands, between informative pairs of markers used in --genome PPC test. 500 if unspecified.
--min [cutoff]
: Specify minimum PI_HAT for inclusion in --genome report.
--max [cutoff]
: Specify maximum PI_HAT for inclusion in --genome report.
--homozyg-match [] : Set minimum concordance across jointly homozygous
variants for a pairwise allelic match to be declared.
--pool-size [ct]
: Set minimum size of pools in '--homozyg group' report.
--read-genome [fn] : Load --genome report for --cluster/--neighbour, instead
of recalculating IBS and PPC test p-values from scratch.
--ppc [p-val]
: Specify minimum PPC test p-value within a cluster.
--mc [max size]
: Specify maximum cluster size.
--mcc [c1] [c2]
: Specify maximum case and control counts per cluster.
--K [min count]
: Specify minimum cluster count.
--ibm [val]
: Specify minimum identity-by-missingness.
--match [f] {mv} : Use covariate values to restrict clustering.
Without --match-type, two samples can only be in the same cluster if all covariates match. The optional second parameter specifies a covariate value to treat as missing.
--match-type [f] : Refine interpretation of --match file.
The --match-type file is expected to be a single line with as many entries as the --match file has covariates; '0' entries specify 'negative matches' (i.e. samples with equal covariate values cannot be in the same cluster), '1' entries specify 'positive matches' (default), and '-1' causes the corresponding covariate to be ignored.
--qmatch [f] {m} : Force all members of a cluster to have similar
--qt [fname]
quantitative covariate values. The --qmatch file contains the covariate values, while the --qt file is a list of nonnegative tolerances (and '-1's marking covariates to skip).
--pca-cluster-names [...] : These can be used individually or in combination
--pca-clusters [fname]
to define a list of clusters to use in the basic --pca computation. (--pca-cluster-names expects a space-delimited sequence of cluster names, while --pca-clusters expects a file with one cluster name per line.) All samples outside those clusters will then be projected on to the calculated PCs.
--mds-plot [dims] <by-cluster> <eigendecomp> <eigvals> :
Multidimensional scaling analysis.
Requires --cluster.
--cell [thresh]
: Skip some --model tests when a contingency table entry is smaller than the given threshold.
--condition [var ID] <dominant | recessive> : Add one variant as a --linear
or --logistic covariate.
--condition-list [f] <dominant | recessive> : Add variants named in the file
as --linear/--logistic covs.
--parameters [...]
: Include only the given covariates/interactions in the --linear/--logistic models, identified by a list of 1-based indices and/or ranges of them.
--tests <all> {...} : Perform a (joint) test on the specified term(s) in the
--linear/--logistic model, identified by 1-based indices and/or ranges of them. If permutation was requested, it is based on this test. * Note that, when --parameters is also present, the
indices refer to the terms remaining AFTER pruning by
* You can use '--tests all' to include all terms.
--vif [max VIF]
: Set VIF threshold for --linear multicollinearity check (default 50).
--xchr-model [code] : Set the X chromosome --linear/--logistic model.
0 = skip sex and haploid chromosomes 1 (default) = add sex as a covariate on X chromosome 2 = code male genotypes 0/2 instead of 0/1 3 = test for interaction between genotype and sex
--lasso-select-covars {cov(s)...} : Subject some or all covariates to LASSO
model selection.
--adjust <gc> <log10> <qq-plot>
: Report some multiple-testing corrections.
--lambda [val]
: Set genomic control lambda for --adjust.
--ci [size]
: Report confidence intervals for odds ratios.
--pfilter [val]
: Filter out association test results with higher p-values.
--aperm [min perms - 1] {max perms} {alpha} {beta} {init interval} {slope} :
Set up to six parameters controlling adaptive permutation tests. * The first two control the minimum and maximum number of permutations that
may be run for each variant; default values are 5 and 1000000.
* The next two control the early termination condition.
100% * (1 - beta/2T) confidence interval is calculated for each empirical p-value, where T is the total number of variants; whenever this confidence interval doesn't contain alpha, the variant is exempted from further permutation testing. Default values are 0 and 1e-4.
* The last two control when the early termination condition is checked.
a check occurs at permutation #p, the next check occurs after [slope]p + [init interval] more permutations (rounded down). Default initial interval is 1, and default slope is 0.001.
: Save best max(T) permutation test statistics.
--mperm-save-all : Save all max(T) permutation test statistics.
--set-p [p-val]
: Adjust set test significant variant p-value ceiling (default 0.05).
--set-r2 {v} <write>
: Adjust set test significant variant pairwise r^2
ceiling (default 0.5). 'write' causes violating pairs to be dumped to {output prefix}.ldset.
--set-max [ct]
: Adjust set test maximum # of significant variants considered per set (default 5).
--set-test-lambda [v] : Specify genomic control correction for set test.
--border [kbs]
: Extend --annotate range intervals by given # kbs.
--annotate-snp-field [nm] : Set --annotate variant ID field name.
--clump-p1 [pval] : Set --clump index var. p-value ceiling (default 1e-4).
--clump-p2 [pval] : Set --clump secondary p-value threshold (default 0.01).
--clump-r2 [r^2]
: Set --clump r^2 threshold (default 0.5).
--clump-kb [kbs]
: Set --clump kb radius (default 250).
--clump-snp-field [n...]
: Set --clump variant ID field name (default 'SNP'). With multiple field names, earlier names
take precedence over later ones.
--clump-field [name...]
: Set --clump p-value field name (default 'P').
: Let --clump non-index vars. join multiple clumps.
: Request extended --clump report.
--clump-annotate [hdr...] : Include named extra fields in --clump-verbose and
--clump-best reports. (Field names can be separated with spaces or commas.)
--clump-range [filename]
: Report overlaps between clumps and regions.
--clump-range-border [kb] : Stretch regions in --clump-range file.
: Extract --clump index vars. from only first file.
: Exclude clumps which contain secondary results
from only one file.
: Report best proxy for each --clump index var.
--meta-analysis-snp-field [n...] : Set --meta-analysis variant ID, A1/A2
--meta-analysis-a1-field [n...]
allele, p-value, and/or effective sample
--meta-analysis-a2-field [n...]
size field names. Defauls are 'SNP',
--meta-analysis-p-field [n...]
'A1', 'A2', 'P', and 'NMISS',
--meta-analysis-ess-field [n...]
respectively. When multiple parameters are given to these flags, earlier names take precedence over later ones. Note that, if the numbers of cases and controls are unequal, effective sample size should be
4 / (1/[# cases] + 1/[# controls]).
: When a variant appears multiple times in in the same file, report that.
--gene-list-border [kbs]
: Extend --gene-report regions by given # of kbs.
--gene-subset [filename]
: Specify gene name subset for --gene-report.
--gene-report-snp-field [] : Set --gene-report variant ID field name (default
'SNP'). Only relevant with --extract.
--gap [kbs]
: Set '--fast-epistasis case-only' min. gap (default 1000).
--epi1 [p-value] : Set --{fast-}epistasis reporting threshold (default
5e-6 for 'boost', 1e-4 otherwise).
--epi2 [p-value] : Set threshold for contributing to SIG_E count (def. 0.01).
--je-cellmin [n] : Set required number of observations per 3x3x2 contingency
table cell for joint-effects test (default 5).
--q-score-range [range file] [data file] {i} {j} <header> :
Apply --score to subset(s) of variants in the primary score list based on e.g. p-value ranges. * The first file should have range labels in the first column, p-value
lower bounds in the second column, and upper bounds in the third column. Lines with too few entries, or nonnumeric values in the second or third column, are ignored.
* The second file should contain a variant ID and a p-value on each
nonempty line (except possibly the first).
Variant IDs are read from
column #i and p-values are read from column #j, where i defaults to 1 and j defaults to i+1. The 'header' modifier causes the first nonempty line of this file to be skipped.
--parallel [k] [n] : Divide the output matrix into n pieces, and only compute
the kth piece. The primary output file will have the piece number included in its name, e.g. plink.rel.13 or plink.rel.13.gz if k is 13. Concatenating these files in order will yield the full matrix of interest. (Yes, this can be done before unzipping.) N.B. This generally cannot be used to directly write a symmetric square matrix. Choose square0 or triangle shape instead, and postprocess as necessary.
--memory [val]
: Set size, in MB, of initial workspace malloc attempt. (Practically mandatory when using GNU parallel.)
--threads [val]
: Set maximum number of concurrent threads. This has one known limitation: some BLAS/LAPACK linear algebra operations are multithreaded in a way that PLINK cannot control. If this is problematic, you should recompile against single-threaded BLAS/LAPACK.
--d [char]
: Change variant/covariate range delimiter (normally '-').
--seed [val...]
: Set random number seed(s). Each value must be an integer between 0 and 4294967295 inclusive.
--perm-batch-size [val] : Set number of permutations per batch for some
permutation tests.
--output-min-p [p] : Specify minimum p-value to write to reports.
: Use slower, more crash-resistant logging method.

Primary methods paper: Chang CC, Chow CC, Tellier LCAM, Vattikuti S, Purcell SM, Lee JJ (2015) Second-generation PLINK: rising to the challenge of larger and richer datasets. GigaScience, 4.

For further documentation and support, consult the main webpage ( ) and/or the mailing list ( ).