Data format

To run LDAK, you require genotypes and phenotypes, while it is common to also include covariates. When performing gene-based analysis, you additionally require a gene annotations file.

Genotypes

LDAK accepts genetic data in three formats, using one of the following arguments:

To filter either samples or predictors, see Data Filtering. LDAK is usually applied to SNP data, in which case all predictors take values between 0 and 2 (representing the count of the A1 allele). However, LDAK can also be applied to other datatypes; for this your data should be in either gen or SP format and you should use the option --SNP-data NO.

Notably, LDAK is not able to process data of pgen format and vcf format. For application of LDAK-KVIK on these data types, we recommend first converting the genotype data to Binary PLINK format.

Phenotypes

Phenotype files should be in PLINK format. The first two columns should provide sample IDs, with subsequent columns providing the phenotype values. This can be one phenotype, or multiple phenotypes. An example phenotype input file looks like this:

FID IID Pheno1 Pheno2
1 1 0.25 25
2 2 0.42 12
3 3 0.32 36
...

LDAK accepts phenotype files both with and without a header. In case a header is included, the first two elements must be named either FID & IID or ID1 & ID2. If a phenotype file contains more than one phenotype (i.e., more than three columns), the user must specify the phenotype to analyse in LDAK using the option --mpheno <integer> (to specify by number) or --pheno-name <name> (to specify by name). Alternatively, some functions (most notably REML) allow for testing of all phenotypes by adding –mpheno -1.

Missing values should be denoted by NA. Note that whereas PLINK also treats -9 as missing, this is not the case in LDAK. Binary phenotypes should only take values:

• 0 (control), 1 (case) or NA; or
• 1 (control), 2 (case) or NA

When a phenotypic values is NA for a particular sample, then that sample is excluded from the analysis.

Covariates

Covariate files should be in PLINK format. The first two columns should provide the sample FIDs and IIDs, with subsequent columns providing covariate values. For example:

FID IID PC1 PC2 PC3 Age Sex
1 1 0.42 -0.12 1.23 41 0
2 2 -0.1 0.23 0.49 64 1
3 3 0.21 -0.14 -0.23 27 0
...

If you have quantitative covariates, you should use --covar, while if you have categorical covariates, you should use --factors. Note that if a categorical covariate has U unique values, LDAK will (internally) replace it with U-1 indicator variables (LDAK will give an error if the total number of indicator variables is greater than half the sample size).

If providing quantitative covariates, you can use --covar-numbers to specify a subset of covariates (using commas and/or dashes to specify multiple covariates). For example, --covar-numbers 1,2,4-6,8 tells LDAK to retain Covariates 1, 2, 4, 5, 6 and 8. If the covariate file has a header name, you can instead use --covar-names (use commas to specify multiple covariates).

Missing covariate values should be denoted by NA (missing values are replaced by the mean value of the corresponding covariate).

Gene annotations

Gene-based association analysis requires a gene annotations file, which should be in ‘Browser Extensible Data’ format. This means that it should contains one row for each gene, and four columns, that report the gene name and chromosome, and its start and end basepairs. For example, if the file contained a single line, with entries “ABC 7 0 10”, this would indicate there is one gene, called ABC, which spans the first ten basepairs of Chromosome 7.

It is possible to download gene annotation files (hg37 and hg38) in a Linux terminal using:

wget https://dougspeed.com/wp-content/uploads/RefSeq_GRCh37.txt
wget https://dougspeed.com/wp-content/uploads/RefSeq_GRCh38.txt

The gene annotations files can also be downloaded from GitHub.

Data filtering

For most commands, it is possible to filter on SNPs or filter on samples using the following arguments:

In addition, for many commands, you can add --pheno <phenofile> and LDAK will only consider samples for which phenotypes are available.

Note that if you would like to filter predictors based on minor allele frequency, variance, missingness or information score, you must first remake the data (this filtering can not be done on-the-fly).