LDAK is a command-line software. Each LDAK command starts with the name of the executable file, followed by arguments. The command must include one main argument, which tells LDAK which feature to use and either the prefix of the output files or the name of the output folder. It will usually also include one or more other arguments. For example, if you are using the Linux version of LDAK, you might use the command
./ldak5.1.linux --calc-stats results --bfile data
The main argument is --calc-stats results, which tells LDAK to calculate predictor statistics and save the output files with prefix results. Meanwhile, the other argument --bfile data tells LDAK that the data are stored in Binary PLINK format in the files data.bed, data.bim and data.fam.
Below is a list of the possible main arguments (this list appears whenever you run LDAK without specifying a main argument).
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Heritability analysis using individual-level data:
--cut-weights <folder> - cut predictors into sections ready for calculating weightings
--calc-weights <folder> - calculate weightings for one section
--join-weights <folder> - join weightings across sections
--calc-weights-all <folder> - calculate weightings for all sections in one step
--cut-kins <folder> - cut predictors into partitions ready for calculating kinships
--calc-kins <folder> - calculate kinship matrix for one partition
--join-kins <folder> - join kinship matrices across partitions
--calc-kins-direct <outfile> - calculate kinship matrix in one step
--reml <outfile> - regress phenotype on kinships and/or regions
--calc-blups <outfile> - calculate BLUP effect size estimates (using results from --reml)
--he <outfile> - perform Haseman-Elston Regression (an alternative to --reml)
--pcgc <outfile> - perform PCGC Regression (recommended instead of --reml for binary traits)
--fast-he <outfile> - perform fast (but approximate) Haseman-Elston Regression
--fast-pcgc <outfile> - perform fast (but approximate) PCGC Regression
Association testing:
--linear <outfile> - perform standard or mixed-model linear regression
--logistic <outfile> - perform logistic regression
--solve-null <outfile> - perform two-part mixed-model linear regression (for complex models)
--cut-genes <folder> - cut predictors into genes (or fixed-length chunks)
--calc-genes-kins <folder> - calculate kinships for each gene/chunk in one partition
--calc-genes-reml <folder> - calculate association test for each gene in one partition
--join-genes-reml <folder> - join association test results across partitions (also computes more accurate p-values)
Heritability analysis using summary statistics:
--calc-tagging <outfile> - calculate tagging file for use with --sum-her or --sum-cor
--join-tagging <outfile> - join tagging files across predictors
--merge-tagging <outfile> - combine tagging files across categories
--reduce-tagging <outfile> - extract categories from a tagging file
--sum-hers <outfile> - estimate heritabilities from summary statistics
--sum-cors <outfile> - estimate genetic correlations from summary statistics
--calc-exps <outfile> - calculate expected heritability tagged by predictors (using results from --sum-her)
--calc-posts <outfile> - calculate posterior estimate of predictor effect sizes (using results from --calc-exps)
Prediction:
--ridge <outfile> - construct a ridge regression prediction model
--bolt <outfile> - construct a Bolt prediction model
--bayesr <outfile> - construct a BayesR prediction model
--calc-cors <outfile> - calculate predictor-predictor correlations for use with --mega-prs
--join-cors <outfile> - join correlations across predictors
--mega-prs <outfile> - construct lasso, ridge regression, Bolt and BayesR prediction models from summary statistics
--pseudo-summaries <outfile> - generate partial summary statistics (used to train MegaPRS prediction models)
Other features:
--thin <outfile> - prune predictors based on pairwise correlations
--thin-tops <outfile> - prune highly-associated predictors based on pairwise correlations
--find-tags <outfile> - search for the predictors most correlated with those in a scorefile
--remove-tags <outfile> - search for all predictors correlated with those in a target file
--filter <outfile> - fiter samples based on relatedness
--add-grm <outfile> - combine kinship matrices
--sub-grm <outfile> - subtract from one kinship matrix the contributions of other kinship matrices
--convert-gz <outfile> - convert kinship matrix saved in gz (old GCTA) format
--convert-raw <outfile> - convert kinship matrix saved as a text file
--calc-sim-grm <outfile> - calculate correlations between pairs of kinship matrices
--pca <outfile> - compute the principal component axes of a kinship matrix
--calc-pca-loads <outfile> - calculate the principal component predictor loadings (using results from --pca)
--decompose <outfile> - eigen-decompose a kinship matrix
--adjust-grm <outfile> - adjust a kinship matrix for covariates
--gxemm-iid / --gxemm-free <outfile> - calculate environmental kinship matrices
--calc-stats <outfile> - calculate predictor allele frequencies, call-rates and possibly information scores
--calc-scores <outfile> - calculate one or more polygenic risk scores
--make-phenos <outfile> - simulate phenotypes
--make-snps <outfile> - simulate SNP data
--jackknife <outfile> - measure prediction accuracy, obtaining standard deviations via block jackknifing
--cut-folds <outfile> - divide samples in preparation for performing K-fold cross-validation
--find-gaussian <outfile> - estimate the selection-related parameter alpha (using results from --sum-hers)
--make-bed / --make-sp / --make-sped / --make-speed / --make-gen <outfile> - convert to bed / sp / sped / speed / gen format
--condense-bed / condense-sp / condense-sped / condense-speed <outfile> - condense to bed / sp / sped / speed format
--calc-sim-data <outfile> - calculate concordance between two datasets
