Single-SNP Analysis

LDAK can perform one-predictor-at-a-time analysis, either using standard linear regression or mixed-model linear regression. The argument for standard linear regression is

--linear <output>

which has only two required options, but many optional ones.

--bfile/--gen/--sp/--speed <prefix> - specifies the data files (see File Formats).

--pheno <phenofile> - specifies the response (in PLINK format). Individuals without a phenotype will be excluded. If <phenofile> contains more than one phenotype, specify which should be used with --mpheno.

--covar <covarfile> - provide covariates (in PLINK format) as fixed effects in the regression. I would generally include 11 covariates: sex, plus 5 in-data PCs and 5 projection PCs (see Principal Component Analysis for how to compute the latter).

--top-preds <list_of_predictors> - provide a list of predictors  to include as fixed effects. Usually, these represent (a pruned subset of) highly-associated predictors, in which case LDAK will perform a conditional analysis.

--prevalence <float> - if the phenotype is binary, then specify the population prevalence to convert effect sizes to the liability scale.

--permute YES - the phenotypic values will be shuffled. This is useful if wishing to perform permutation analysis to see the distribution of p-values / test statistics when there is no true signal.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

The commands for performing mixed-model linear regression are identical to those for performing standard linear regression, except that you must also specify a kinship matrix.

Usually, this is achieved by adding --grm <grmstem> to specify a kinship matrix. For heritability estimation, we advise computing weighted kinship matrices, but for mixed-model linear regression, it normally suffices to construct an unweighted kinship matrix using a pruned subset of predictors. For this, we can use the following commands

--thin <output> - to prune the predictors (e.g., using --window-prune 0.1 and --window-cm 1).

--calc-kins-direct <grmstem> with --ignore-weights YES, --power -1 and --extract <output>.in - to compute the kinship matrix.

See Get Kinships for detailed instructions for constructing kinship matrices.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

As when performing REML, an alternative to providing a kinship matrix is to use --region-number and --region-prefix to provide one or more regions. In this case, you must use --weights to specify the predictor weightings (or --ignore-weights YES to set them to 1)  and --power to indicate how to scale predictors (we advise using -0.25). In theory, this can be used to include significantly-associated regions as random effects, but it is not an option we have explored in detail.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

The main output files is <output>.assoc, which provides full results for each predictor. <output>.summaries provides the results necessary for performing analyses using SumHer, while <output>.pvalues contains just the p-values (these can be supplied when cutting genes prior to Gene-based Analysis). <output>.coeff provides estimates of the fixed effects, while <output>.score contains six polygenic risk scores (which can be used to construct Profile Scores).