This page lists only the novel tools contained within LDAK. If you would like to see a full list of features (i.e., novel tools, tools adopted from other software, and auxiliary tools), please click here.
Note that some of these tools exist due to inadequacies in existing versions. In particular, most existing tools assume the GCTA Model, and do not let the user change this. Therefore, I have developed generalized versions that allow the user to specify the Heritability Model.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
LDAK weightings (arguments --cut-weights, --calc-weights, --join-weights, --calc-weights-all). A tool for finding weights that account for uneven linkage disequilibrium.
REML (argument --reml). A tool for estimating variance components. This improves the basic REML algorithm in the following ways: when provided with one kinship matrix, it reduces runtime by first performing an eigen-decomposition; it can accommodate regions; it reduces memory requirements by allowing kinship matrices to be read-on-the-fly, rather than stored.
Fast HE and PCGC Regression (arguments --fast-he, --fast-pcgc). A generalized version of the method for estimating SNP heritability by Pazokitoroudi et al..
Generalized mixed-model association analysis (arguments --linear, --solve-null). A tool for performing efficient mixed-model association analysis with more than kinship matrix.
GBAT (arguments --cut-genes, --calc-genes-reml, --join-genes-reml). A tool for performing highly-efficient and powerful gene/set-based association testing. This is a generalized version of the method of Listgarten et al., that also can be run using only summary statistics.
SumHer (arguments --calc-tagging, --sum-hers, --sum-cors). A tool for investigating genetic architecture using summary statistics. This begun as a generalized version of LD Score Regression (i.e., a tool for estimating SNP heritability, heritability enrichments and genetic correlations), but now also provides a measure of model fit and enables estimation of the selection-related parameter alpha.
Ridge Predict (argument --ridge). A generalized version of ridge-regression, that is also highly-efficient.
MegaPRS (argument --mega-prs with option --model). This contains four tools for constructing prediction models using summary statistics, which are (in effect) generalized and more efficient versions of lassosum, SBLUP, LDPred and SBayesR.
Pseudo Summaries (argument --pseudo-summaries). This is a tool for dividing a single set of summary statistics into two sets that mimic independent training and test summary statistics.
Quality Control (arguments depend on aim). I have developed robust tests for inflation due to cryptic relatedness (population structure or familial relatedness) and due to genotyping errors.