Introduction

March 2020. Our most recent work, "Evaluating and improving heritabilty models using summary statistics", has just been published in Nature GeneticsĀ (for a free version click here). This work provides a major update to SumHer, our tool for performing heritability analyses using summary statistics. See the SumHer section for more details.

I am currently updating LDAK to incorporate the new features. If you send me an email (see Help page for address), I will let you know as soon as the new version is ready.

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

LDAK does the following:

Calculates LDAK SNP weightings
Computes weighted and unweighted kinship matrices
Performs generalised REML (also computes BLUP estimates)
Efficient mixed-model analysis
Gene-based Association Testing
MultiBLUP Prediction

LDAK also includes SumHer, our tool for performing heritability analysis using summary statistics (i.e., estimate SNP heritability, confounding bias, heritability enrichments, genetic correlations, and the selection-related parameter alpha).

LDAK is additionally able to:

Manipulate kinship matrices (add/subtracting/converting)
Perform Haseman-Elston and PCGC Regression
Compute Genetic risk scores
Perform principal component analysis
Generate simulated genotypes and phenotypes
Accommodate and convert many data formats

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

LDAK is explained in the following publications

Speed et al., AJHG 2012 (explains the LDAK SNP weightings)
Speed et al., Nature Genetics 2017 (explains the importance of the heritability model)
Speed and Balding, Nature Genetics 2019 (introduces SumHer)
Speed et al., Nature Genetics 2020 (compares heritability models using summary statistics)

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

The original use of LDAK was to adjust for linkage disequilibrium (LD) by calculating SNP weightings which downweight the contribution of SNPs in highly tagged regions. These SNP weightings could then be used to compute LD-adjusted kinship matrices, and in turn better estimate SNP heritability. The idea is outlined here, or see our AJHG paper for full details.

Originally, LDAK required individual-level data (genotypes and phenotypes), but we have now added SumHer, which allows you to perform heritability analysis using just summary statistics.

LDAK also includes MultiBLUP, our method for creating linear prediction models. This improves upon standard BLUP by allowing for multiple kinship matrices (see our Genome Research paper here). Either the user can provide these kinship matrices, or they can be determined using Adaptive MultiBLUP.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Please email doug<at>aias<dot>au<dot>dk with any questions or bug reports.