Home
Adj-HE (Adjusted HE) is a computationally efficient method to estimate Single Nucleotide Polymorphism (SNP)-heritability in presence of population substructure for biobank-scale data. It is a simplification of the Haseman- Elston regression (HE). For details of this statistical method, please refer/cite:
sudo docker build -f Dockerfile -t adjhe:1.1 .
sudo docker save adjhe:1.1 > AdjHE_1.1.tar
module load singularity
cd ~/tools/Basu_herit
singularity build Estimate.sif docker-archive://AdjHE_1.1.tar
Start an interactive job on mesabi that you will use to build the container. The size of the /tmp directory allocated with the –tmp flag will be dependent on the size of the container you are trying to build. A rule of thumb that has seemed to work okay is to request at least twice the size of the container you are trying to build.
srun -N 1 --ntasks-per-node=1 --tmp=100g --mem-per-cpu=30g -t 3:00:00 -p interactive --pty bash
See simulation 2 in Section 3 "Results" in the paper. the efficiency of this method of estimation allows for users to do some amount of model selection nicluding with the number of PC's which could drastically impact heritability estimates.
Lin, Z., Seal, S., & Basu, S. (2020). Estimating SNP heritability in presence of population substructure in large biobank-scale data. bioRxiv. https://doi.org/10.1101/2020.08.05.236901