Feature
Compute Enrichment score output for LDSC for a better effect size estimation.
Helpful code: https://github.com/kkdey/GSSG/tree/master/code/ldsc
"We further define the Enrichment score (E-score) of a gene program as the difference between the heritability enrichment of the SNP annotation corresponding to the gene program of interest and the SNP annotation corresponding to a gene program assigning a probabilistic grade of 1 to all protein-coding genes with at least one enhancer-gene link in the relevant tissue (Methods). We use the p-value of the E-score as our primary metric, assessing statistical significance using a genomic block-jackknife as in our previous work11 [=S-LSSC] because the p-values can be compared across datasets, whereas the E-score magnitude can vary substantially in gene programs dominated by a smaller (or larger) number of genes. We primarily focus on E-scores greater than 2, because E-scores that are statistically significant but small in magnitude may have more limited biological importance, as the cell types underlying these E-scores may be tagging other causal cell types "
Challenge
This requires "h2" mode of LDSC which takes significant computing time
Reference
Taking from:
Identifying disease-critical cell types and cellular processes across the human body by integration of single-cell profiles and human genetics
Karthik A. Jagadeesh, Kushal K. Dey, Daniel T. Montoro, Steven Gazal, Jesse M. Engreitz, Ramnik J. Xavier, Alkes L. Price, Aviv Regev
https://www.biorxiv.org/content/10.1101/2021.03.19.436212v2
Feature
Compute Enrichment score output for LDSC for a better effect size estimation.
Helpful code: https://github.com/kkdey/GSSG/tree/master/code/ldsc
"We further define the Enrichment score (E-score) of a gene program as the difference between the heritability enrichment of the SNP annotation corresponding to the gene program of interest and the SNP annotation corresponding to a gene program assigning a probabilistic grade of 1 to all protein-coding genes with at least one enhancer-gene link in the relevant tissue (Methods). We use the p-value of the E-score as our primary metric, assessing statistical significance using a genomic block-jackknife as in our previous work11 [=S-LSSC] because the p-values can be compared across datasets, whereas the E-score magnitude can vary substantially in gene programs dominated by a smaller (or larger) number of genes. We primarily focus on E-scores greater than 2, because E-scores that are statistically significant but small in magnitude may have more limited biological importance, as the cell types underlying these E-scores may be tagging other causal cell types "
Challenge
This requires "h2" mode of LDSC which takes significant computing time
Reference
Taking from:
Identifying disease-critical cell types and cellular processes across the human body by integration of single-cell profiles and human genetics
Karthik A. Jagadeesh, Kushal K. Dey, Daniel T. Montoro, Steven Gazal, Jesse M. Engreitz, Ramnik J. Xavier, Alkes L. Price, Aviv Regev
https://www.biorxiv.org/content/10.1101/2021.03.19.436212v2