Closed humanpaingeneticslab closed 9 months ago
@humanpaingeneticslab great questions!
Q1. h2_snp = 2 x F x (1-F) x ( BETA_MEAN^2 + BETA_SD^2 )
Q2. Yes, these are LD-corrected.
Q3. We used SNPs with MAF>0.1%. There are various studies suggesting that this captures most of the heritability, though it's not a closed question (see e.g. Schoech et al. 2019 and Gazal et al. 2018)
@humanpaingeneticslab can I close the issue?
Yes, many thanks!
On Nov 5, 2023, at 5:38 AM, Omer Weissbrod @.***> wrote:
@humanpaingeneticslab can I close the issue? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>
Dear polyfun users,
This isn't an issue but rather few simple questions - couldn't find a better place to ask :-)
Q1. Is it possible from polyfun's output to derive the per-SNP heritability (h2_snp)? Polyfun's output have these columns: "CHR SNP BP A1 A2 SNPVAR N Z P PIP BETA_MEAN BETA_SD CREDIBLE_SET". Might need allele frequencies F too (I can get from other sources)?
[ ] h2_snp = BETA_MEAN [ ] h2_snp = BETA_MEAN^2 [ ] h2_snp = 2 x F x (1-F) x BETA_MEAN^2 [ ] h2_snp = BETA_MEAN^2 + BETA_SD^2 [ ] h2_snp = 2 x F x (1-F) x ( BETA_MEAN^2 + BETA_SD^2 ) [ ] h2_snp = SNPVAR [ ] h2_snp = SNPVAR^2 [ ] h2snp = 2 x F x (1-F) x SNPVAR^2 [ ] none of the above, but rather: ____
Q2. If so, are these per-SNP heritability estimates LD-corrected? In other words, if I sum heritabilities over all SNPs I won't get like 10000% (from summing up high-LD SNPs) :-/
Q3. How low of a minor allele frequency the heritability estimate would be good for? There's a MAF floor I guess in the LD matrices used for input to polyfun, right?
Many Thanks!