Error in if (max(abs(pred - pred0)) < 1e-06) break

[Mahantesh-Biradar]

I'm using the ‘1.12.1’ version of bigsnpr. I have seem to have trouble in the LD calculation step. How can I address this issue?

# calculate LD
> for (chr in 1:22) {
+   # Extract SNPs that are included in the chromosome
+   ind.chr <- which(info_snp$chr == chr)
+   ind.chr2 <- info_snp$`_NUM_ID_`[ind.chr]
+   # Calculate the LD
+   corr0 <- snp_cor(
+     genotype2,
+     ind.col = ind.chr2,
+     ncores = NCORES,
+     infos.pos = POS2[ind.chr2],
+     size = 3 / 1000
+   )
+   if (chr == 1) {
+     ld <- Matrix::colSums(corr0^2)
+     corr <- as_SFBM(corr0, tmp)
+   } else {
+     ld <- c(ld, Matrix::colSums(corr0^2))
+     corr$add_columns(corr0, nrow(corr))
+   }
+ }

There were 11 warnings (use warnings() to see them)

> warnings()
Warning messages:
1: NA or NaN values in the resulting correlation matrix.
2: NA or NaN values in the resulting correlation matrix.
3: NA or NaN values in the resulting correlation matrix.
4: NA or NaN values in the resulting correlation matrix.
5: NA or NaN values in the resulting correlation matrix.
6: NA or NaN values in the resulting correlation matrix.
7: NA or NaN values in the resulting correlation matrix.
8: NA or NaN values in the resulting correlation matrix.
9: NA or NaN values in the resulting correlation matrix.
10: NA or NaN values in the resulting correlation matrix.
11: NA or NaN values in the resulting correlation matrix.

> # We assume the fam order is the same across different chromosomes
> fam.order <- as.data.table(obj.bigSNP$fam)
> # Rename fam order
> setnames(fam.order,
+          c("family.ID", "sample.ID"),
+          c("FID", "IID"))
> #4. Perform LD score regression
> df_beta <- info_snp[,c("beta", "beta_se", "n_eff", "_NUM_ID_")]
> ldsc <- snp_ldsc(   ld, 
+                     length(ld), 
+                     chi2 = (df_beta$beta / df_beta$beta_se)^2,
+                     sample_size = df_beta$n_eff, 
+                     blocks = NULL)
Error in if (max(abs(pred - pred0)) < 1e-06) break : 
  missing value where TRUE/FALSE needed
> h2_est <- ldsc[["h2"]]
Error: object 'ldsc' not found
> h2_est #[1] 0.457358
Error: object 'h2_est' not found

[privefl]

NA or NaN values in the resulting correlation matrix

This has already been discussed in other issues here. Please have a look at them by searching through the issues.

Please close the issue if you find an answer to your question.
Otherwise, please provide more details on why your problem is different from those already reported.

[Mahantesh-Biradar]

Thanks Florian. I'm looking into this and let me update this asap.

[Mahantesh-Biradar]

Hi Florian,

I have already excluded rare variants (MAF < 0.01) and I have tried to run this step without imputing missing values and also imputation using genotype2 <- snp_fastImputeSimple(genotype).

In both the cases, I see the following error: NA or NaN values in the resulting correlation matrix.

[privefl]

You should be able to locate missing values with something like Matrix::which(is.na(corr0), arr.ind = TRUE).

[Mahantesh-Biradar]

Thanks Florian

[privefl]

So, what was the problem?

[Mahantesh-Biradar]

I did the QC of genotype data again and filtered the SNPs based on HWE. Then there was no issue at this step.

…

On Sun, Mar 12, 2023 at 2:53 PM Florian Privé ***@***.***> wrote: So, what was the problem? — Reply to this email directly, view it on GitHub <#414 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB4AOXYMSWNMM5TD3KT2UGLW3XPQLANCNFSM6AAAAAAVVMTT3M> . You are receiving this because you modified the open/close state.Message ID: ***@***.***>