h1, h12, rho are with NAN

[StellaXia]

Hi Mingxuan,

Thanks a lot for your great work of XPASS and detailed examples.

I applied your software on our data with unknown population ancestry information, therefore, I chose the sd_method to be "Chromosome".

The code is as following.
fit <-XPASS(file_z1 = train_sumstats,file_z2 = ukb_sumstats,file_ref1 = ref_Train,
file_ref2 = ref_EUR,
file_cov1 = cov_Train,file_cov2 = cov_EUR,
file_predGeno = test,
compPRS=T,
sd_method="Chromosome",compPosMean = T,
file_out = paste0("BMI_GCV",GCV))

However, the results of h1, h12 and rho are Inf or NAN. Grateful if there is any suggestions!

Summary statistics file 1: Train_GCV1_sumstats_format.txt
Summary statistics file 2: BMI_ukb_sumstat_format_all.txt
Reference file 1: Train.GCV1
Reference file 2: 1000G.EUR.QC.hm3.ind
Covariates file 1: covars_v2.GCV1.txt
Covariates file 2: 1000G.EUR.QC.hm3.ind.pc10.txt
Test genotype file : Test.GCV1
Reading data from summary statisitcs...
560899 and 3777871 SNPs found in summary statistics files 1 and 2.
Reading SNP info from reference panels...
560899 and 1313833 SNPs found in reference panel 1 and 2.
Reading SNP info from test genotype file...
560899 SNPs found in test file.
107096 SNPs are matched in all files.
147 SNPs are removed because of ambiguity; 106949 SNPs remained.
Calculating kinship matrix from the both reference panels...
0 SNPs are removed because of no variation; 106949 SNPs remained.
13463 SNPs in the second reference panel are alligned for alleles according to the first.
0 SNPs have different minor alleles in population 1, z-scores are corrected according to reference panel.
6288 SNPs have different minor alleles in population 2, z-scores are corrected according to reference panel.
Assigning SNPs to LD Blocks...
Calculate PVE...
h1 h2 h12 rho
[1,] Inf 0.17083593 NaN NaN
[2,] NA 0.01244687 NA NA
Error in if (abs(fit$H[1, 4]) > 1) { :
missing value where TRUE/FALSE needed
Calls: XPASS
Execution halted

Thanks!

Stella.

[mxcai]

Hi Stella,

Thank you for your interest in our package.

It seems the estimation of heritability works well for the UKB BMI data (BMI_ukb) but produces Inf value with your dataset (Train_GCV1). This may be caused by NA values in the summary data or covariates file. Would you double check if there is any such missing values in your own input files?

Best,
Mingxuan