Some issues encountered during the use of snp_ldpred2_auto

[JhGoodLuck]

hi,
I would like to know if LDpred-auto can be applied to smaller summary data (such as cis-QTL, with SNP numbers ranging from several hundred to several thousand).
I am currently testing the LDpred2 method using a pQTL dataset. After calculating the corr and h2_est, I can obtain a set of estimates using snp_ldpred2_inf. However, the beta_est obtained by snp_ldpred2_auto are all NA. I am not sure if this is due to the insufficient number of SNPs, or if there are any other possible reasons.
The data I tested has a heritability h2_est of 0.016 and a SNP number of 745. vec_p_init = seq_log(1e-4, 0.9, length.out = 10).
Thanks

[privefl]

Have you tried LDpred2-grid as well?

What do you use for use_MLE, allow_jump_sign, and shrink_corr in LDpred2-auto?
In your case, it is very important that you use use_MLE = FALSE.

[JhGoodLuck]

Yes, I tested 168 sets of parameters using snp_ldpred2_grid and they all ran normally. Only one set had a NA value. But when using snp_ldpred2_auto would all the values be NA. For the other parameters, I used the default settings throughout the process and use_MLE was set to TRUE.