Open iamyingzhou opened 7 months ago
I've just come to the realization that association tests using JIT pre-trained prediction models differ from the MR-JTI method. Is there a connection between the two?
Here is the code that I am using now Rscript ${src_dir}/predixcan_r.r \ --asso_test \ --db_path ${dir}/JTI_Liver.db \ --cov_path ${dir}/JTI_Liver.txt.gz \ --gwas_path ${dir}/LDLq.txt.gz \ --gwas_variant_col rsid \ --gwas_beta_col beta \ --gwas_se_col se \ --gwas_eff_allele_col eff_allele \ --gwas_ref_allele_col ref_allele \ --asso_out_path ${dir}/result_ldl_liver.txt \ --parallel
Hi Yingzhou, Thanks for reaching out! predixcan_r.r --asso_test performs GReX-trait association test (the last step of TWAS), while MR-JTI is a Mendelian randomization test. For P+z-score, I'm afraid not. We recommend using beta+se. Dan
On Tue, Nov 28, 2023 at 11:12 PM Yingzhou Shi @.***> wrote:
I've just come to the realization that association tests using JIT pre-trained prediction models differ from the MR-JTI method. Is there a connection between the two?
Here is the code that I am using now Rscript ${src_dir}/predixcan_r.r \ --asso_test \ --db_path ${dir}/JTI_Liver.db \ --cov_path ${dir}/JTI_Liver.txt.gz \ --gwas_path ${dir}/LDLq.txt.gz \ --gwas_variant_col rsid \ --gwas_beta_col beta \ --gwas_se_col se \ --gwas_eff_allele_col eff_allele \ --gwas_ref_allele_col ref_allele \ --asso_out_path ${dir}/result_ldl_liver.txt \ --parallel
— Reply to this email directly, view it on GitHub https://github.com/gamazonlab/MR-JTI/issues/11#issuecomment-1830044364, or unsubscribe https://github.com/notifications/unsubscribe-auth/AH5HAMLCJQKZDS2PWQGBIHTYGX5M5AVCNFSM6AAAAAA75XXSTSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMZQGA2DIMZWGQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Thanks for your wonderful work. I'm looking to conduct summary-stat-based association test on a meta-analysis GWAS. The output from the METAL software provides only the P-value and Z-score columns, without beta and standard error. Is it feasible to run association test with just the P-value and Z-score ?