the out put file like this:
variable beta beta_CI_lower beta_CI_upper CI_significance
expression -0.709540602 -0.901773199 -0.551193179 sig
ldsc -0.062807264 -0.233198711 0.105520592 nonsig
rs3902354 0 0 0 nonsig
rs68104325 0 0 0 nonsig
rs585362 0 0 0 nonsig
rs17035665 0 0 0 nonsig
the variable column inclueded expression,ldsc and followed by 51 rsid. I do not understand how this output to determin the causal inference for a gene. I searched your result in you paper S8_MR-JTI:
geneid genename Bonferroni_adjusted_CI_lower Bonferroni_adjusted_CI_upper Significance map_to_well-known_genes map_to_mouse_cholesterol_modules
ENSG00000002834 LASP1 0.000 0.000 nonsig NA Lasp1
ENSG00000003056 M6PR 0.000 0.000 nonsig NA M6pr
ENSG00000004142 POLDIP2 0.180 0.525 sig NA NA
ENSG00000005981 ASB4 0.000 0.000 nonsig NA NA
Since there were not any gene id in output file,how could transfored the mrjti_result.csv to S8_MR-JTI result?
Problem2:
The input file like this in example:
rsid effect_allele_gwas ldscore eqtl_beta eqtl_se eqtl_p gwas_beta gwas_se gwas_p
rs10449300 G 16.3387 0.122073958 0.114470828 0.287482148 -0.00067683 0.0025788 0.79296
rs10494095 G 9.46725 0.021090531 0.121112593 0.861926553 -0.0025397 0.0032695 0.43728
rs10857790 T 11.4308 0.056750248 0.137096339 0.679345557 -0.0059203 0.0030763 0.054298
I do have download the eQTL info in GTEx, but since the eqtl_beta, eqtl_se, eqtl_p were caculated for a target gene and I found that some SNPs are eQTL for more than one gene of interest in the GTEx data, so beta (slope) and standard_error (slope_se) differ among genes. So i am not sure how to merge this info to the file.
Problem3:
You recomend SNPs need to be pruned before runing MR-JTI, does it mean i prune first before GWAS and then selected SNPs within the Bonferroni correction genes' window? Total number of genes tested (Bonferroni correction will be applied), how should i selected snps for this genes? by gene position window? The MR-JTI was run by once for all selected genes or for genes one by one?
I am very interested in this method and i am new for this, sorroy for so many easy problems. look forward for your help.
Thanks.
The script provides the MR-JTI test for each gene at a time. The first line "expression ... " gives you the results (the effect of the expression on the trait) for this gene. The S8_MR-JTI summarized the results across multiple genes.
You will need to manually generate a file for each gene and use it as the input. It is fine that a SNP is eQTL for multiple genes, especially when the gene expressions are correlated.
You can perform pruning anytime since it only depends on the LD. Or, you can clump the SNP based on the p-value from GWAS. Yes. We suggest using the SNPs within 1 Mb of the gene (1 Mb on both sides of the gene body). The script runs for one gene at a time.
Please let me know if you have any questions.
Dan
Hi @egamazon @zdangm @gamazonlab Thank you for the wonderful method. I met some problems when i try to use the MR-JTI analysis.
First, i run the MR-JTI with the example file,mrjti_example.txt. script_dir='${path}/MR-JTI-master/mr'
example_file_dir='${path}/MR-JTI-master/mr'
Rscript ${script_dir}/MR-JTI.r \ --df_path ${example_file_dir}/mrjti_example.txt \ --n_genes 100 \ --out_path ${example_file_dir}/mrjti_result.csv
the out put file like this: variable beta beta_CI_lower beta_CI_upper CI_significance expression -0.709540602 -0.901773199 -0.551193179 sig ldsc -0.062807264 -0.233198711 0.105520592 nonsig rs3902354 0 0 0 nonsig rs68104325 0 0 0 nonsig rs585362 0 0 0 nonsig rs17035665 0 0 0 nonsig
the variable column inclueded expression,ldsc and followed by 51 rsid. I do not understand how this output to determin the causal inference for a gene. I searched your result in you paper S8_MR-JTI: geneid genename Bonferroni_adjusted_CI_lower Bonferroni_adjusted_CI_upper Significance map_to_well-known_genes map_to_mouse_cholesterol_modules ENSG00000002834 LASP1 0.000 0.000 nonsig NA Lasp1 ENSG00000003056 M6PR 0.000 0.000 nonsig NA M6pr ENSG00000004142 POLDIP2 0.180 0.525 sig NA NA ENSG00000005981 ASB4 0.000 0.000 nonsig NA NA
Since there were not any gene id in output file,how could transfored the mrjti_result.csv to S8_MR-JTI result?
Problem2: The input file like this in example: rsid effect_allele_gwas ldscore eqtl_beta eqtl_se eqtl_p gwas_beta gwas_se gwas_p rs10449300 G 16.3387 0.122073958 0.114470828 0.287482148 -0.00067683 0.0025788 0.79296 rs10494095 G 9.46725 0.021090531 0.121112593 0.861926553 -0.0025397 0.0032695 0.43728 rs10857790 T 11.4308 0.056750248 0.137096339 0.679345557 -0.0059203 0.0030763 0.054298 I do have download the eQTL info in GTEx, but since the eqtl_beta, eqtl_se, eqtl_p were caculated for a target gene and I found that some SNPs are eQTL for more than one gene of interest in the GTEx data, so beta (slope) and standard_error (slope_se) differ among genes. So i am not sure how to merge this info to the file.
Problem3: You recomend SNPs need to be pruned before runing MR-JTI, does it mean i prune first before GWAS and then selected SNPs within the Bonferroni correction genes' window? Total number of genes tested (Bonferroni correction will be applied), how should i selected snps for this genes? by gene position window? The MR-JTI was run by once for all selected genes or for genes one by one?
I am very interested in this method and i am new for this, sorroy for so many easy problems. look forward for your help. Thanks.