INFO - 35 % of model's snps used in every tissue

[montenegrina]

Hello,

I am running this:

TISSUE_GTEx=(Adipose_Subcutaneous Adipose_Visceral_Omentum Adrenal_Gland Artery_Aorta Artery_Coronary Artery_Tibial Brain_Anterior_cingulate_cortex_BA24 Brain_Caudate_basal_ganglia Brain_Cerebellar_Hemisphere Brain_Cerebellum Brain_Cortex Brain_Frontal_Cortex_BA9 Brain_Hippocampus Brain_Hypothalamus Brain_Nucleus_accumbens_basal_ganglia Brain_Putamen_basal_ganglia Breast_Mammary_Tissue Cells_EBV-transformed_lymphocytes Cells_Transformed_fibroblasts Colon_Sigmoid Colon_Transverse Esophagus_Gastroesophageal_Junction Esophagus_Mucosa Esophagus_Muscularis Heart_Atrial_Appendage Heart_Left_Ventricle Liver Lung Muscle_Skeletal Nerve_Tibial Ovary Pancreas Pituitary Prostate Skin_Not_Sun_Exposed_Suprapubic Skin_Sun_Exposed_Lower_leg Small_Intestine_Terminal_Ileum Spleen Stomach Testis Thyroid Uterus Vagina Whole_Blood) for tissue in ${TISSUE_GTEx[@]} do python2 ./single_tissue_association_test.py \ --model_db_path sample_data/weight_db_GTEx/${tissue}.db \ --covariance sample_data/covariance_tissue/${tissue}.txt.gz \ --gwas_folder sample_data/GWAS \ --gwas_file_pattern METAANALYSIS1_SE_BETA.TBL \ --snp_column MarkerName \ --effect_allele_column Allele1 \ --non_effect_allele_column Allele2 \ --beta_column Effect \ --pvalue_column P-value \ --output_file sample_data/results_meta/${tissue}.csv done

So far I have processed 25 tissues and in every tissue there is always: INFO - 35 % of model's snps used

I am running this on META GWAS results: METAANALYSIS1_SE_BETA.TBL I got from running METAL software on 3 cohorts.

using for DEFAULTWEIGHT to specify the number of cases+controls for each cohort, so I have this:

DEFAULTWEIGHT 1211 DEFAULTWEIGHT 1363 DEFAULTWEIGHT 17012

To run your software I need BETA, to get BETA with METAL I can add this in my script: SCHEME STDERR

so my results in meta gwas file look like this:

MarkerName Allele1 Allele2 Effect StdErr P-value Direction rs11693919 t c -0.3396 0.0652 1.894e-07 —

In your manuscript you mention you used "inverse-variance weighted approach in METAL" can you please share with me METAL commands you used to do that?

I used your pre-calculated covaiance matrices you provided. Is that the issue?

Can you please advise, Ana

[montenegrina]

I also found out that "inverse-variance weighted approach in METAL" simply means commenting SCHEME STDERR in metal input script. Therefore I run the analysis again and there is 0.9999 correlation between p values I get with Z score and inverse-variance weighted method.

So I run UTMOST again across all tissues with the new META analysis done with inverse-variance weighted approach.

But again every tissue has at most:

INFO - 35 % of model's snps used

I also checked my alleles: A1 and A2 and they are the same as (REF and ALT) reported in HaploReg