PrediXcan issue with HCP data about metaxcan

[Heroico]

Hi Amy,

If a gene was imputed as 0 for all subjects, then it is likely that PrediXcan is not able to match the variants in your GWAS to the variants in the models. This can be caused by several issues.

Feel free to post here the full command line you are issuing to run PrediXcan.py.

Also, the --prediction_summary_output argument generates an additional file that might be useful.
It contains, for each gene in the database, how many variants are in its model, and how many of these variants are also available in the GWAS.

You can also implement a sanity check.
The prediction model files are sqlite databases.
You can access them programmatically in tools like R or python, and check if the variants in the model coincide with the variants in your genotypes.
For example, assuming you assemble a table with your variants like so (in a text file):

id      effect_allele non_effect_allele
rsid1   C             G
rsid2   A             T
....

Then you can run in R

library(RSQLite)
library(tidyverse)
db <- dbConnect(RSQLite::SQLite(), "my_model.db")
weights <- dbReadTable(db, "weights")
dbDisconnect()

# then load variants and alleles
variants <- read_tsv("my_variants.txt.gz")
# aligned variants
weights %>% inner_join(variants, by=c("rsid"="id", "ref_allele"="non_effect_allele", "eff_allele"="effect_allele")) %>% nrow
# swapped variants
weights %>% inner_join(variants, by=c("rsid"="id", "ref_allele"="effect_allele", "eff_allele"="non_effect_allele")) %>% nrow

[amymiles]

Thanks very much for the prompt response. I ran PrediXcan as follows:

python PrediXcan.py
--predict
--dosages data/dosages
--dosages_prefix chr
--samples samples.fam
--weights data/weights/gtex_v7_Brain_Cortex_imputed_europeans_tw_0.5_signif.db
--output_prefix results/${DATE}_cortex

When I included the '--prediction_summary_output' flag, I received the following error message: PrediXcan.py: error: unrecognized arguments: --prediction_summary_output

Please let me know if there are errors in the code that I have missed. If not, I will look into running the sanity check, as described. As a side note, I initially ran the above command with python 2.7.8, but I have also tried with python 3.8.1.

Thanks again for all of your help,
Amy

[Heroico]

Hi Amy, this looks like a command meant for the deprecated implementation:
https://github.com/hakyimlab/PrediXcan

You can find a tutorial for the new implementation in this repository (https://github.com/hakyimlab/MetaXcan) following this link.

Best,

Alvaro

[Heroico]

Closing due to inactivity