Allow custom column names for predixcan import about focus

[bpprins]

The error was caused by gene label names being taken from the "gene" field in the weights table, which originally is the "gene_id" field in the PrediXcan GTEx modeling scripts output (gtex_v7_nested_cv_elnet.R) - which I used to fill with the probe names, and using the "gene_name" / "genename" field to populate with actual gene names according to GENCODE genes.

[quattro]

Hi @bpprins , thank you for your interest in focus! Did you resolve this? I can update the predixcan-import to use user-specified column names, but I'll have to inspect the predixcan scripts.

[bpprins]

Hello @quattro (Nicholas), yes I've resolved it with a workaround, thanks for asking. It seems you have a lot going on so the workaround I think is fine for now. If anyone else is using the same pipeline as I do, may be then it's worthwhile later on to be able to specify columns: I have used the PredictDB_Pipeline_GTEx_v7 pipeline to create models, which can be found on their github right over here :

https://github.com/hakyimlab/PredictDB_Pipeline_GTEx_v7/tree/master/model_training/scripts.

Specifically I have used, gtex_v7_nested_cv_elnet.R for model training, the output of which are then fed to make_dbs.R and then I used filter_dbs.R to create the dbs, which I then converted to FOCUS format.

Cheers,

Bram

[quattro]

Thanks for the update Bram. I'll re-open this to remind me for the next release. Really looking forward to seeing how the plasma protein weights turn out. Cheers, Nick