search

xinhe-lab / ctwas

package for the causal TWAS project
https://xinhe-lab.github.io/ctwas/
MIT License
31 stars 12 forks source link

excluding genes with poor performance #33

Open Takahiro0113 opened 1 month ago

Takahiro0113 commented 1 month ago

Thank you for creating a nice software! I'm wondering if it makes sense to exclude genes with poor prediction accuracy from prediction model as it could be noisy information. Do you have any idea?

kevinlkx commented 1 month ago

Thanks for your question. ctwas assumes the genes included in the prediction model already have good prediction accuracy. You should only include genes with good prediction accuracy in the model, e.g. FUSION has a heritability cutoff when training the model.