search

Kevin-Haigis-Lab / speclet

A Bayesian hierarchical model to discover tissue-specific cancer driver genes and synthetic lethal interactions from CRISPR/Cas9 LoF screens.
GNU General Public License v3.0
0 stars 0 forks source link

Treat cancer genes as standard covariates in the single lineage model #174

Closed jhrcook closed 2 years ago

jhrcook commented 2 years ago

The cancer genes co-mutation variable required special parameterization with a 3D matrix and complicated indexing in the model of all the lineages. This is no longer necessary for the single-lineage model where the cancer genes can be treated as regular coefficients in the data frame. Make sure to still drop cancer genes that are never mutated in the cell lines, though.

jhrcook commented 2 years ago

Also need to remember to account for the cancer gene's mutation effect is already a variable in the model. This avoids non-identifiability between $m$ and $w$.

jhrcook commented 2 years ago

I don't think this would actually work because $w$ is not a standard variable. It has dimensions $[\text{gene} \times \text{cancer gene}]$, not just a variable per cancer gene.