search

Bishop-Laboratory / autoPLIER-analysis

Analysis for the autoPLIER project
2 stars 1 forks source link

LV interpretability #51

Open Julian-Borbeck opened 2 years ago

Julian-Borbeck commented 2 years ago

Count number of Trivial LVs Of the non Trivial LVs compute expectation of number of Pathways each LV represents ( p0 weighting, p1 weighting, p2 weighting)

millerh1 commented 2 years ago

This is low priority because you can just reduce the number of LVs in the model

millerh1 commented 2 years ago

The goal of this issue would be to create some function that can take some set of LVs and their pathway loadings and collapse them down to get the aggregate weighting of pathways as a way of interpreting.

If for example, LVs 1, 5, 10, 12, and 21 are significantly predictive of "Cardiac Muscle", then we want to collapse them to get the set of pathways that are most predict of "Cardiac Muscle".