search

monarch-initiative / genophenocorr

Genotype Phenotype Correlation
https://monarch-initiative.github.io/genophenocorr/stable
MIT License
4 stars 1 forks source link

Multiple testing correction strategy #126

Open ielis opened 6 months ago

ielis commented 6 months ago

Peter's info:

I think we need a more general interface than just MC type, because we will also want to choose which HPO terms to test using flexible criteria. Probably something like

class MultipleCorrectionStrategy

    Attributes
    MC type (Bonferroni, ....)
    Min annot (int) - e.g., do not test if just one patient is annotated to a term
    hierarchy strategy (probably an enum) – what should we do with testing terms that are parent/child of each other? There are some strategies that we should definitely use. For instance, if the parent has the same annots as the child, just test the child.

There are more sophisticated ways of doing this, and I spoke with a statistician who might want to help us with this, but it would be good to implement the above and then send it to Mike Love for possible collaboration.

ielis commented 4 months ago

We need to prepare a notebook that presents the nuts and bolts of the analysis.

We need to extract the code that runs the G/P analysis into a separate class and present it in the notebook.

pnrobinson commented 1 month ago

See https://github.com/monarch-initiative/genophenocorr/pull/160