search

nhejazi / methyvim

:package: :microscope: R/methyvim: Targeted, Robust, and Model-free Differential Methylation Analysis
https://code.nimahejazi.org/methyvim/
MIT License
3 stars 0 forks source link

screening procedure: tmle.npvi #2

Closed nhejazi closed 5 years ago

nhejazi commented 7 years ago

The non-parametric variable importance measure implemented in the tmle.npvi package could be used as a screening procedure for identifying sites to be tested more rigorously.

The general screening algorithm would works as follows:

  1. For each CpG site measured via a methylation assay, construct an input matrix for tmle.npvi using the unscaled outcome (Y), the methylation measurements of the given site (A), and any background covariates at the subject-level. The matrix would thus be of dimension = n x (2 + #(baseline covars)), where n denotes the sample size (in subjects).

  2. Using the input matrix described above, the tmle.npvi procedure could then be run with its default settings, using the "learning" flavor. This will essentially result in a GLM being fit to each CpG site.

  3. Using the resultant p-values, a subset of the original CpG sites could be selected to being more thoroughly assessed via either tmle or tmle.npvi, based on the type of exposure in question.

nhejazi commented 6 years ago

The package tmle.npvi is outdated on CRAN, relative to the GitHub version. Until the CRAN version of the package supports the most recent set of functions for manipulating the results of the NPVI TML estimation technique, a screening procedure based on this cannot be implemented.