Disentangled methylation dimension reduction

[jjc2718]

PCA generates a dense, ~Gaussian feature matrix, which may not play nicely with elastic net. We want to see if we can do better with a less "entangled" or sparser dimension reduction method.

VAEs have been used successfully for methylation data:

• https://www.biorxiv.org/content/10.1101/433763v5.full.pdf
• https://pdfs.semanticscholar.org/64a5/c234078f2facb3916a475c4af63a2ef90b70.pdf
• https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-020-3443-8

Maybe this would work better?

[jjc2718]

Could use this package: https://github.com/Christensen-Lab-Dartmouth/MethylNet