Biostar Beta. Not for public use.
De Novo pathway generation
Entering edit mode
20 months ago
Dirk • 80

I'm very interested in looking into a "de novo" generation of gene/protein interaction pathways that are built strictly from gene expression (e.g. rnaseq) datasets. I've looked around and have found that most of the work is done on gene enrichment or pathway enrichment, but I am very curious if we could suss out what "pathways" exist from a purely data-driven standpoint. I have a dataset of ~120,000 expression profiles for ~30k genes--is there any way to extract gene interactions from there without relying on pathway commons/KEGG/etc?

Entering edit mode

I should note that my first idea is to create an adjacency matrix for each of the genes that use the weights of a neural network between the genes as the weights in the matrix and create an undirected acyclic graph from there, but wasn't sure if there was a better route.


Login before adding your answer.

Similar Posts
Loading Similar Posts
Powered by the version 2.1