Open rbutleriii opened 1 month ago
Hi @rbutleriii,
Apologies for the delayed response. You can infer pathway activity by treating pathways as sources and each edge of the network as a node with a specific mode as the target. Alternatively, you could take an iterative approach: start by computing activities from the "leaf" sources, and then iteratively calculate activities until you reach the "root." We’ve implemented a similar approach to this in a recent preprint: https://doi.org/10.1101/2024.07.15.603538 Hope this is helpful!
Hi,
Wondering if it it is possible to incorporate more complex prior information into the mlm or one of the other available models. I have gene data that is stored in a set of igraphs for each pathway I am analyzing, with edge values connecting genes and node values that represent scores for the genes themselves.
It doesn't seem like the network data.frame can have columns for
Pathway
,source_gene
,target_gene
,mor
. Although, a workaround would be if the framework could accept multiple frames of prior information. One network would be the node scores, and the other would be a simplified neighbor score for each node, similar to NGSEA generating a mean of the absolute value of the gene-based scores of its network neighbors, or alternately some other metric of network centrality.