search

PathwayAndDataAnalysis / causalpath

A project for exploring differentially active signaling paths related to proteomics datasets
GNU Lesser General Public License v3.0
37 stars 10 forks source link

Question on the use of causalpath in transcriptomic data #25

Open LiuCanidk opened 2 months ago

LiuCanidk commented 2 months ago

Hi, thanks for developing this nice tool!

I wonder if this tool can be applied into transcriptomic data although the original design is for proteomic data. I think it is all about abundance data and causal correlation inferrence so it may be applied into RNA gene expression data. But I am not sure about this extension of causalpath, so I would like to submit this new issue. And if so, any additional key points should I take consideration into trnascriptomic analysis with causalpath?

Any disucussion would be greatly appreciated!

ozgunbabur commented 2 months ago

Hi LiuCanidk,

Yes, CausalPath can analyze just the RNA expression data, even though this is not its primary use case. It requires setting up the parameters carefully. Here are important points:

  1. The RNAseq data needs to be formatted for CausalPath. Make sure there is an "ID" column, a "Symbols" column, and a "Feature" column. You can use the gene symbol in both the ID and the Symbols columns. Let the Feature column display the letter "R" for all the rows, indicating these are for RNA expression.

  2. Use the parameter "data-type-for-expressional-targets = rna" in the parameters.txt file.

3 (Optional). Use the parameter "use-expression-for-activity-evidence = true". CausalPath uses proteomic changes as evidence for activity change by default. It does not do that for RNA expression by default because a change in RNA may not often affect the protein activity. But since RNA data is all we have here, we can just try it this way for exploration. But the results will be noisy for sure. If this option is not used, make sure network significance calculation is turned on ("calculate-network-significance = true"), otherwise, results will be always empty. In the second case, CausalPath will infer activated and inhibited transcription factors through data label randomization and connect them to their causal downstream RNA changes on the result network. This is a less noisy and better option if any differential TF is inferred.

I think it is best if we work on an example. You can share your data (partial data is fine), and we can craft a parameters.txt file that will work.

Ozgun

LiuCanidk commented 2 months ago

Thanks a lot! I will try causalpath on my RNA data and share the result soon.