Question Regarding Estimating Ligand-Target Matrix Using Custom GR Network

[peipp410]

Hi there,

I have used my own method to estimate the regulatory strength of a gene regulatory network, and the results are stored in an n_
tf × n_hvg matrix (gr_network). I would like to know if it's possible to download the existing ligand_tf_matrix_nsga2r_final.rds and then multiply it by my own gr_network matrix (by selecting matching gene features) to obtain the final ligand_target_matrix.

Is this approach valid? Thank you for your help!

[Eisuan]

Dear user,
I think it depends on your final aim.

If your purpose is to perform ligand-target gene prioritization, I recommend integrating your GRN information as an additional source into NicheNet's GRN and then recomputing the ligand-target gene matrix. You can find the code here.
This process requires you to convert your network into an unweighted version.

You can otherwise manually perform the matrix multiplication between the PPR score and GRN matrices (the former is the one you linked), as described in the original paper. You will need to make some adjustments because the PPR score matrix does not describe ligand-to-TF relationships only. However, these manipulations (outside the framework I suggested) can cause profound and unpredictable changes in the model's performance.

Bests,
Daniele

[peipp410]

Thank you so much for your response! I really appreciate your help. I plan to try the first method you suggested and see how it works. Your advice has been incredibly helpful!