Cell-specific network inference using local Transfer Entropy.
To get started, please see the documentation.
Code for reproducing the manuscript figures can be found here.
Command-line script is available in src/locaTE_cmd.jl, which can be called with arguments
usage: locaTE_cmd.jl [--tau TAU] [--k_lap K_LAP] [--lambda1 LAMBDA1]
[--lambda2 LAMBDA2] [--outdir OUTDIR]
[--suffix SUFFIX] [--cutoff CUTOFF] [--gpu]
[--maxiter MAXITER] [-h] [X] [X_rep] [P] [R]
positional arguments:
X path to counts matrix, X
X_rep path to dim-reduced representation of X
P path to transition matrix
R path to kernel matrix
optional arguments:
--tau TAU power for transition matrix (type: Int64,
default: 1)
--k_lap K_LAP number of neighbours for Laplacian (type: Int64,
default: 15)
--lambda1 LAMBDA1 (type: Float64, default: 5.0)
--lambda2 LAMBDA2 (type: Float64, default: 0.01)
--outdir OUTDIR (default: "./")
--suffix SUFFIX (default: "")
--cutoff CUTOFF (type: Float64, default: 0.0)
--gpu
--maxiter MAXITER (type: Int64, default: 1000)
-h, --help show this help message and exit
Example:
JULIA_NUM_THREADS=32 julia locaTE_cmd.jl --tau 1 --lambda1 25 --lambda2 0.001 --outdir locaTE_output/ --cutoff 0.3 X.npy X_pca.npy P_velo_dot.npy R.npy
For further examples, consult the examples/ directory.
See the preprint.
Zhang, S.Y. and Stumpf, M.P., 2023. Dynamical information enables inference of gene regulation at single-cell scale. bioRxiv, pp.2023-01.