Command line utility
A script src/locaTE_cmd.jl that can be run from command line is included, which expects inputs to be passed as .npy or .csv files.
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
locaTE-cmd: utility for running locaTE workflow from the command-line.
Input matrices can be supplied as .npy (binary) or .csv (text).
positional arguments:
X Path to counts matrix X
X_rep Path to dimensionality-reduced representation of
X. From this, the kNN graph will be constructed.
P Path to transition matrix P encoding dynamics.
R Path to kernel matrix R encoding neighbourhood
information.
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 lambda1 (λ1), strength of Laplacian
regularization. (type: Float64, default: 5.0)
--lambda2 LAMBDA2 lambda2 (λ2), strength of Lasso regularization.
(type: Float64, default: 0.01)
--outdir OUTDIR Output directory (default: "./")
--suffix SUFFIX Suffix to append to output files. (default: "")
--cutoff CUTOFF Cutoff below which expression values will be set
to zero. Can be helpful for expression value
binning in some datasets with artifactually small
counts. (type: Float64, default: 0.0)
--gpu GPU acceleration, recommended for large datasets
when available.
--maxiter MAXITER Maximum iterations for denoising regression.
(type: Int64, default: 1000)
-h, --help show this help message and exit