Scvi Tools

Configuration File: remote_tools/scvi_tools.json Tool Type: Remote Tools Count: 2

This page contains all tools defined in the scvi_tools.json configuration file.

Available Tools

run_scvi_differential_expression (Type: RemoteTool)

Train scVI on a single-cell count matrix and run Bayesian differential expression between two gro…

run_scvi_differential_expression tool specification

Tool Information:

  • Name: run_scvi_differential_expression

  • Type: RemoteTool

  • Description: Train scVI on a single-cell count matrix and run Bayesian differential expression between two groups of cells, returning per-gene log-fold-changes and probabilities of differential expression (proba_de). Input is a server-accessible .h5ad of raw UMI counts.

Parameters:

  • adata_path (string) (required) Server-accessible path or URL to an .h5ad AnnData of RAW UMI counts.

  • groupby (string) (required) obs column defining the groups to compare (e.g. ‘cell_type’).

  • group1 (string) (required) Value in groupby for the foreground group.

  • group2 (string) (optional) Value in groupby for the reference group (optional; default = rest).

  • batch_key (string) (optional) obs column naming the batch/sample (optional).

  • max_epochs (integer) (optional) Training epochs (default: scVI heuristic).

Example Usage:

query = {
    "name": "run_scvi_differential_expression",
    "arguments": {
        "adata_path": "example_value",
        "groupby": "example_value",
        "group1": "example_value"
    }
}
result = tu.run(query)

run_scvi_integration (Type: RemoteTool)

Train scVI (single-cell Variational Inference; Lopez 2018 / scvi-tools, Gayoso 2022) on a single-…

run_scvi_integration tool specification

Tool Information:

  • Name: run_scvi_integration

  • Type: RemoteTool

  • Description: Train scVI (single-cell Variational Inference; Lopez 2018 / scvi-tools, Gayoso 2022) on a single-cell RNA-seq count matrix and return the batch-corrected low-dimensional latent representation of cells — the standard input to clustering/UMAP. Corrects batch effects across samples/technologies when a batch key is supplied. Input is a server-accessible .h5ad of raw UMI counts.

Parameters:

  • adata_path (string) (required) Server-accessible path or URL to an .h5ad AnnData of RAW UMI counts.

  • batch_key (string) (optional) obs column naming the batch/sample for integration (optional; empty = no batch correction).

  • n_latent (integer) (optional) Latent dimensionality (default 10).

  • max_epochs (integer) (optional) Training epochs (default: scVI heuristic min(round(20000/n_cells*400), 400)).

  • n_top_genes (integer) (optional) Subset to this many highly-variable genes before training (default 2000; 0 = all).

  • accelerator (string) (optional) ‘auto’ (default), ‘cpu’, or ‘gpu’.

Example Usage:

query = {
    "name": "run_scvi_integration",
    "arguments": {
        "adata_path": "example_value"
    }
}
result = tu.run(query)