scVI Hyperparameter Metrics Pipeline

This pipeline performs a grid search to optimize hyperparameters for scVI. The grid is constructed by combining elements from predefined lists of model input parameters. After running scVI with each set of parameters, integration metrics are calculated using scib-metrics.

Repository Contents

• main.nf – The Nextflow pipeline that orchestrates the entire workflow.
• nextflow.config – Configuration file that:
  • Submits processes to LSF on Sanger's HPC.
  • Ensures the correct environment is set via a Singularity container (absolute path).
  • Defines global default parameters.
• examples/inputs.py – Example .py file specifying the path to anndata and configurations for scVI and scib-metrics.
• examples/run_scVI_metrics.sh – Example script to execute the pipeline.
• Dockerfile – Defines a Docker image with scVI and scib-metrics version 0.2.2.

Pipeline Arguments

• --input_file – Path to a .py file specifying:
  • The path to anndata.
  • Configuration details for scVI and scib-metrics.

Optional parameters:

• --help — Display this help message
• --umap — Set this flag to calculate umap