An updated command line interface with functionality
Compute scores: scdrs compute-score. This perform the same function as python compute_score.py
Downstream analyses: scdrs perform-downstream, which perform the same function as python compute_downstream.py
Munge .gs file: scdrs munge-gs, which takes a gene set score file such as MAGMA association z-scores (or p-values) and format to the scdrs .gs format
The downstream analyses now only handle either one of the (1) group analysis (2) correlation analysis (3) gene analysis rather than performing all at once as in python compute_downstream.py. I think this way is cleaner.
The existing files are not changed to ensure the compatibility.
Several utility functions are moved from cli script to scdrs.util module as these can be also reused in python API.
scdrs.util.convert_species_name: taken from python compute_score.py
scdrs.util.load_h5ad: Load h5ad file and optionally filter out cells and perform normalization
scdrs.util.load_drs_score: Load drs scores, could be multiple score files, depend on the score_path parameter pattern. Either it is path to a single score file, such as /path/to/trait.full_score.gz or a file pattern for multiple score files, such as /path/to/@.full_score.gz
scdrs.util.downstream_group_analysis: Perform the group-level downstream analysis for scDRS results
scdrs.util.downstream_corr_analysis: Perform the correlation between cell-level variables with scDRS scores
scdrs.util.downstream_gene_analysis: Perform the correlation between gene-level variables with scDRS scores
Discussion points
Whether to move the downstream code to scdrs.method
An updated command line interface with functionality
scdrs compute-score. This perform the same function as python compute_score.pyscdrs perform-downstream, which perform the same function as python compute_downstream.pyscdrs munge-gs, which takes a gene set score file such as MAGMA association z-scores (or p-values) and format to the scdrs .gs formatThe downstream analyses now only handle either one of the (1) group analysis (2) correlation analysis (3) gene analysis rather than performing all at once as in
python compute_downstream.py. I think this way is cleaner.The existing files are not changed to ensure the compatibility.
Several utility functions are moved from cli script to
scdrs.utilmodule as these can be also reused in python API.scdrs.util.convert_species_name: taken from python compute_score.pyscdrs.util.load_h5ad: Load h5ad file and optionally filter out cells and perform normalizationscdrs.util.load_drs_score: Load drs scores, could be multiple score files, depend on thescore_pathparameter pattern. Either it is path to a single score file, such as /path/to/trait.full_score.gz or a file pattern for multiple score files, such as /path/to/@.full_score.gzscdrs.util.downstream_group_analysis: Perform the group-level downstream analysis for scDRS resultsscdrs.util.downstream_corr_analysis: Perform the correlation between cell-level variables with scDRS scoresscdrs.util.downstream_gene_analysis: Perform the correlation between gene-level variables with scDRS scoresDiscussion points