Skip to content

README-params.md

Latest commit

 

History

History
71 lines (65 loc) · 5.56 KB

README-params.md

File metadata and controls

71 lines (65 loc) · 5.56 KB

Parameters

Grouping parameters

  • experiment_key_column: Column in cell metadata used to generate pseudobulk datasets and calculate cell type proportions (combined with term proportion_covariate_column)
  • anndata_cell_label: Column in cell metadata representing cell types to perform differential expression and gene set enrichment

Differential expression parameters

Parameters are applied within each cell type as denoted by anndata_cell_label

  • mean_cp10k_filter: Remove genes with mean counts per 10,000 (CP10K) expression <= mean_cp10k_filter
  • models: List of configurations for differential expression methods to run
    • filter_options: Options for the optional pre-filter.
      • filter: Value to set as minimum for filter.
      • modality: [cp10k|counts]
      • metric: Default is mean. TODO -- alternate metrics implemented?
      • by_comparison: TODO -- what is this?
    • method: String in the format package::resolution::model. Possible values for each:
      • package: "mast", "edger", "deseq"
      • resolution: "singlecell" or "pseudobulk"
      • model: Options:
        • For MAST, "bayesglm", "glmer" (needed for random effect models), or "glm"
        • For edgeR, "glmQLFit" or "glmLRT"
        • For DESeq2, "glmGamPoi" (recommended), "parametric", "local", or "mean"
    • formula: Formula to model the gene expression. Terms should be columns in cell metadata (Ex: "~ sex + age + disease_status"). The pipeline also supports the following operations:
    • variable_target: Term in formula to test (e.g., disease_status)
    • variable_continuous: Terms in formula that should be cast as continuous covariates
    • variable_discrete: Terms in formula that should be cast as discrete (i.e., categorical) covariates
    • variable_discrete_level: Reference information for discrete covariates, formatted as "cov_1::ref;;cov_2::ref"
    • pre_filter_genes: Logical (e.g., true or false) to apply mean_cp10k_filter before or after performing differential expression
    • proportion_covariate_column: Column in cell metadata to calculate the proportion of cells from each experiment (defined by experiment_key_column) representing each value. For instance, if same as anndata_cell_label, pipeline will calculate the proportion of each cell type for each experiment key.
    • include_proportion_covariates: Logical (e.g., true or false) to include proportions from proportion_covariate_column in formula
    • ruvseq: Logical (e.g., true or false) to run RUVSeq
    • ruvseq_n_empirical_genes: Number of empirical genes to use as input into RUVSeq. If value<1, we will take the proportion of total genes (value * # genes total). If value>1, we will use value as the number of genes
    • ruvseq_min_pvalue: Number representing minimum p-value threshold for empirical genes. Only genes with p-value > value will be used as empirical genes
    • ruvseq_k: Number of RUVSeq factors to adjust for
  • de_merge_config: Configuration for merge settings
    • ihw_correction: Configuration for IHW correction
      • covariates: Comma-separated list of covariates to include in IHW correction (e.g., "cell_label,disease_status")
      • alpha: See IHW documentation
  • de_plot_config: Parameters for plotting differential expression results
    • mean_expression_filter: List of mean expression thresholds to drop for plots for each group in anndata_cell_label. For example: if gene A expression is 0 counts in cluster 1 and 10 in cluster 2, it will be dropped from cluster 1 but not cluster 2.
  • goenrich_config:
    • go_terms: Ontology terms: MF (Molecular Function), CC (Cellular Component), BP (Biological Process). Multiple terms can be specified, separated by commas (e.g., 'BP,MF,CC').
    • clustering_method: Method to cluster terms. Options: "binary_cut", "louvain", "mclust"
  • gsea_config: Parameters for running gene set analyses
    • fgsea_parameters: List of alternate configurations for fgsea
      • sample_size: See fGSEA documentation
      • score_type: See fGSEA documentation
      • min_set_size: See fGSEA documentation
      • max_set_size: See fGSEA documentation
      • eps: See fGSEA documentation
      • database: Comma-separated list of databases to test for enrichments. Detailed descriptions of databases can be found here. Options:
        • c2.cgp: Chemical and genetic perturbations
        • c2.cp.biocarta: BioCarta
        • c2.cp.kegg: KEGG
        • c2.cp.reactome: Reactome
        • c2.cp: PID
        • c5.bp: GO biological process
        • c5.cc: GO cellular component
        • c5.mf: GO molecular function
        • c6.all: Oncogenic signatures
        • c7.all: Immunologic signatures
        • all: All gene sets (c2.cp.reactome, c2.cp.kegg, c5.bp, c5.cc, c5.mf)
    • gsea_summarize_parameters: Parameters to summarize GSEA data
      • distance_metric: Metric to calculate distance between terms. Options: "kappa", "jaccard", "dice", "overlap"
      • clustering_method: Method to cluster terms. Options: "binary_cut", "louvain", "mclust"