Function describe_var

src/memory/statistics/qc.rs:240–288

Calculate gene-level (variable) quality control metrics. Computes QC statistics for each gene including detection rates across cells, mean expression levels, and dropout percentages. ## Parameters `adata` - AnnData object containing the dataset `x` - Expression matrix to analyze `expr_type` - Label for expression type (e.g., "counts", "UMI") `_var_type` - Label for variable type (unused, kept fo

(
    adata: &IMAnnData,
    x: &anndata_memory::IMArrayElement,
    expr_type: &str,
    _var_type: &str,
    log1p: bool,
)

/// - Assess gene expression distributions
/// - Quality control before feature selection
fn describe_var(
    adata: &IMAnnData,
    x: &anndata_memory::IMArrayElement,
    expr_type: &str,
    _var_type: &str,
    log1p: bool,
) -> anyhow::Result<DataFrame> {
    let n_obs = adata.n_obs();

    let n_cells_by_counts: Vec<u32> = x.nonzero_whole(&Direction::COLUMN)?;
    let total_counts: Vec<f64> = x.sum_whole(&Direction::COLUMN)?;

    let mean_counts: Vec<f64> = total_counts
        .iter()
        .map(|&total| total / n_obs as f64)
        .collect();

    let mut columns = vec![];

    let col_name = format!("n_cells_by_{}", expr_type);
    columns.push(Column::new(col_name.into(), n_cells_by_counts.clone()));

    let col_name = format!("mean_{}", expr_type);
    columns.push(Column::new(col_name.into(), mean_counts.clone()));

    if log1p {
        let log_values: Vec<f64> = mean_counts.iter().map(|&x| (x + 1.0).ln()).collect();
        let col_name = format!("log1p_mean_{}", expr_type);
        columns.push(Column::new(col_name.into(), log_values));
    }

    let pct_dropout: Vec<f64> = n_cells_by_counts
        .iter()
        .map(|&n| (1.0 - n as f64 / n_obs as f64) * 100.0)
        .collect();
    let col_name = format!("pct_dropout_by_{}", expr_type);
    columns.push(Column::new(col_name.into(), pct_dropout));

    let col_name = format!("total_{}", expr_type);
    columns.push(Column::new(col_name.into(), total_counts.clone()));

    if log1p {
        let log_values: Vec<f64> = total_counts.iter().map(|&x| (x + 1.0).ln()).collect();
        let col_name = format!("log1p_total_{}", expr_type);
        columns.push(Column::new(col_name.into(), log_values));
    }

    DataFrame::new(columns).map_err(Into::into)
}

/// Calculate comprehensive quality control metrics for single-cell data.
///