Function normalize_dispersions

src/memory/processing/hvg/mod.rs:355–382

Normalize dispersions by subtracting expected values and dividing by standard deviation. This creates z-scores for dispersion values, allowing identification of genes with unusually high variability relative to genes with similar expression levels. ## Parameters `log_dispersions` - Log-transformed dispersion values `bin_indices` - Bin assignment for each gene `bin_means` - Expected dispersion fo

(
    log_dispersions: &[f64],
    bin_indices: &[usize],
    bin_means: &[f64],
    bin_stds: &[f64],
)

///
/// Higher values indicate genes that are more variable than expected.
fn normalize_dispersions(
    log_dispersions: &[f64],
    bin_indices: &[usize],
    bin_means: &[f64],
    bin_stds: &[f64],
) -> anyhow::Result<Vec<f64>> {
    let mut normalized_dispersions = vec![0.0; log_dispersions.len()];

    for (i, &disp) in log_dispersions.iter().enumerate() {
        let bin_idx = bin_indices[i];

        if bin_idx >= bin_means.len() {
            normalized_dispersions[i] = f64::NAN;
            continue;
        }

        let mean = bin_means[bin_idx];
        let std = bin_stds[bin_idx];

        if disp.is_nan() || mean.is_nan() || std.is_nan() || std == 0.0 {
            normalized_dispersions[i] = f64::NAN;
        } else {
            normalized_dispersions[i] = (disp - mean) / std;
        }
    }

    Ok(normalized_dispersions)
}

/// Select highly variable genes based on dispersion scores and expression filters.
///