pylimma.vooma

pylimma.vooma(y, design=None, block=None, correlation=None, predictor=None, span=None, legacy_span=False, plot=False, save_plot=False, *, out_layer='vooma_E', weights_layer='vooma_weights', key='vooma', layer=None)[source]

voom-like weights for non-count expression data.

Similar to voom but for continuous log-expression data (e.g., microarray). Estimates the mean-variance relationship and computes observation weights.

Parameters:

  • y (ndarray) – Expression matrix (log-scale), shape (n_genes, n_samples).

  • design (ndarray, optional) – Design matrix. If None, uses an intercept-only model.

  • block (ndarray, optional) – Factor indicating blocking structure.

  • correlation (float, optional) – Intra-block correlation (required if block is specified).

  • predictor (ndarray, optional) – Precision predictor, shape (n_genes,) or (n_genes, n_samples). When given, the variance trend is fitted against a linear combination of average log-expression and the row-mean predictor, and sample-specific weights are derived from the predictor.

  • span (float, optional) – LOWESS span. If None, chosen adaptively.

  • legacy_span (bool, default False) – If True, use the legacy adaptive-span rule (small_n=10, power=0.5); otherwise use small_n=50, power=1/3. Ignored if span is given.

  • save_plot (bool, default False) – If True, include trend data in output.

  • plot (bool)

  • out_layer (str)

  • weights_layer (str)

  • key (str)

  • layer (str | None)

Returns:

Endarray

Expression matrix (same as input y).

weightsndarray

Precision weights, shape (n_genes, n_samples).

designndarray

Design matrix.

spanfloat

LOWESS span used.

voom_xydict, optional

Trend data (only if save_plot=True).

voom_linedict, optional

LOWESS fit (only if save_plot=True).

Return type:

dict