grelu.visualize#

Functions#

`_collect_preds_and_labels`(→ pandas.DataFrame)	Function to collect predictions and labels for a region of interest into a dataframe
`plot_distribution`(values[, title, method, figsize])	Given a 1-D sequence of values, plot a histogram or density plot of their distribution.
`plot_pred_distribution`(preds, labels[, tasks, bins, ...])	Plot the density of predictions and regression labels for a given task.
`plot_pred_scatter`(preds, labels[, tasks, bins, ...])	Plot a scatterplot of predictions and regression labels for a given task.
`plot_binary_preds`(preds, labels[, tasks, bins, figsize])	Plot a box plot of predictions for each classification label
`plot_calibration_curve`(probs, labels[, tasks, bins, ...])	Plots a calibration curve for a classification model
`add_highlights`(→ None)	Add highlights to a matplotlib axis
`plot_evolution`(df[, figsize])	Plot change in scores and predictions over multiple rounds of directed evolution
`plot_gc_match`(positives, negatives[, binwidth, ...])	Plot a histogram comparing GC content distribution in positive and negative regions.
`plot_attributions`(attrs[, start_pos, end_pos, ...])	Plot base-level importance scores across a sequence.
`plot_ISM`(ism_preds[, start_pos, end_pos, figsize, method])	Return in silico mutagenesis plot
`plot_tracks`(tracks[, start_pos, end_pos, titles, ...])	Plot genomic coverage tracks
`plot_attention_matrix`(attn[, start_pos, end_pos, ...])	Plot a bin x bin matrix of attentiomn weights derived

Module Contents#

grelu.visualize._collect_preds_and_labels(preds: numpy.ndarray | pandas.DataFrame, labels: numpy.ndarray | anndata.AnnData, tasks: List[int] | List[str] | None = None, bins: List[int] | None = None) → pandas.DataFrame[source]#: Function to collect predictions and labels for a region of interest into a dataframe

grelu.visualize.plot_distribution(values: List | numpy.ndarray | pandas.Series, title: str = 'metric', method: str = 'histogram', figsize: Tuple[int, int] = (4, 3), **kwargs)[source]#

Given a 1-D sequence of values, plot a histogram or density plot of their distribution.

Parameters:

values – 1-D sequence of numbers to plot
title – Plot title
method – Either “histogram” or “density”
figsize – Tuple containing (width, height)
**kwargs – Additional arguments to pass to geom_histogram (if method == “histogram”) or geom_density (if method == “density”).

Returns:

histogram or density plot

Plot the density of predictions and regression labels for a given task.

Parameters:

preds – Model predictions
labels – True labels
tasks – List of task names or indices. If None, all tasks will be used.
bins – List of relevant bins in the predictions and labels. If None, all bins will be used.
figsize – Tuple containing (width, height)
**kwargs – Additional arguments to pass to geom_density()

Returns:

Density plots

grelu.visualize.plot_pred_scatter(preds: numpy.ndarray | pandas.DataFrame, labels: numpy.ndarray | anndata.AnnData, tasks: List[int] | List[str] | None = None, bins: List[int] | None = None, density: bool = False, figsize: Tuple[int, int] = (4, 3), **kwargs)[source]#

Plot a scatterplot of predictions and regression labels for a given task.

Parameters:

preds – Model predictions
labels – True labels
tasks – List of task names or indices. If None, all tasks will be used.
bins – List of relevant bins in the predictions and labels. If None, all bins will be used.
density – If true, color the points by local density.
figsize – Tuple containing (width, height)
**kwargs – Additional arguments to pass to geom_point (if density = False) or geom_pointdensity (if density = True).

Returns:

Scatter plots

Plot a box plot of predictions for each classification label

Parameters:

preds – Model predictions
labels – True labels
tasks – List of task names or indices. If None, all tasks will be used.
bins – List of relevant bins in the predictions and labels. If None, all bins will be used.
figsize – Tuple containing (width, height)
**kwargs – Additional arguments to pass to geom_boxplot

Returns:

Box plots

grelu.visualize.plot_calibration_curve(probs: numpy.ndarray | pandas.DataFrame, labels: numpy.ndarray | anndata.AnnData, tasks: List[int] | List[str] | None = None, bins: List[int] | None = None, aggregate: bool = True, figsize: Tuple[int, int] = (4, 3), show_legend: bool = True)[source]#

Plots a calibration curve for a classification model

Parameters:

probs – Model predictions
labels – True classification labels
tasks – List of task names or indices. If None, all tasks will be used.
bins – List of relevant bins in the predictions and labels. If None, all bins will be used.
figsize – Tuple containing (width, height)
show_legend – If True, the legend is displayed. If False, no legend is displayed.

Returns:

Line plots showing the calibration between true and predicted probabilities for each task (if aggregate=False) or for all tasks combined (if aggregate=True)

grelu.visualize.add_highlights(ax, centers: int | List[int] | None = None, width: int | None = None, starts: int | List[int] | None = None, ends: int | List[int] | None = None, positions: int | List[int] | None = None, ymin: float = -10, ymax: float = 20, facecolor: str | None = 'yellow', alpha: float | None = 0.15, edgecolor: str | None = None) → None[source]#: Add highlights to a matplotlib axis

grelu.visualize.plot_evolution(df: pandas.DataFrame, figsize: Tuple[float, float] = (4, 3), **kwargs)[source]#

Plot change in scores and predictions over multiple rounds of directed evolution

Parameters:

df – Dataframe produced by grelu.design.evolve
figsize – Tuple containing (width, height)
**kwargs – Additional arguments to pass to geom_boxplot.

grelu.visualize.plot_gc_match(positives: pandas.DataFrame, negatives: pandas.DataFrame, binwidth: float = 0.1, genome: str = 'hg38', figsize: Tuple[int, int] = (4, 3), **kwargs)[source]#

Plot a histogram comparing GC content distribution in positive and negative regions.

Parameters:

positives – Genomic intervals
negatives – Genomic intervals
binwidth – Resolution at which to bin GC content
genome – Name of the genome
figsize – Tuple containing (width, height)
**kwargs – Additional arguments to pass to geom_bar

Returns: Bar plot

grelu.visualize.plot_attributions(attrs: numpy.ndarray, start_pos: int = 0, end_pos: int = -1, figsize: Tuple[int] = (20, 2), ticks: int = 10, highlight_centers: List[int] | None = None, highlight_width: int | List[int] = 5, highlight_positions: List[int] | None = None, ylim: Tuple[float, float] | None = None, facecolor: str | None = 'yellow', edgecolor: str | None = None, alpha: float | None = 0.15)[source]#

Plot base-level importance scores across a sequence.

Parameters:

attrs – A numpy array of shape (4, L)
start_pos – Start position along the sequence
end_pos – End position along the sequence.
figsize – Tuple containing (width, height)
ticks – Frequency of ticks on the x-axis
highlight_centers – List of positions where highlights are centered
highlight_width – Width of each highlighted region
highlight_positions – List of individual positions to highlight.
ylim – Axis limits for the y-axis
facecolor – Face color for highlight box
edgecolor – Edge color for highlight box
alpha – Opacity of highlight box

grelu.visualize.plot_ISM(ism_preds: pandas.DataFrame, start_pos: int | None = None, end_pos: int | None = None, figsize: Tuple[float, float] = (8, 1.5), method: str = 'heatmap', **kwargs)[source]#

Return in silico mutagenesis plot

Parameters:

ism_preds – ISM dataframe produced by grelu.model.interpret.ISM_predict
start_pos – Start position of region to plot
end_pos – End position of region to plot
figsize – Tuple containing (width, height)
method – ‘heatmap’ or ‘logo’
**kwargs – Additional arguments to be passed to sns.heatmap (in case type=’heatmap’)
'logo' (or plot_attributions (in case type =) –

Returns:

Heatmap or sequence logo for the specified region.

grelu.visualize.plot_tracks(tracks: numpy.ndarray, start_pos: int = 0, end_pos: int = None, titles: List[str] | None = None, figsize: Tuple[float, float] = (20, 1.5), highlight_intervals: pandas.DataFrame | None = None, facecolor: str | None = 'yellow', edgecolor: str | None = None, alpha: float | None = 0.15, annotations: Dict[str, pandas.DataFrame] = {})[source]#

Plot genomic coverage tracks

Parameters:

tracks – Numpy array of shape (T, L)
start_pos – Coordinate at which the tracks start
end_pos – Coordinate at which the tracks end
titles – List containing a title for each track
figsize – Tuple of (width, height)
highlight_intervals – A pandas dataframe containing genomic intervals to highlight
facecolor – Face color for highlight box
edgecolor – Edge color for highlight box
alpha – Opacity of highlight box
annotations – Dictionary of (key, value) pairs where the keys are strings and the values are pandas dataframes containing annotated genomic intervals

grelu.visualize.plot_attention_matrix(attn: numpy.ndarray, start_pos: int = 0, end_pos: int | None = None, highlight_intervals: pandas.DataFrame | None = None, figsize: Tuple[int, int] = (5, 4), **kwargs)[source]#

Plot a bin x bin matrix of attentiomn weights derived from transformer layers in a model.

Parameters:

attn – A square numpy array containing attention weights.
start_pos – The start coordinate of the genomic region
end_pos – The end coordinate of the genomic region
highlight_intervals – A pandas dataframe containing genomic intervals to highlight
figsize – A tuple containing (width, height)
**kwargs – Additional arguments to pass to sns.heatmap