def hfss_plot_convergences_report(convergence_t: pd.core.frame.DataFrame,
convergence_f: pd.core.frame.DataFrame,
fig: mpl.figure.Figure = None,
_display=True):
"""Plot convergence frequency vs. pass number if fig is None.
Plot delta frequency and solved elements vs. pass number.
Plot delta frequency vs. solved elements.
Args:
convergence_t (pandas.core.frame.DataFrame): Convergence vs pass number of the eigenemode freqs.
convergence_f (pandas.core.frame.DataFrame): Convergence vs pass number of the eigenemode freqs.
fig (matplotlib.figure.Figure, optional): A mpl figure. Defaults to None.
_display (bool, optional): Display the plot? Defaults to True.
"""
if fig is None:
fig = plt.figure(figsize=(11, 3.))
# Grid spec and axes; height_ratios=[4, 1], wspace=0.5
gs = mpl.gridspec.GridSpec(1, 3, width_ratios=[1.2, 1.5, 1])
axs = [fig.add_subplot(gs[i]) for i in range(3)]
ax0t = axs[1].twinx()
plot_convergence_f_vspass(axs[0], convergence_f)
plot_convergence_max_df(axs[1], convergence_t.iloc[:, 1])
plot_convergence_solved_elem(ax0t, convergence_t.iloc[:, 0])
plot_convergence_maxdf_vs_sol(axs[2], convergence_t.iloc[:, 1],
convergence_t.iloc[:, 0])
fig.tight_layout(w_pad=0.1) # pad=0.0, w_pad=0.1, h_pad=1.0)
def style_figure(fig: matplotlib.figure.Figure, title: str, bottom: float = 0.2) -> None:
"""Stylize the supplied matplotlib Figure instance."""
fig.tight_layout()
if bottom is not None:
fig.subplots_adjust(bottom=bottom)
fig.suptitle(title)
fig.legend(ncol=10, handlelength=0.75, handletextpad=0.25, columnspacing=0.5, loc='lower left')
def fig_square(fig: mpl.figure.Figure) -> None:
"""Adjust figure width so that it is square, and tight layout.
Parameters
----------
fig : mpl.figure.Figure
`Figure` instance.
"""
fig.set_size_inches(mpl.figure.figaspect(1))
fig.tight_layout()
def apply(self, fig: matplotlib.figure.Figure) -> None:
# It shouldn't hurt to align the labels if there's only one.
fig.align_ylabels()
# Adjust the layout.
fig.tight_layout()
adjust_default_args = dict(
# Reduce spacing between subplots
hspace=0,
wspace=0,
# Reduce external spacing
left=0.10,
bottom=0.105,
right=0.98,
top=0.98,
)
adjust_default_args.update(self.edge_padding)
fig.subplots_adjust(**adjust_default_args)
def plot_convergences(self,
convergence_t: pd.DataFrame = None,
convergence_f: pd.DataFrame = None,
fig: mpl.figure.Figure = None,
_display: bool = True):
"""Creates 3 plots, useful to determin the convergence achieved by the renderer:
* convergence frequency vs. pass number if fig is None.
* delta frequency and solved elements vs. pass number.
* delta frequency vs. solved elements.
Args:
convergence_t (pd.DataFrame): Convergence vs pass number of the eigenemode freqs.
convergence_f (pd.DataFrame): Convergence vs pass number of the eigenemode freqs.
fig (matplotlib.figure.Figure, optional): A mpl figure. Defaults to None.
_display (bool, optional): Display the plot? Defaults to True.
"""
if convergence_t is None:
convergence_t = self.convergence_t
if convergence_f is None:
convergence_f = self.convergence_f
if fig is None:
fig = plt.figure(figsize=(11, 3.))
# Grid spec and axes; height_ratios=[4, 1], wspace=0.5
gspec = mpl.gridspec.GridSpec(1, 3, width_ratios=[1.2, 1.5, 1])
axs = [fig.add_subplot(gspec[i]) for i in range(3)]
ax0t = axs[1].twinx()
plot_convergence_f_vspass(axs[0], convergence_f)
plot_convergence_max_df(axs[1], convergence_t.iloc[:, 1])
plot_convergence_solved_elem(ax0t, convergence_t.iloc[:, 0])
plot_convergence_maxdf_vs_sol(axs[2], convergence_t.iloc[:, 1],
convergence_t.iloc[:, 0])
fig.tight_layout(w_pad=0.1) # pad=0.0, w_pad=0.1, h_pad=1.0)
def _trends_helper(
adata: AnnData,
models: Dict[str, Dict[str, Any]],
gene: str,
ln_key: str,
lineage_names: Optional[Sequence[str]] = None,
same_plot: bool = False,
sharey: Union[str, bool] = False,
show_ylabel: bool = True,
show_lineage: bool = True,
show_xticks_and_label: bool = True,
lineage_cmap=None,
abs_prob_cmap=cm.viridis,
gene_as_title: bool = False,
legend_loc: Optional[str] = "best",
fig: mpl.figure.Figure = None,
axes: Union[mpl.axes.Axes, Sequence[mpl.axes.Axes]] = None,
**kwargs,
) -> None:
"""
Plot an expression gene for some lineages.
Parameters
----------
%(adata)s
%(model)s
gene
Name of the gene in `adata.var_names``.
ln_key
Key in ``adata.obsm`` where to find the lineages.
lineage_names
Names of lineages to plot.
same_plot
Whether to plot all lineages in the same plot or separately.
sharey
Whether the y-axis is being shared.
show_ylabel
Whether to show y-label on the y-axis. Usually, only the first column will contain the y-label.
show_lineage
Whether to show the lineage as the title. Usually, only first row will contain the lineage names.
show_xticks_and_label
Whether to show x-ticks and x-label. Usually, only the last row will show this.
lineage_cmap
Colormap to use when coloring the the lineage. If `None`, use colors from ``adata.uns``.
abs_prob_cmap:
Colormap to use when coloring in the absorption probabilities, if they are being plotted.
gene_as_title
Whether to use the gene names as titles (with lineage names as well) or on the y-axis.
legend_loc
Location of the legend. If `None`, don't show any legend.
fig
Figure to use.
ax
Ax to use.
**kwargs
Keyword arguments for :meth:`cellrank.ul.models.BaseModel.plot`.
Returns
-------
%(just_plots)s
"""
n_lineages = len(lineage_names)
if same_plot:
axes = [axes] * len(lineage_names)
fig.tight_layout()
axes = np.ravel(axes)
percs = kwargs.pop("perc", None)
if percs is None or not isinstance(percs[0], (tuple, list)):
percs = [percs]
same_perc = False # we need to show colorbar always if percs differ
if len(percs) != n_lineages or n_lineages == 1:
if len(percs) != 1:
raise ValueError(
f"Percentile must be a collection of size `1` or `{n_lineages}`, got `{len(percs)}`."
)
same_perc = True
percs = percs * n_lineages
hide_cells = kwargs.pop("hide_cells", False)
show_cbar = kwargs.pop("show_cbar", True)
if same_plot:
lineage_colors = (
lineage_cmap.colors
if lineage_cmap is not None and hasattr(lineage_cmap, "colors")
else adata.uns.get(f"{_colors(ln_key)}", cm.Set1.colors)
)
else:
lineage_colors = (
"black" if not mcolors.is_color_like(lineage_cmap) else lineage_cmap
)
for i, (name, ax, perc) in enumerate(zip(lineage_names, axes, percs)):
if same_plot:
if gene_as_title:
title = gene
ylabel = "expression" if show_ylabel else None
else:
title = ""
ylabel = gene
else:
if gene_as_title:
title = None
ylabel = "expression" if i == 0 else None
else:
title = (
(name if name is not None else "no lineage") if show_lineage else ""
)
ylabel = gene if i == 0 else None
models[gene][name].plot(
ax=ax,
fig=fig,
perc=perc,
show_cbar=False,
title=title,
hide_cells=hide_cells or (same_plot and i == n_lineages - 1),
same_plot=same_plot,
lineage_color=lineage_colors[i]
if same_plot and name is not None
else lineage_colors,
abs_prob_cmap=abs_prob_cmap,
ylabel=ylabel,
**kwargs,
)
if sharey in ("row", "all", True) and i == 0:
plt.setp(ax.get_yticklabels(), visible=True)
if show_xticks_and_label:
plt.setp(ax.get_xticklabels(), visible=True)
else:
ax.set_xlabel(None)
if not same_plot and same_perc and show_cbar and not hide_cells:
vmin = np.min([models[gene][ln].w_all for ln in lineage_names])
vmax = np.max([models[gene][ln].w_all for ln in lineage_names])
norm = mcolors.Normalize(vmin=vmin, vmax=vmax)
for ax in axes:
[
c
for c in ax.get_children()
if isinstance(c, mpl.collections.PathCollection)
][0].set_norm(norm)
divider = make_axes_locatable(ax)
cax = divider.append_axes("right", size="2.5%", pad=0.1)
_ = mpl.colorbar.ColorbarBase(
cax, norm=norm, cmap=abs_prob_cmap, label="absorption probability"
)
if same_plot and lineage_names != [None] and legend_loc is not None:
ax.legend(lineage_names, loc=legend_loc)
def _trends_helper(
models: Dict[str, Dict[str, Any]],
gene: str,
transpose: bool = False,
lineage_names: Optional[Sequence[str]] = None,
same_plot: bool = False,
sharey: Union[str, bool] = False,
show_ylabel: bool = True,
show_lineage: Union[bool, np.ndarray] = True,
show_xticks_and_label: Union[bool, np.ndarray] = True,
lineage_cmap: Optional[Union[mpl.colors.ListedColormap, Sequence]] = None,
lineage_probability_color: Optional[str] = None,
abs_prob_cmap=cm.viridis,
gene_as_title: bool = False,
legend_loc: Optional[str] = "best",
fig: mpl.figure.Figure = None,
axes: Union[mpl.axes.Axes, Sequence[mpl.axes.Axes]] = None,
**kwargs,
) -> None:
"""
Plot an expression gene for some lineages.
Parameters
----------
%(adata)s
%(model)s
gene
Name of the gene in `adata.var_names``.
ln_key
Key in ``adata.obsm`` where to find the lineages.
lineage_names
Names of lineages to plot.
same_plot
Whether to plot all lineages in the same plot or separately.
sharey
Whether the y-axis is being shared.
show_ylabel
Whether to show y-label on the y-axis. Usually, only the first column will contain the y-label.
show_lineage
Whether to show the lineage as the title. Usually, only first row will contain the lineage names.
show_xticks_and_label
Whether to show x-ticks and x-label. Usually, only the last row will show this.
lineage_cmap
Colormap to use when coloring the the lineage. When ``transpose``, this corresponds to the color of genes.
lineage_probability_color
Actual color of 1 ``lineage``. Only used when ``same_plot=True`` and ``transpose=True`` and
``lineage_probability=True``.
abs_prob_cmap:
Colormap to use when coloring in the absorption probabilities, if they are being plotted.
gene_as_title
Whether to use the gene names as titles (with lineage names as well) or on the y-axis.
legend_loc
Location of the legend. If `None`, don't show any legend.
fig
Figure to use.
ax
Ax to use.
**kwargs
Keyword arguments for :meth:`cellrank.ul.models.BaseModel.plot`.
Returns
-------
%(just_plots)s
"""
n_lineages = len(lineage_names)
if same_plot:
axes = [axes] * len(lineage_names)
fig.tight_layout()
axes = np.ravel(axes)
percs = kwargs.pop("perc", None)
if percs is None or not isinstance(percs[0], (tuple, list)):
percs = [percs]
same_perc = False # we need to show colorbar always if percs differ
if len(percs) != n_lineages or n_lineages == 1:
if len(percs) != 1:
raise ValueError(
f"Percentile must be a collection of size `1` or `{n_lineages}`, got `{len(percs)}`."
)
same_perc = True
percs = percs * n_lineages
hide_cells = kwargs.pop("hide_cells", False)
show_cbar = kwargs.pop("cbar", True)
show_prob = kwargs.pop("lineage_probability", False)
if same_plot:
if not transpose:
lineage_colors = (lineage_cmap.colors if lineage_cmap is not None
and hasattr(lineage_cmap, "colors") else
lineage_cmap)
else:
# this should be fine w.r.t. to the missing genes, since they are in the same order AND
# we're also passing the failed models (this is important)
# these are actually gene colors, bu w/e
if lineage_cmap is not None:
lineage_colors = (
lineage_cmap.colors if hasattr(lineage_cmap, "colors") else
[c for _, c in zip(lineage_names, lineage_cmap)])
else:
lineage_colors = _create_categorical_colors(n_lineages)
else:
lineage_colors = (("black" if not mcolors.is_color_like(lineage_cmap)
else lineage_cmap), ) * n_lineages
if n_lineages > len(lineage_colors):
raise ValueError(
f"Expected at least `{n_lineages}` colors, found `{len(lineage_colors)}`."
)
lineage_color_mapper = {
ln: lineage_colors[i]
for i, ln in enumerate(lineage_names)
}
successful_models = {
ln: models[gene][ln]
for ln in lineage_names if models[gene][ln]
}
if show_prob and same_plot:
minns, maxxs = zip(*[
models[gene][n]._return_min_max(show_conf_int=kwargs.get(
"conf_int", False), ) for n in lineage_names
])
minn, maxx = min(minns), max(maxxs)
kwargs["loc"] = legend_loc
kwargs["scaler"] = lambda x: (x - minn) / (maxx - minn)
else:
kwargs["loc"] = None
if isinstance(show_xticks_and_label, bool):
show_xticks_and_label = [show_xticks_and_label] * len(lineage_names)
elif len(show_xticks_and_label) != len(lineage_names):
raise ValueError(
f"Expected `show_xticks_label` to be the same length as `lineage_names`, "
f"found `{len(show_xticks_and_label)}` != `{len(lineage_names)}`.")
if isinstance(show_lineage, bool):
show_lineage = [show_lineage] * len(lineage_names)
elif len(show_lineage) != len(lineage_names):
raise ValueError(
f"Expected `show_lineage` to be the same length as `lineage_names`, "
f"found `{len(show_lineage)}` != `{len(lineage_names)}`.")
last_ax = None
ylabel_shown = False
cells_shown = False
for i, (name, ax, perc) in enumerate(zip(lineage_names, axes, percs)):
model = models[gene][name]
if isinstance(model, FailedModel):
if not same_plot:
ax.remove()
continue
if same_plot:
if gene_as_title:
title = gene
ylabel = "expression" if show_ylabel else None
else:
title = ""
ylabel = gene
else:
if gene_as_title:
title = None
ylabel = "expression" if not ylabel_shown else None
else:
title = ((name if name is not None else "no lineage")
if show_lineage[i] else "")
ylabel = gene if not ylabel_shown else None
model.plot(
ax=ax,
fig=fig,
perc=perc,
cbar=False,
title=title,
hide_cells=cells_shown if not hide_cells else True,
same_plot=same_plot,
lineage_color=lineage_color_mapper[name],
lineage_probability_color=lineage_probability_color,
abs_prob_cmap=abs_prob_cmap,
lineage_probability=show_prob,
ylabel=ylabel,
**kwargs,
)
if sharey in ("row", "all", True) and not ylabel_shown:
plt.setp(ax.get_yticklabels(), visible=True)
if show_xticks_and_label[i]:
plt.setp(ax.get_xticklabels(), visible=True)
else:
ax.set_xlabel(None)
last_ax = ax
ylabel_shown = True
cells_shown = True
if not same_plot and same_perc and show_cbar and not hide_cells:
vmin = np.min([model.w_all for model in successful_models.values()])
vmax = np.max([model.w_all for model in successful_models.values()])
norm = mcolors.Normalize(vmin=vmin, vmax=vmax)
for ax in axes:
children = [
c for c in ax.get_children()
if isinstance(c, mpl.collections.PathCollection)
]
if len(children):
children[0].set_norm(norm)
divider = make_axes_locatable(last_ax)
cax = divider.append_axes("right", size="2%", pad=0.1)
_ = mpl.colorbar.ColorbarBase(
cax,
norm=norm,
cmap=abs_prob_cmap,
label="absorption probability",
ticks=np.linspace(norm.vmin, norm.vmax, 5),
)
if same_plot and lineage_names != [None] and legend_loc is not None:
handles = [
mpl.lines.Line2D([], [], color=lineage_color_mapper[ln], label=ln)
for ln in successful_models.keys()
]
last_ax.legend(handles=handles, loc=legend_loc)
