# code adapted from https://github.com/lmcinnes/umap/blob/7e051d8f3c4adca90ca81eb45f6a9d1372c076cf/umap/plot.py
import warnings
import numpy as np
import pandas as pd
from pandas.api.types import is_categorical_dtype
from scipy.sparse import issparse
from numbers import Number
import matplotlib.colors
import matplotlib.cm
from ..configuration import _themes, set_figure_params, reset_rcParams
from .utils import (
despline,
despline_all,
deaxis_all,
set_spine_linewidth,
scatter_with_colorbar,
scatter_with_legend,
_select_font_color,
arrowed_spines,
is_gene_name,
is_cell_anno_column,
is_list_of_lists,
is_layer_keys,
_matplotlib_points,
_datashade_points,
save_fig,
)
from ..tools.utils import (
update_dict,
get_mapper,
log1p_,
flatten,
)
from ..tools.moments import calc_1nd_moment
from ..docrep import DocstringProcessor
docstrings = DocstringProcessor()
[docs]@docstrings.get_sectionsf("scatters")
def scatters(
adata,
basis="umap",
x=0,
y=1,
color='ntr',
layer="X",
highlights=None,
labels=None,
values=None,
theme=None,
cmap=None,
color_key=None,
color_key_cmap=None,
background=None,
ncols=4,
pointsize=None,
figsize=(6, 4),
show_legend="on data",
use_smoothed=True,
aggregate=None,
show_arrowed_spines=False,
ax=None,
sort='raw',
save_show_or_return="show",
save_kwargs={},
return_all=False,
add_gamma_fit=False,
frontier=False,
contour=False,
ccmap=None,
calpha=2.3,
sym_c=False,
smooth=False,
dpi=100,
inset_dict={},
**kwargs
):
"""Plot an embedding as points. Currently this only works
for 2D embeddings. While there are many optional parameters
to further control and tailor the plotting, you need only
pass in the trained/fit umap model to get results. This plot
utility will attempt to do the hard work of avoiding
overplotting issues, and make it easy to automatically
colour points by a categorical labelling or numeric values.
This method is intended to be used within a Jupyter
notebook with ``%matplotlib inline``.
Parameters
----------
adata: :class:`~anndata.AnnData`
an Annodata object
basis: `str`
The reduced dimension.
x: `int` (default: `0`)
The column index of the low dimensional embedding for the x-axis.
y: `int` (default: `1`)
The column index of the low dimensional embedding for the y-axis.
color: `string` (default: `ntr`)
Any column names or gene expression, etc. that will be used for coloring cells.
layer: `str` (default: `X`)
The layer of data to use for the scatter plot.
highlights: `list` (default: None)
Which color group will be highlighted. if highligts is a list of lists - each list is relate to each color element.
labels: array, shape (n_samples,) (optional, default None)
An array of labels (assumed integer or categorical),
one for each data sample.
This will be used for coloring the points in
the plot according to their label. Note that
this option is mutually exclusive to the ``values``
option.
values: array, shape (n_samples,) (optional, default None)
An array of values (assumed float or continuous),
one for each sample.
This will be used for coloring the points in
the plot according to a colorscale associated
to the total range of values. Note that this
option is mutually exclusive to the ``labels``
option.
theme: string (optional, default None)
A color theme to use for plotting. A small set of
predefined themes are provided which have relatively
good aesthetics. Available themes are:
* 'blue'
* 'red'
* 'green'
* 'inferno'
* 'fire'
* 'viridis'
* 'darkblue'
* 'darkred'
* 'darkgreen'
cmap: string (optional, default 'Blues')
The name of a matplotlib colormap to use for coloring
or shading points. If no labels or values are passed
this will be used for shading points according to
density (largely only of relevance for very large
datasets). If values are passed this will be used for
shading according the value. Note that if theme
is passed then this value will be overridden by the
corresponding option of the theme.
color_key: dict or array, shape (n_categories) (optional, default None)
A way to assign colors to categoricals. This can either be
an explicit dict mapping labels to colors (as strings of form
'#RRGGBB'), or an array like object providing one color for
each distinct category being provided in ``labels``. Either
way this mapping will be used to color points according to
the label. Note that if theme
is passed then this value will be overridden by the
corresponding option of the theme.
color_key_cmap: string (optional, default 'Spectral')
The name of a matplotlib colormap to use for categorical coloring.
If an explicit ``color_key`` is not given a color mapping for
categories can be generated from the label list and selecting
a matching list of colors from the given colormap. Note
that if theme
is passed then this value will be overridden by the
corresponding option of the theme.
background: string or None (optional, default 'None`)
The color of the background. Usually this will be either
'white' or 'black', but any color name will work. Ideally
one wants to match this appropriately to the colors being
used for points etc. This is one of the things that themes
handle for you. Note that if theme
is passed then this value will be overridden by the
corresponding option of the theme.
ncols: int (optional, default `4`)
Number of columns for the figure.
pointsize: `None` or `float` (default: None)
The scale of the point size. Actual point cell size is calculated as `500.0 / np.sqrt(adata.shape[0]) *
pointsize`
figsize: `None` or `[float, float]` (default: None)
The width and height of a figure.
show_legend: bool (optional, default True)
Whether to display a legend of the labels
use_smoothed: bool (optional, default True)
Whether to use smoothed values (i.e. M_s / M_u instead of spliced / unspliced, etc.).
aggregate: `str` or `None` (default: `None`)
The column in adata.obs that will be used to aggregate data points.
show_arrowed_spines: bool (optional, default False)
Whether to show a pair of arrowed spines representing the basis of the scatter is currently using.
ax: `matplotlib.Axis` (optional, default `None`)
The matplotlib axes object where new plots will be added to. Only applicable to drawing a single component.
sort: `str` (optional, default `raw`)
The method to reorder data so that high values points will be on top of background points. Can be one of
{'raw', 'abs', 'neg'}, i.e. sorted by raw data, sort by absolute values or sort by negative values.
save_show_or_return: `str` {'save', 'show', 'return'} (default: `show`)
Whether to save, show or return the figure.
save_kwargs: `dict` (default: `{}`)
A dictionary that will passed to the save_fig function. By default it is an empty dictionary and the save_fig function
will use the {"path": None, "prefix": 'scatter', "dpi": None, "ext": 'pdf', "transparent": True, "close":
True, "verbose": True} as its parameters. Otherwise you can provide a dictionary that properly modify those keys
according to your needs.
return_all: `bool` (default: `False`)
Whether to return all the scatter related variables. Default is False.
add_gamma_fit: `bool` (default: `False`)
Whether to add the line of the gamma fitting. This will automatically turn on if `basis` points to gene names
and those genes have went through gamma fitting.
frontier: `bool` (default: `False`)
Whether to add the frontier. Scatter plots can be enhanced by using transparency (alpha) in order to show area
of high density and multiple scatter plots can be used to delineate a frontier. See matplotlib tips & tricks
cheatsheet (https://github.com/matplotlib/cheatsheets). Originally inspired by figures from scEU-seq paper:
https://science.sciencemag.org/content/367/6482/1151. If `contour` is set to be True, `frontier` will be
ignored as `contour` also add an outlier for data points.
contour: `bool` (default: `False`)
Whether to add an countor on top of scatter plots. We use `tricontourf` to plot contour for non-gridded data.
The shapely package was used to create a polygon of the concave hull of the scatters. With the polygon we
then check if the mean of the triangulated points are within the polygon and use this as our condition to
form the mask to create the contour. We also add the polygon shape as a frontier of the data point (similar
to when setting `frontier = True`). When the color of the data points is continuous, we will use the same cmap
as for the scatter points by default, when color is categorical, no contour will be drawn but just the
polygon. cmap can be set with `ccmap` argument. See below.
ccmap: `str` or `None` (default: `None`)
The name of a matplotlib colormap to use for coloring or shading points the contour. See above.
calpha: `float` (default: `2.3`)
alpha value for identifying the alpha hull to influence the gooeyness of the border. Smaller numbers don't
fall inward as much as larger numbers. Too large, and you lose everything!
sym_c: `bool` (default: `False`)
Whether do you want to make the limits of continuous color to be symmetric, normally this should be used for
plotting velocity, jacobian, curl, divergence or other types of data with both positive or negative values.
smooth: `bool` or `int` (default: `False`)
Whether do you want to further smooth data and how much smoothing do you want. If it is `False`, no smoothing
will be applied. If `True`, smoothing based on one step diffusion of connectivity matrix (`.uns['moment_cnn']
will be applied. If a number larger than 1, smoothing will based on `smooth` steps of diffusion.
dpi: `float`, (default: 100.0)
The resolution of the figure in dots-per-inch. Dots per inches (dpi) determines how many pixels the figure
comprises. dpi is different from ppi or points per inches. Note that most elements like lines, markers,
texts have a size given in points so you can convert the points to inches. Matplotlib figures use Points per
inch (ppi) of 72. A line with thickness 1 point will be 1./72. inch wide. A text with fontsize 12 points will
be 12./72. inch heigh. Of course if you change the figure size in inches, points will not change, so a larger
figure in inches still has the same size of the elements.Changing the figure size is thus like taking a piece
of paper of a different size. Doing so, would of course not change the width of the line drawn with the same
pen. On the other hand, changing the dpi scales those elements. At 72 dpi, a line of 1 point size is one
pixel strong. At 144 dpi, this line is 2 pixels strong. A larger dpi will therefore act like a magnifying
glass. All elements are scaled by the magnifying power of the lens. see more details at answer 2 by
@ImportanceOfBeingErnest: https://stackoverflow.com/questions/47633546/relationship-between-dpi-and-figure-size
inset_dict: `dict` (default: {})
A dictionary of parameters in inset_ax. Example, something like {"width": "5%", "height": "50%", "loc":
'lower left', "bbox_to_anchor": (0.85, 0.90, 0.145, 0.145), "bbox_transform": ax.transAxes, "borderpad": 0}
See more details at https://matplotlib.org/api/_as_gen/mpl_toolkits.axes_grid1.inset_locator.inset_axes.html
or https://stackoverflow.com/questions/39803385/what-does-a-4-element-tuple-argument-for-bbox-to-anchor-mean-in-matplotlib
kwargs:
Additional arguments passed to plt.scatters.
Returns
-------
result: matplotlib axis
The result is a matplotlib axis with the relevant plot displayed.
If you are using a notbooks and have ``%matplotlib inline`` set
then this will simply display inline.
"""
import matplotlib.pyplot as plt
from matplotlib import rcParams
from matplotlib.colors import to_hex
if contour: frontier = False
if background is None:
_background = rcParams.get("figure.facecolor")
_background = to_hex(_background) if type(_background) is tuple else _background
set_figure_params('dynamo', background=_background)
else:
_background = background
set_figure_params('dynamo', background=_background)
x, y = [x] if type(x) in [int, str] else x, [y] if type(y) in [int, str] else y
if all([is_gene_name(adata, i) for i in basis]):
if x[0] not in ['M_s', 'X_spliced'] and y[0] not in ['M_u', 'X_unspliced']:
if ('M_s' in adata.layers.keys() and 'M_u' in adata.layers.keys()):
x, y = ['M_s'], ['M_u']
elif ('X_spliced' in adata.layers.keys() and 'X_unspliced' in adata.layers.keys()):
x, y = ['X_spliced'], ['X_unspliced']
else:
x, y = ['spliced'], ['unspliced']
if use_smoothed:
mapper = get_mapper()
# check layer, basis -> convert to list
if type(color) is str:
color = [color]
if type(layer) is str:
layer = [layer]
if type(basis) is str:
basis = [basis]
n_c, n_l, n_b, n_x, n_y = (
1 if color is None else len(color),
1 if layer is None else len(layer),
1 if basis is None else len(basis),
1 if x is None else len(x),
1 if y is None else len(y),
)
point_size = (
16000.0 / np.sqrt(adata.shape[0])
if pointsize is None
else 16000.0 / np.sqrt(adata.shape[0]) * pointsize
)
scatter_kwargs = dict(
alpha=0.1, s=point_size, edgecolor=None, linewidth=0, rasterized=True
) # (0, 0, 0, 1)
if kwargs is not None:
scatter_kwargs.update(kwargs)
font_color = _select_font_color(_background)
total_panels, ncols = n_c * n_l * n_b * n_x * n_y, min(max([n_c, n_l, n_b, n_x, n_y]), ncols)
nrow, ncol = int(np.ceil(total_panels / ncols)), ncols
if figsize is None:
figsize = plt.rcParams["figsize"]
if total_panels >= 1 and ax is None:
plt.figure(None, (figsize[0] * ncol, figsize[1] * nrow), facecolor=_background, dpi=dpi)
gs = plt.GridSpec(nrow, ncol, wspace=0.12)
i = 0
axes_list, color_list = [], []
for cur_b in basis:
for cur_l in layer:
if use_smoothed:
cur_l_smoothed = mapper[cur_l]
prefix = cur_l + "_"
# if prefix + cur_b in adata.obsm.keys():
# if type(x) != str and type(y) != str:
# x_, y_ = (
# adata.obsm[prefix + cur_b][:, int(x)],
# adata.obsm[prefix + cur_b][:, int(y)],
# )
# else:
# continue
for cur_c in color:
cur_title = cur_c
if cur_l in ["protein", "X_protein"]:
_color = adata.obsm[cur_l].loc[cur_c, :]
else:
_color = adata.obs_vector(cur_c, layer=None) if cur_l == 'X' else adata.obs_vector(cur_c, layer=cur_l)
for cur_x, cur_y in zip(x, y):
if type(cur_x) is int and type(cur_y) is int:
points = pd.DataFrame(
{
cur_b + "_0": adata.obsm[prefix + cur_b][:, cur_x],
cur_b + "_1": adata.obsm[prefix + cur_b][:, cur_y],
}
)
points.columns = [cur_b + "_0", cur_b + "_1"]
elif is_gene_name(adata, cur_x) and is_gene_name(adata, cur_y):
points = pd.DataFrame(
{
cur_x: adata.obs_vector(k=cur_x, layer=None) if cur_l_smoothed == 'X' else adata.obs_vector(k=cur_x, layer=cur_l_smoothed),
cur_y: adata.obs_vector(k=cur_y, layer=None) if cur_l_smoothed == 'X' else adata.obs_vector(k=cur_y, layer=cur_l_smoothed),
}
)
# points = points.loc[(points > 0).sum(1) > 1, :]
points.columns = [
cur_x + " (" + cur_l_smoothed + ")",
cur_y + " (" + cur_l_smoothed + ")",
]
cur_title = cur_x + ' VS ' + cur_y
elif is_cell_anno_column(adata, cur_x) and is_cell_anno_column(adata, cur_y):
points = pd.DataFrame(
{
cur_x: adata.obs_vector(cur_x),
cur_y: adata.obs_vector(cur_y),
}
)
points.columns = [cur_x, cur_y]
cur_title = cur_x + ' VS ' + cur_y
elif is_cell_anno_column(adata, cur_x) and is_gene_name(adata, cur_y):
points = pd.DataFrame(
{
cur_x: adata.obs_vector(cur_x),
cur_y: adata.obs_vector(k=cur_y, layer=None) if cur_l_smoothed == 'X' else adata.obs_vector(k=cur_y, layer=cur_l_smoothed),
}
)
# points = points.loc[points.iloc[:, 1] > 0, :]
points.columns = [cur_x, cur_y + " (" + cur_l_smoothed + ")"]
cur_title = cur_y
elif is_gene_name(adata, cur_x) and is_cell_anno_column(adata, cur_y):
points = pd.DataFrame(
{
cur_x: adata.obs_vector(k=cur_x, layer=None) if cur_l_smoothed == 'X' else adata.obs_vector(k=cur_x, layer=cur_l_smoothed),
cur_y: adata.obs_vector(cur_y)
}
)
# points = points.loc[points.iloc[:, 0] > 0, :]
points.columns = [cur_x + " (" + cur_l_smoothed + ")", cur_y]
cur_title = cur_x
elif is_layer_keys(adata, cur_x) and is_layer_keys(adata, cur_y):
add_gamma_fit = True
cur_x_, cur_y_ = adata[:, cur_b].layers[cur_x], adata[:, cur_b].layers[cur_y]
points = pd.DataFrame(
{cur_x: flatten(cur_x_),
cur_y: flatten(cur_y_)}
)
# points = points.loc[points.iloc[:, 0] > 0, :]
points.columns = [cur_x, cur_y]
cur_title = cur_b
if aggregate is not None:
groups, uniq_grp = (
adata.obs[aggregate],
adata.obs[aggregate].unique().to_list(),
)
group_color, group_median = (
np.zeros((1, len(uniq_grp))).flatten()
if isinstance(_color[0], Number)
else np.zeros((1, len(uniq_grp))).astype("str").flatten(),
np.zeros((len(uniq_grp), 2)),
)
grp_size = adata.obs[aggregate].value_counts().values
scatter_kwargs = (
{"s": grp_size}
if scatter_kwargs is None
else update_dict(scatter_kwargs, {"s": grp_size})
)
for ind, cur_grp in enumerate(uniq_grp):
group_median[ind, :] = np.nanmedian(
points.iloc[np.where(groups == cur_grp)[0], :2], 0
)
if isinstance(_color[0], Number):
group_color[ind] = np.nanmedian(
np.array(_color)[np.where(groups == cur_grp)[0]]
)
else:
group_color[ind] = (
pd.Series(_color)[np.where(groups == cur_grp)[0]]
.value_counts()
.index[0]
)
points, _color = (
pd.DataFrame(
group_median, index=uniq_grp, columns=points.columns
),
group_color,
)
# https://stackoverflow.com/questions/4187185/how-can-i-check-if-my-python-object-is-a-number
# answer from Boris.
is_not_continous = (
not isinstance(_color[0], Number) or _color.dtype.name == 'category'
)
if is_not_continous:
labels = _color.to_dense() if is_categorical_dtype(_color) else _color
if theme is None:
if _background in ["#ffffff", "black"]:
_theme_ = "glasbey_dark"
else:
_theme_ = "glasbey_white"
else:
_theme_ = theme
else:
values = _color
if theme is None:
if _background in ["#ffffff", "black"]:
_theme_ = (
"inferno"
if cur_l != "velocity"
else "div_blue_black_red"
)
else:
_theme_ = (
"viridis" if not cur_l.startswith("velocity") else "div_blue_red"
)
else:
_theme_ = theme
_cmap = _themes[_theme_]["cmap"] if cmap is None else cmap
_color_key_cmap = (
_themes[_theme_]["color_key_cmap"]
if color_key_cmap is None
else color_key_cmap
)
_background = (
_themes[_theme_]["background"] if _background is None else _background
)
if labels is not None and values is not None:
raise ValueError(
"Conflicting options; only one of labels or values should be set"
)
if total_panels > 1:
ax = plt.subplot(gs[i])
i += 1
# if highligts is a list of lists - each list is relate to each color element
if highlights is not None:
if is_list_of_lists(highlights):
_highlights = highlights[color.index(cur_c)]
_highlights = (
_highlights
if all([i in _color for i in _highlights])
else None
)
else:
_highlights = (
highlights
if all([i in _color for i in highlights])
else None
)
color_out = None
if smooth and not is_not_continous:
knn = adata.obsp['moments_con']
values = calc_1nd_moment(values, knn)[0] if smooth in [1, True] else \
calc_1nd_moment(values, knn**smooth)[0]
if points.shape[0] <= figsize[0] * figsize[1] * 100000:
ax, color_out = _matplotlib_points(
points.values,
ax,
labels,
values,
highlights,
_cmap,
color_key,
_color_key_cmap,
_background,
figsize[0],
figsize[1],
show_legend,
sort=sort,
frontier=frontier,
contour=contour,
ccmap=ccmap,
calpha=calpha,
sym_c=sym_c,
inset_dict=inset_dict,
**scatter_kwargs
)
else:
ax = _datashade_points(
points.values,
ax,
labels,
values,
highlights,
_cmap,
color_key,
_color_key_cmap,
_background,
figsize[0],
figsize[1],
show_legend,
sort=sort,
frontier=frontier,
contour=contour,
ccmap=ccmap,
calpha=calpha,
sym_c=sym_c,
**scatter_kwargs
)
if i == 1 and show_arrowed_spines:
arrowed_spines(ax, points.columns[:2], _background)
else:
despline_all(ax)
deaxis_all(ax)
ax.set_title(cur_title)
axes_list.append(ax)
color_list.append(color_out)
labels, values = None, None # reset labels and values
if add_gamma_fit and cur_b in adata.var_names[adata.var.use_for_dynamics]:
xnew = np.linspace(0, points.iloc[:, 0].max() * 0.80)
k_name = 'gamma_k' if adata.uns['dynamics']['experiment_type'] == 'one-shot' else 'gamma'
if k_name in adata.var.columns:
if (
not ("gamma_b" in adata.var.columns)
or all(adata.var.gamma_b.isna())
):
adata.var.loc[:, "gamma_b"] = 0
ax.plot(
xnew,
xnew * adata[:, cur_b].var.loc[:, k_name].unique()
+ adata[:, cur_b].var.loc[:, "gamma_b"].unique(),
dashes=[6, 2],
c=font_color,
)
else:
raise Exception(
"adata does not seem to have velocity_gamma column. Velocity estimation is required before "
"running this function."
)
if save_show_or_return == "save":
s_kwargs = {"path": None, "prefix": 'scatters', "dpi": None,
"ext": 'pdf', "transparent": True, "close": True, "verbose": True}
s_kwargs = update_dict(s_kwargs, save_kwargs)
save_fig(**s_kwargs)
if background is not None: reset_rcParams()
elif save_show_or_return == "show":
if show_legend:
plt.subplots_adjust(right=0.85)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
plt.tight_layout()
plt.show()
if background is not None: reset_rcParams()
elif save_show_or_return == "return":
if background is not None: reset_rcParams()
if return_all:
return (axes_list, color_list, font_color) if total_panels > 1 else (ax, color_out, font_color)
else:
return axes_list if total_panels > 1 else ax