def plot_categ_spatial(mod,
adata,
sample_col,
color,
n_columns=2,
figure_size=(24, 5.7),
point_size=0.8,
text_size=9):
for_plot = adata.obs[["imagecol", "imagerow", sample_col]]
for_plot["color"] = color
# fix types
for_plot["color"] = pd.Categorical(for_plot["color"], ordered=True)
# for_plot['color'] = pd.to_numeric(for_plot['color'])
for_plot["sample"] = pd.Categorical(for_plot[sample_col], ordered=False)
for_plot["imagecol"] = pd.to_numeric(for_plot["imagecol"])
for_plot["imagerow"] = -pd.to_numeric(for_plot["imagerow"])
ax = (
plotnine.ggplot(
for_plot, plotnine.aes(x="imagecol", y="imagerow", color="color"))
+ plotnine.geom_point(size=point_size) # + plotnine.scale_color_cmap()
+ plotnine.coord_fixed() + plotnine.theme_bw() + plotnine.theme(
panel_background=plotnine.element_rect(
fill="black", colour="black", size=0, linetype="solid"),
panel_grid_major=plotnine.element_line(
size=0, linetype="solid", colour="black"),
panel_grid_minor=plotnine.element_line(
size=0, linetype="solid", colour="black"),
strip_text=plotnine.element_text(size=text_size),
) + plotnine.facet_wrap("~sample", ncol=n_columns) +
plotnine.theme(figure_size=figure_size))
return ax
def plot_counts(counts: pd.DataFrame):
counts.bc_l = pd.Categorical(counts.bc_l, bc_range(counts.bc_l))
counts.bc_r = pd.Categorical(counts.bc_r, bc_range(counts.bc_r))
log = np.log
return (
ggplot(counts, aes('bc_r', 'bc_l', fill='log(Count)'))
+ geom_tile()
# + scale_y_reverse()
+ coord_fixed()
+ scale_fill_cmap('inferno')
+ theme(axis_text_x=element_text(angle=90, vjust=.5))
+ labs(x='Right Barcode', y='Left Barcode')
)
def pictures(self, mode='bw', subset=None, n_random=10):
"""Returns a picture of the selected images.
Creates either a colored or a black-white picture of the selected
images.
Args:
mode: Should the picture be black-white ('bw') or in color
('color')?
subset: Optional list of picture indices that should be included in
the dataframe. If specified, n_random will be ignored.
n_random: Optional number of randomly selected images. If neither
subset nor n_random are specified, all images will be included.
Returns:
A plotnine object including all pictures with their label.
Raises:
NotImplementedError: mode must be either 'bw' or 'color'."""
dataframe = self.rgb_dataframe(subset=subset, n_random=n_random)
if mode == 'bw':
fill_key = 'rgb_bw'
elif mode == 'color':
fill_key = 'rgb'
else:
raise NotImplementedError("Pictures are either in black-white"
"('bw') or in color ('color').")
picture = (
gg.ggplot(dataframe, gg.aes(x='x', y='y', fill=fill_key)) +
gg.geom_tile() + gg.theme_void() +
gg.theme(legend_position='none') + gg.scale_fill_manual(
values={key: key
for key in dataframe[fill_key].unique()}) +
gg.facet_wrap('image_id', labeller=self.labeller) +
gg.scale_y_reverse() + gg.coord_fixed())
return picture
def plot_factor_spatial(
adata,
fact,
cluster_names,
fact_ind=[0],
trans="log",
sample_name=None,
samples_col="sample",
obs_x="imagecol",
obs_y="imagerow",
n_columns=6,
max_col=5000,
col_breaks=[0.1, 100, 1000, 3000],
figure_size=(24, 5.7),
point_size=0.8,
text_size=9,
):
r"""Plot expression of factors / cell types in space.
Convenient but not as powerful as scanpy plotting.
:param adata: anndata object with spatial data
:param fact: pd.DataFrame with spatial expression of factors (W), e.g. mod.spot_factors_df
:param cluster_names: names of those factors to show on a plot
:param fact_ind: index of factors to plot
:param trans: transform colorscale? passed to plotnine.scale_color_cmap
:param sample_name: if anndata object contains multiple samples specify which sample to plot (no warning given if not)
:param samples_col: if anndata object contains multiple which .obs columns specifies sample?
:param obs_x: which .obs columns specifies x coordinate?
:param obs_y: which .obs columns specifies y coordinate?
:param n_columns: how many factors / clusters to plot in each row (plotnine.facet_grid)
:param max_col: colorscale maximum expression in fact
:param col_breaks: colorscale breaks
:param figure_size: figures size works weirdly (only x axis has an effect, use 24 for 6-column plot, 12 for 3, 8 for 2 ...).
:param point_size: point size of spots
:param text_size: text size
"""
if sample_name is not None:
sample_ind = np.isin(adata.obs[samples_col], sample_name)
else:
sample_ind = np.repeat(True, adata.shape[0])
# adata.obsm['X_spatial'][:,0] vs adata.obs['imagecol'] & adata.obs['imagerow']
for_plot = np.concatenate(
(
adata.obs[obs_x].values.reshape((adata.obs.shape[0], 1)),
-adata.obs[obs_y].values.reshape((adata.obs.shape[0], 1)),
fact.iloc[:, fact_ind[0]].values.reshape((adata.obs.shape[0], 1)),
np.array([
cluster_names[fact_ind[0]] for j in range(adata.obs.shape[0])
]).reshape((adata.obs.shape[0], 1)),
),
1,
)
for_plot = pd.DataFrame(
for_plot,
index=adata.obs.index,
columns=["imagecol", "imagerow", "weights", "cluster"])
# select only correct sample
for_plot = for_plot.loc[sample_ind, :]
for i in fact_ind[1:]:
for_plot1 = np.concatenate(
(
adata.obs[obs_x].values.reshape((adata.obs.shape[0], 1)),
-adata.obs[obs_y].values.reshape((adata.obs.shape[0], 1)),
fact.iloc[:, i].values.reshape((adata.obs.shape[0], 1)),
np.array([cluster_names[i]
for j in range(adata.obs.shape[0])]).reshape(
(adata.obs.shape[0], 1)),
),
1,
)
for_plot1 = pd.DataFrame(
for_plot1,
index=adata.obs.index,
columns=["imagecol", "imagerow", "weights", "cluster"])
# select only correct sample
for_plot1 = for_plot1.loc[sample_ind, :]
for_plot = pd.concat((for_plot, for_plot1))
for_plot["imagecol"] = pd.to_numeric(for_plot["imagecol"])
for_plot["imagerow"] = pd.to_numeric(for_plot["imagerow"])
for_plot["weights"] = pd.to_numeric(for_plot["weights"])
for_plot["cluster"] = pd.Categorical(for_plot["cluster"],
categories=cluster_names[fact_ind],
ordered=True)
# print(np.log(np.max(for_plot['weights'])))
ax = (plotnine.ggplot(
for_plot, plotnine.aes("imagecol", "imagerow", color="weights")) +
plotnine.geom_point(size=point_size) +
plotnine.scale_color_cmap("magma",
trans=trans,
limits=[0.1, max_col],
breaks=col_breaks + [max_col]) +
plotnine.coord_fixed() + plotnine.theme_bw() + plotnine.theme(
panel_background=plotnine.element_rect(
fill="black", colour="black", size=0, linetype="solid"),
panel_grid_major=plotnine.element_line(
size=0, linetype="solid", colour="black"),
panel_grid_minor=plotnine.element_line(
size=0, linetype="solid", colour="black"),
strip_text=plotnine.element_text(size=text_size),
) + plotnine.facet_wrap("~cluster", ncol=n_columns) +
plotnine.ggtitle("nUMI from each cell type") +
plotnine.theme(figure_size=figure_size))
return ax
on="site").merge(loc_df,
on="site_location")
# Plot total number of cells per well
cell_count_totalcells_df = (cell_count_df.groupby(
["x_loc", "y_loc", "well", "site_location",
"site"])["total_cell_count"].mean().reset_index())
plate = cell_count_df["plate"].unique()[0]
os.makedirs(output_figuresdir, exist_ok=True)
by_well_gg = (
gg.ggplot(cell_count_totalcells_df, gg.aes(x="x_loc", y="y_loc")) +
gg.geom_point(gg.aes(fill="total_cell_count"), size=10) +
gg.geom_text(gg.aes(label="site_location"), color="lightgrey") +
gg.facet_wrap("~well") + gg.coord_fixed() + gg.theme_bw() +
gg.ggtitle(f"Total Cells/Well\n{plate}") + gg.theme(
axis_text=gg.element_blank(),
axis_title=gg.element_blank(),
strip_background=gg.element_rect(colour="black", fill="#fdfff4"),
) + gg.labs(fill="Cells") + gg.scale_fill_cmap(name="magma"))
output_file = pathlib.Path(output_figuresdir,
"plate_layout_cells_count_per_well.png")
if check_if_write(output_file, force, throw_warning=True):
by_well_gg.save(output_file, dpi=300, verbose=False)
# Plot cell category ratios per well
ratio_df = pd.pivot_table(
cell_count_df,
values="cell_count",
def test_coord_fixed():
assert p + coord_fixed(0.5) == 'coord_fixed'
def test_coord_fixed():
assert p + coord_fixed(0.5) == 'coord_fixed'
cell_count_df.groupby(["x_loc", "y_loc", "well", "site_location", "site"])[
"total_cell_count"
]
.mean()
.reset_index()
)
plate = cell_count_df["plate"].unique()[0]
os.makedirs(output_figuresdir, exist_ok=True)
by_well_gg = (
gg.ggplot(cell_count_totalcells_df, gg.aes(x="x_loc", y="y_loc"))
+ gg.geom_point(gg.aes(fill="total_cell_count"), size=10)
+ gg.geom_text(gg.aes(label="site_location"), color="lightgrey")
+ gg.facet_wrap("~well")
+ gg.coord_fixed()
+ gg.theme_bw()
+ gg.ggtitle(f"Total Cells/Well\n{plate}")
+ gg.theme(
axis_text=gg.element_blank(),
axis_title=gg.element_blank(),
strip_background=gg.element_rect(colour="black", fill="#fdfff4"),
)
+ gg.labs(fill="Cells")
+ gg.scale_fill_cmap(name="magma")
)
output_file = pathlib.Path(output_figuresdir, "plate_layout_cells_count_per_well.png")
if check_if_write(output_file, force, throw_warning=True):
by_well_gg.save(output_file, dpi=300, verbose=False)
feed = {}
feed.update(tr_isotropic.params["range"].tf_feed_entry)
feed.update(tr_anis_2D.params["maxrange"].tf_feed_entry)
feed.update(tr_anis_2D.params["minrange_fct"].tf_feed_entry)
feed.update(tr_anis_2D.params["azimuth"].tf_feed_entry)
feed.update(tr_proj.params["azimuth"].tf_feed_entry)
feed.update(tr_proj.params["range"].tf_feed_entry)
with tf.Session(graph=g) as session:
session.run(init, feed_dict=feed)
out_identity = test_identity.eval(session=session, feed_dict=feed)
out_isotropic = test_isotropic.eval(session=session, feed_dict=feed)
out_anis_2D = test_anis_2D.eval(session=session, feed_dict=feed)
out_proj = test_proj.eval(session=session, feed_dict=feed)
# visual testing
df = pd.DataFrame({"x": test_identity[:, 0], "y": test_identity[:, 1]})
fig_id = p9.ggplot(df) +\
p9.geom_point(p9.aes("x", "y")) +\
p9.coord_fixed()
df = pd.DataFrame({"x": test_isotropic[:, 0], "y": test_isotropic[:, 1]})
fig_iso = p9.ggplot(df) +\
p9.geom_point(p9.aes("x", "y")) +\
p9.coord_fixed()
df = pd.DataFrame({"x": test_anis_2D[:, 0], "y": test_anis_2D[:, 1]})
fig_anis2D = p9.ggplot(df) +\
p9.geom_point(p9.aes("x", "y")) +\
p9.coord_fixed()