def plot_revigo(
rev,
outline=2,
expand_points=(1.05, 1.2),
figure_size=(8, 8),
font_size=8,
point_size=3,
point_alpha=0.7,
palette='RdPu',
dispensability_cutoff=1.,
show_all_labels=False,
text_column='name',
term_size_limit=None,
):
import plotnine as p9
import matplotlib.patheffects as path_effects
pe = [
path_effects.Stroke(linewidth=2, foreground='white'),
path_effects.Normal()
]
if not show_all_labels:
lbl_df = rev[(rev.eliminated == 0)
& (rev.dispensability < dispensability_cutoff)]
if term_size_limit is not None:
lbl_df = lbl_df[lbl_df.term_size < term_size_limit]
else:
lbl_df = rev
g = (p9.ggplot(p9.aes(x='plot_X', y='plot_Y'), data=rev) +
p9.geom_point(p9.aes(fill='neglog10', size='frequency'),
color='black',
alpha=point_alpha) +
p9.geom_text(p9.aes(label=text_column),
data=lbl_df,
size=font_size,
adjust_text={
'expand_points': expand_points,
'arrowprops': {
'arrowstyle': '-'
},
'x': rev.plot_X.values,
'y': rev.plot_Y.values
},
path_effects=pe) + p9.theme_bw() +
p9.scale_fill_distiller(type='seq', palette=palette, direction=1) +
p9.labs(x='Semantic similarity space',
y='',
fill='-log10(adj. p-value)',
size='Term frequency') +
p9.scale_size_continuous(range=(2, 7), trans='log10') +
p9.theme(figure_size=figure_size,
axis_text_x=p9.element_blank(),
axis_text_y=p9.element_blank(),
axis_ticks=p9.element_blank()))
return g
def test_text_aesthetics():
p = (
ggplot(df, aes(y='y', label='label')) +
geom_text(aes('x', label='label'), size=15, ha='left') + geom_text(
aes('x+1', angle='angle'), size=15, va='top', show_legend=False) +
geom_text(
aes('x+2', label='label', alpha='z'), size=15, show_legend=False) +
geom_text(aes('x+3', color='factor(z)'), size=15, show_legend=False) +
geom_text(aes('x+5', size='z'), ha='right', show_legend=False) +
scale_size_continuous(range=(12, 30)) +
scale_y_continuous(limits=(-0.5, n - 0.5)))
assert p == 'text_aesthetics'
def test_text_aesthetics():
p = (ggplot(df, aes(y='y', label='label')) +
geom_text(aes('x', label='label'), size=15, ha='left') +
geom_text(aes('x+1', angle='angle'),
size=15, va='top', show_legend=False) +
geom_text(aes('x+2', label='label', alpha='z'),
size=15, show_legend=False) +
geom_text(aes('x+3', color='factor(z)'),
size=15, show_legend=False) +
geom_text(aes('x+5', size='z'),
ha='right', show_legend=False) +
scale_size_continuous(range=(12, 30)) +
scale_y_continuous(limits=(-0.5, n-0.5)))
assert p == 'text_aesthetics'
def ggpca(x,
y=None,
center='col',
scale='none',
rlab=False,
clab=False,
cshow=None,
rsize=4,
csize=2,
lsize=10,
lnudge=0.03,
ralpha=0.6,
calpha=1.0,
clightalpha=0,
rname='sample',
cname='variable',
lname='',
grid=True,
printit=False,
xsvd=None,
invert1=False,
invert2=False,
colscale=None,
**kwargs):
if cshow is None:
cshow = x.shape[1]
if rlab is not None and isinstance(rlab, bool):
rlab = x.index if rlab else ''
if clab is not None and isinstance(clab, bool):
clab = x.columns if clab else ''
if y is not None:
pass
x = x.loc[:, x.isnull().sum(axis=0) == 0]
if xsvd is None:
xsvd = svdForPca(x, center, scale)
rsf = np.max(xsvd[0].iloc[:, 0]) - np.min(xsvd[0].iloc[:, 0])
csf = np.max(xsvd[2].iloc[0, :]) - np.min(xsvd[2].iloc[0, :])
sizeRange = sorted([csize, rsize])
alphaRange = sorted([calpha, ralpha])
ggd = pd.DataFrame({
'PC1': xsvd[0].iloc[:, 0] / rsf,
'PC2': xsvd[0].iloc[:, 1] / rsf,
'label': rlab,
'size': rsize,
'alpha': ralpha
})
cclass = []
if cshow > 0:
cdata = pd.DataFrame({
'PC1': xsvd[2].iloc[0, :] / csf,
'PC2': xsvd[2].iloc[1, :] / csf,
'label': clab,
'size': csize,
'alpha': calpha
})
if cshow < x.shape[1]:
cscores = cdata['PC1']**2 + cdata['PC2']**2
keep = cscores.sort_values(ascending=False).head(cshow).index
if clightalpha > 0:
cdata.loc[~cdata.index.isin(keep), 'label'] = ''
cdata.loc[~cdata.index.isin(keep), 'alpha'] = clightalpha
alphaRange = [
np.min([alphaRange[0], clightalpha]),
np.max([alphaRange[1], clightalpha])
]
else:
cdata = cdata.loc[cdata.index.isin(keep)]
ggd = pd.concat([cdata, ggd])
cclass = [cname] * cdata.shape[0]
if invert1:
ggd['PC1'] = -ggd['PC1']
if invert2:
ggd['PC2'] = -ggd['PC2']
if y is not None:
ggd['class'] = cclass + list(y.loc[x.index])
else:
ggd['class'] = cclass + ([rname] * x.shape[0])
ggo = gg.ggplot(
ggd,
gg.aes(x='PC1',
y='PC2',
color='class',
size='size',
alpha='alpha',
label='label'))
ggo += gg.geom_hline(yintercept=0, color='lightgray')
ggo += gg.geom_vline(xintercept=0, color='lightgray')
ggo += gg.geom_point()
ggo += gg.theme_bw()
ggo += gg.geom_text(nudge_y=lnudge, size=lsize, show_legend=False)
if colscale is None and len(ggd['class'].unique()) < 8:
colscale = [
'darkslategray', 'goldenrod', 'lightseagreen', 'orangered',
'dodgerblue', 'darkorchid'
]
colscale = colscale[0:(len(ggd['class'].unique()) - 1)] + ['gray']
if len(colscale) == 2 and cshow > 0:
colscale = ['black', 'darkgray']
if len(colscale) == 2 and cshow == 0:
colscale = ['black', 'red']
if len(colscale) == 3:
colscale = ['black', 'red', 'darkgray']
ggo += gg.scale_color_manual(values=colscale, name=lname)
ggo += gg.scale_size_continuous(guide=False, range=sizeRange)
ggo += gg.scale_alpha_continuous(guide=False, range=alphaRange)
ggo += gg.xlab('PC1 (' +
str(np.round(100 * xsvd[1][0]**2 /
((xsvd[1]**2).sum()), 1)) +
'% explained var.)')
ggo += gg.ylab('PC2 (' +
str(np.round(100 * xsvd[1][1]**2 /
((xsvd[1]**2).sum()), 1)) +
'% explained var.)')
if not grid:
ggo += gg.theme(panel_grid_minor=gg.element_blank(),
panel_grid_major=gg.element_blank(),
panel_background=gg.element_blank())
ggo += gg.theme(axis_ticks=gg.element_blank(),
axis_text_x=gg.element_blank(),
axis_text_y=gg.element_blank())
if printit:
print(ggo)
return ggo
dataset = "dmso_treated"
umap_resistant_type_gg = (
gg.ggplot(embedding_df, gg.aes(x="x", y="y"))
+ gg.geom_point(
gg.aes(fill="Metadata_clone_type", shape="Metadata_batch", size="Metadata_cell_count"),
color='black', alpha=0.6)
+ gg.theme_bw()
+ gg.xlab("UMAP (X)")
+ gg.ylab("UMAP (Y)")
+ gg.ggtitle("DMSO treated samples")
+ gg.scale_shape_manual(name="Batch", values=[".", "+", "x"])
+ gg.scale_fill_manual(name="Clone type", values=["#1F8AA5", "#E98831"])
+ gg.scale_size_continuous(name="Cell count")
+ gg.theme(
strip_text=gg.element_text(size=6, color="black"),
strip_background=gg.element_rect(colour="black", fill="#fdfff4"),
)
)
file = os.path.join("figures", "umap", f"{dataset}_umap_resistant_type")
for extension in save_file_extensions:
umap_resistant_type_gg.save(filename='{}{}'.format(file, extension), height=3, width=3.5, dpi=400)
umap_resistant_type_gg
# In[10]: