import scanpy as sc
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import scrublet as scr
%matplotlib inline
plt.rcParams['pdf.fonttype'] = 42
def us(adata,gene,groups=None, show=False, exclude =None,figsize=None,**kwargs):
from matplotlib import rcParams
if figsize:
rcParams['figure.figsize'] = figsize
if ',' in gene:
gene = gene.split(',')
if groups:
if ',' in groups:groups = groups.split(',')
sc.pl.umap(adata,color=gene,color_map='OrRd',groups=groups, show=show, **kwargs)
else:
if exclude: # list to not show
groups = [x for x in set(adata.obs[gene]) if x not in exclude]
sc.pl.umap(adata,color=gene,color_map='OrRd',groups=groups, show=show, **kwargs)
else:
sc.pl.umap(adata,color=gene,color_map='OrRd',show=show, **kwargs)
rcParams['figure.figsize'] = [3,3]
def dotraw(adata, value, key):
if value not in adata.obs.columns:
if value not in adata.raw.var_names:
print('no'); return
idx = adata.raw.var_names.get_loc(value)
array = adata.raw.X.toarray()[:,idx]
else:
array = np.array(adata.obs[value])
df = pd.DataFrame(array, index=adata.obs.index).merge(adata.obs[[key]],
left_index=True, right_index=True)
df['mask'] = df[0].where(df[0] == 0, 1)
ratio = df.groupby(key)['mask'].mean()
mean = df.groupby(key)[0].mean()
res = pd.concat([ratio,mean], axis=1)
res.columns = ['ratio', 'mean']
return res
def us(adata,gene,groups=None, show=False, exclude =None,figsize=None,**kwargs):
from matplotlib import rcParams
if figsize:
rcParams['figure.figsize'] = figsize
else: rcParams['figure.figsize'] = (3,3)
if ',' in gene:
gene = gene.split(',')
if groups:
sc.pl.umap(adata,color=gene,color_map='OrRd',groups=groups, show=show, **kwargs)
else:
if exclude: # list to not show
groups = [x for x in set(adata.obs[gene]) if x not in exclude]
sc.pl.umap(adata,color=gene,color_map='OrRd',groups=groups, show=show, **kwargs)
else:
sc.pl.umap(adata,color=gene,color_map='OrRd',show=show, **kwargs)
def size(a,b): plt.gcf().set_size_inches(a,b)
def bind(adata, a,b):
order = []
for x in adata.obs[a].cat.categories:
for y in adata.obs[b].cat.categories:
order.append(x+'_'+y)
adata.obs[a+'_'+b] = [x+'_'+y for x,y in zip(adata.obs[a], adata.obs[b])]
adata.obs[a+'_'+b] = adata.obs[a+'_'+b].astype('category')
order = [x for x in order if x in set(adata.obs[a+'_'+b])]
adata.obs[a+'_'+b] = adata.obs[a+'_'+b].cat.reorder_categories(order)
def deg_single(adata, key, target, reference, method='wilcoxon', log=False, filter_res=False, pvalcut = 0.05, foldcut = 1):
import sys
sc.tl.rank_genes_groups(adata, key, groups = [target, reference], reference=reference, method='wilcoxon')
name = pd.DataFrame(adata.uns['rank_genes_groups']['names'])
logf = pd.DataFrame(adata.uns['rank_genes_groups']['logfoldchanges'])
pval = pd.DataFrame(adata.uns['rank_genes_groups']['pvals_adj'])
df = pd.concat([name,logf,pval], axis=1)
df.columns = ['name','logFC', 'pval']
if log:
df['logpval'] = [np.log2(sys.float_info.max-1) if x == 0 else -np.log2(x) for x in df['pval']]
if filter_res:
df = df[df.pval < pvalcut].copy()
df = df[abs(df.logpval) > foldcut].copy()
return df
def degplot(deg, topgene=20, dotsize=3, xlim=None, pvalcut=0.05, foldcut=0):
import seaborn as sns
import matplotlib.pyplot as plt
from adjustText import adjust_text
if 'logpval' not in deg.columns:
deg['logpval'] = [np.log2(sys.float_info.max-1) if x == 0 else -np.log2(x) for x in deg['pval']]
deg = deg[deg.pval < pvalcut].copy()
deg = deg[abs(deg.logpval) > foldcut].copy()
deg = deg.sort_values("logFC", ascending=False)
sns.scatterplot(data = deg, x='logFC', y='logpval', s=dotsize, color='grey')
lim = np.max([abs(deg['logFC'].max()), abs(deg['logFC'].min())])
plt.xlim(-lim-1,lim+1)
if xlim:
plt.xlim(-xlim,xlim)
plt.axvline(x=0, color='black', ls='--', lw=0.7)
for i in range(topgene):
if deg[(deg.pval < pvalcut) & (deg.logFC > foldcut)].shape[0] < i+1: continue
plt.text(deg.loc[deg.index[i],'logFC'], deg.loc[deg.index[i], 'logpval'],
deg.loc[deg.index[i], 'name'], fontsize=6, color='red')
for i in range(topgene):
if deg[(deg.pval < pvalcut) & (deg.logFC < foldcut)].shape[0] < i+1: continue
plt.text(deg.loc[deg.index[-(i+1)],'logFC'], deg.loc[deg.index[-(i+1)], 'logpval'],
deg.loc[deg.index[-(i+1)], 'name'], fontsize=6, color='blue')
adjust_text(plt.gca().texts, arrowprops=dict(arrowstyle='-', lw=0.3, color='grey'), precision = 0.3)
plt.ylabel('log [Adjusted P-value]')
plt.xlabel('log [Fold Changes]')
sns.despine(left=True)
for i in range(topgene):
if deg[(deg.pval < pvalcut) & (deg.logFC > foldcut)].shape[0] < i+1: continue
plt.plot(deg.loc[deg.index[i],'logFC'], deg.loc[deg.index[i], 'logpval'],
color='red', markersize=dotsize/3, marker='o')
for i in range(topgene):
if deg[(deg.pval < pvalcut) & (deg.logFC < foldcut)].shape[0] < i+1: continue
plt.plot(deg.loc[deg.index[-(i+1)],'logFC'], deg.loc[deg.index[-(i+1)], 'logpval'],
color='blue', markersize=dotsize/3, marker='o')
plt.rcParams['figure.dpi'] = 300
