def plot_cy_embedding(clust_object, folder):
"""Cytokine embedding"""
plt.figure(901, figsize=(13, 9.7))
cyplot.plotModuleEmbedding(clust_object.dmatDf,
clust_object.labels,
method='kpca')
colors = palettable.colorbrewer.get_map(
'Set1', 'qualitative', len(np.unique(clust_object.labels))).mpl_colors
colorLegend(colors, [
'%s%1.0f' % (clust_object.sampleStr, i)
for i in np.unique(clust_object.labels)
],
loc='lower left')
plt.figure(901).savefig(os.path.join(
folder, '%s_embedding.png' % clust_object.name),
dpi=300)
def plot_color_legend(clust_object, folder):
"""color_label_legend"""
plt.figure(48, figsize=(3, 3))
plt.clf()
axh = plt.subplot(1, 1, 1)
colorLegend(palettable.colorbrewer.qualitative.Set3_6.mpl_colors,
['%s' % s for s in clust_object.modDf.columns],
loc=10)
axh.spines['right'].set_color('none')
axh.spines['left'].set_color('none')
axh.spines['top'].set_color('none')
axh.spines['bottom'].set_color('none')
axh.set_xticks([])
axh.set_yticks([])
axh.set_facecolor('white')
plt.figure(48).savefig(os.path.join(
folder, '%s_color_label_legend.png' % clust_object.name),
dpi=300)
def plotEpitopeCDF(df, ba):
plt.clf()
ic50Ticks = [20, 50, 200, 500, 1000, 2000, 5000, 10000, 20000]
ECDF = sm.distributions.empirical_distribution.ECDF
colors = palettable.colorbrewer.qualitative.Set1_3.mpl_colors
for x,locus in enumerate('ABC'):
obsInd = df.HLA.str.slice(stop=1) == locus
ecdf = ECDF(df['log_IC50'].loc[obsInd])
plt.step(ecdf.x, ecdf.y, '-', color=colors[x], lw=3)
ref = [v for k,v in ba.items() if k[0][0]==locus and not k in df.ba_key]
ecdf = ECDF(ref)
plt.step(ecdf.x, ecdf.y, '--', color=colors[x], alpha=1)
plt.ylim((0,1))
colorLegend(colors=colors[:3], labels=['HLA-A', 'HLA-B', 'HLA-C'])
plt.xlim((0,15))
plt.xticks(np.log(ic50Ticks), ic50Ticks)
plt.xlabel('Predicted HLA binding $IC_{50}$ (nM)')
plt.ylabel('Fraction of epitopes')
tickSz='large',
vRange=(0, 1))
plt.figure(43, figsize=(15.5, 9.5))
colInds = cy.plotting.plotHierClust(1 - rcyc.pwrel,
rcyc.Z,
labels=rcyc.labels,
titleStr='Pairwise reliability (%s)' %
rcyc.name,
vRange=(0, 1),
tickSz='large')
plt.figure(901, figsize=(13, 9.7))
cy.plotting.plotModuleEmbedding(rcyc.dmatDf,
rcyc.labels,
method='kpca',
txtSize='large')
colors = palettable.colorbrewer.get_map('Set1', 'qualitative',
len(np.unique(rcyc.labels))).mpl_colors
colorLegend(colors,
['%s%1.0f' % (rcyc.sampleStr, i) for i in np.unique(rcyc.labels)],
loc='lower left')
"""df here should have one column per module and the genotype column"""
ptidDf = longDf[['ptid', 'sample', 'genotype',
'dpi']].drop_duplicates().set_index('ptid')
df = rcyc.modDf.join(ptidDf)
ind = df.genotype == 'WT'
col = 'SERUM1'
stats.ranksums(df[col].loc[ind], df[col].loc[~ind])
def plotHLAFreq(hlaDf,twoDigit=True,groupColumn=None,stackColumn=None,loci='AB',cutoff=0.05,annotateCount=True):
"""Bar plot of HLA allele expression (optionally by group)
Requires "hlas" column in hlaDf which is a list of alleles"""
allHLAs=[]
for hlas in hlaDf.hlas:
allHLAs.extend(hlas)
if twoDigit:
allHLAs=[h[:4] for h in allHLAs]
uHLA=list(set(allHLAs))
if not groupColumn is None:
uGroup=sorted(hlaDf[groupColumn].unique().tolist())
groupS=hlaDf[groupColumn]
else:
uGroup=['All']
groupS=pd.Series(data=[uGroup[0] for i in arange(hlaDf.shape[0])], index=hlaDf.index, name='Group')
if not stackColumn is None:
uStack=sorted(hlaDf[stackColumn].unique().tolist())
stackS=hlaDf[stackColumn]
else:
uStack=['All']
stackS=pd.Series(data=[uStack[0] for i in arange(hlaDf.shape[0])], index=hlaDf.index, name='Stack')
counts=zeros((len(uGroup), len(uStack), len(uHLA)))
for hlas, group, stack in zip(hlaDf.hlas, groupS, stackS):
if twoDigit:
tmp={h[:4] for h in hlas}
else:
tmp=set(hlas)
for h in tmp:
"""For each person's unique alleles, increment the counter"""
hlai=uHLA.index(h)
groupi=uGroup.index(group)
stacki=uStack.index(stack)
counts[groupi, stacki, hlai]+=1
sHLA=sorted(uHLA, key=lambda h: counts[:,:, uHLA.index(h)].sum(), reverse=True)
mycm=array(['darkgreen', 'fuchsia', 'saddlebrown', 'lightseagreen', 'gold', 'royalblue', 'tomato', 'thistle', 'tan'])
myalphas=linspace(1, 0.3, len(uStack))
N=hlaDf.shape[0]
groupN=objhist(groupS)
w=1/(len(uGroup)+2)
clf()
for locusi, locus in enumerate(loci):
subplot(len(loci), 1, locusi+1)
xlabs=[]
x=0
other=zeros((len(uGroup), len(uStack)))
for hla in sHLA:
hlai=uHLA.index(hla)
if hla[0]==locus:
if counts[:,:, hlai].sum()/N > cutoff:
xlabs.append(hla)
for groupi, group in enumerate(uGroup):
for stacki, stack in enumerate(uStack):
bar(left=x+w*groupi-len(uGroup)*w/2,
height=1e2*counts[groupi, stacki, hlai]/groupN[group],
bottom=1e2*counts[groupi, :stacki, hlai].sum()/groupN[group],
width=w, color=mycm[groupi], alpha=myalphas[stacki])
if counts[groupi, stacki, hlai]>1 and annotateCount:
annotate('%d' % counts[groupi, stacki, hlai],
xy=(x+w*groupi-len(uGroup)*w/2+w/2, 1e2*counts[groupi, :(stacki+1), hlai].sum()/groupN[group]),
ha='center', va='top', xytext=(0, -4), textcoords='offset points')
x+=1
else:
for groupi, group in enumerate(uGroup):
for stacki, stack in enumerate(uStack):
other[groupi, stacki]+=counts[groupi, stacki, hlai]
xlabs.append('Other')
for groupi, group in enumerate(uGroup):
for stacki, stack in enumerate(uStack):
bar(left=x+w*groupi,
height=1e2*other[groupi, stacki]/groupN[group],
bottom=1e2*other[groupi, :stacki].sum()/groupN[group],
width=w, color=mycm[groupi], alpha=myalphas[stacki])
if len(uGroup)>1:
sorti=argsort(uGroup)
colorLegend(colors=mycm[:len(uGroup)][sorti], labels=[uGroup[i] for i in sorti], title=groupColumn)
ylabel('% of participants', size='x-large')
xticks(arange(len(xlabs)), xlabs, size='x-large')
xlim((-0.5, len(xlabs)+1))
def plot_modules(clust_object, folder):
'''Plot cytomod object modules information'''
"""Hierarchical clustering heatmap"""
plt.figure(41, figsize=(15.5, 9.5))
# colInds = plotHColCluster(ds[s].cyDf, method='complete', metric='pearson-signed', col_labels=ds[s].labels, col_dmat=ds[s].dmatDf)
colInds = plotHColCluster(clust_object.cyDf,
method='complete',
metric='pearson-signed',
col_labels=clust_object.labels,
save_path=os.path.join(
folder, '%s_hierchical_clust_heatmap.png' %
clust_object.name))
# plt.figure(41).savefig(os.path.join(folder, '%s_hierchical_clust_heatmap.png' % clust_object.name))
"""Heatmaps of pairwise reliability"""
plt.figure(43, figsize=(15.5, 9.5))
colInds = cyplot.plotHierClust(1 - clust_object.pwrel,
clust_object.Z,
labels=clust_object.labels,
titleStr='Pairwise reliability (%s)' %
clust_object.name.replace('_', ' '),
vRange=(0, 1))
plt.figure(43).savefig(os.path.join(folder,
'%s_pwrel.png' % clust_object.name),
dpi=300)
"""color_label_legend"""
plt.figure(48, figsize=(3, 3))
plt.clf()
axh = plt.subplot(1, 1, 1)
colorLegend(palettable.colorbrewer.qualitative.Set3_6.mpl_colors,
['%s' % s for s in clust_object.modDf.columns],
loc=10)
axh.spines['right'].set_color('none')
axh.spines['left'].set_color('none')
axh.spines['top'].set_color('none')
axh.spines['bottom'].set_color('none')
axh.set_xticks([])
axh.set_yticks([])
axh.set_facecolor('white')
plt.figure(48).savefig(os.path.join(folder, 'color_label_legend.png'),
dpi=300)
"""Plot intra-module correlation"""
plt.figure(50, figsize=(15, 9))
for lab in list(
cy.labels2modules(clust_object.labels,
clust_object.dropped).keys()):
cyplot.plotModuleCorr(clust_object.cyDf,
clust_object.labels,
lab,
dropped=clust_object.dropped)
plt.figure(50).savefig(os.path.join(
folder, '%s_mod_corr_%s.png' % (clust_object.name, lab)),
dpi=300)
"""Cytokine embedding"""
plt.figure(901, figsize=(13, 9.7))
cyplot.plotModuleEmbedding(clust_object.dmatDf,
clust_object.labels,
method='kpca')
colors = palettable.colorbrewer.get_map(
'Set1', 'qualitative', len(np.unique(clust_object.labels))).mpl_colors
colorLegend(colors, [
'%s%1.0f' % (clust_object.sampleStr, i)
for i in np.unique(clust_object.labels)
],
loc='lower left')
plt.figure(901).savefig(os.path.join(folder,
'%sembed.png' % clust_object.name),
dpi=300)