def add_rs_bars(self, rs, options, sim_rs=[]):
r_width = 0.2
s_width = (1.0 - r_width) / len(sim_rs)
myticks, myticklabels = [], []
for i, k in enumerate(self.r_key):
myticks.append(i)
myticklabels.append(' > ' + str(k))
self.ax.bar(i,
rs[i],
width=r_width,
color=options.plotColors['RESULTS'])
iLoc = i + r_width
for b in sorted([bh[i] for bh in sim_rs], reverse=True):
self.ax.bar(iLoc,
b,
width=s_width,
color=options.plotColors['SIMS'])
iLoc += s_width
self.ax.set_xticks(myticks)
self.ax.set_xticklabels(myticklabels, horizontalalignment='left')
items = [
Rect((0, 0), 1, 1, fc=options.plotColors['RESULTS']),
Rect((0, 0), 1, 1, fc=options.plotColors['SIMS'])
]
self.ax.legend(items, ['Results', 'Simulation'],
loc='upper right',
bbox_to_anchor=(1, 1),
fontsize=10)
return self
def fill_boxes(self,opts,vals,colors,alt_colors,JITTERS=None,SIG_STARS=None):
self.bp = self.axes[-1].boxplot(vals,positions=self.pos, widths=self.width, patch_artist=True, showmeans=True,whis=0.7)
for i,opt in enumerate(opts):
clr,ac = colors[i], alt_colors[i]
self.bp['boxes'][i].set_edgecolor(clr)
self.bp['boxes'][i].set_linewidth(1)
plt.setp(self.bp['medians'][i], color=clr, linewidth=3)
plt.setp(self.bp['means'][i], marker='h',markersize=9,markerfacecolor=clr)
plt.setp(self.bp['caps'][(i*2)+1], color=clr,linewidth=1)
plt.setp(self.bp['caps'][(i*2)], color=clr,linewidth=1)
plt.setp(self.bp['whiskers'][i*2], color=clr,linewidth=1)
plt.setp(self.bp['whiskers'][1+(i*2)], color=clr,linewidth=1)
plt.setp(self.bp['fliers'][i], markerfacecolor=clr, markeredgecolor = clr, marker='s',markersize=2.0)
if clr == ac: self.items[opt] = Rect((0,0),1,1,fc=clr)
else: self.items[opt] = Rect((0,0),1,1,fc=a,ec=b,hatch='*')
for patch, clevel in zip(self.bp['boxes'], colors):
patch.set_facecolor(clevel); patch.set_alpha(0.5)
if JITTERS != None:
for i,(xJ,yJ) in enumerate(JITTERS):
plt.scatter(xJ, yJ, c=colors[i], alpha=0.7,s=4,zorder=9)
if SIG_STARS != None:
for i,SIG in enumerate(SIG_STARS):
if SIG: plt.scatter(self.pos[i],max(vals[i])*1.2,s=100,marker='*',color='gold')
def finish(self):
if self.options.verbose: sys.stderr.write('\nAdding Legend...')
plt.axis('off')
items,labels=[],[]
leg_items,leg_labels,leg_ids = [],[],[]
for ID in self.multi_legend:
items.append([])
labels.append([])
for a,b in self.multi_colors[ID].items():
items[-1].append(Rect((0,0),1,1,fc=b))
labels[-1].append(a)
leg_ids.append(ID)
maxLen= max([len(x) for x in labels])
leg_items,leg_labels = [[] for x in range(maxLen)],[[] for x in range(maxLen)]
for I,L in zip(items,labels):
while len(I) < maxLen:
I.append(Rect((0,0),1,1,fc='white',ec='white'))
L.append('')
for k in range(len(L)):
leg_items[k].append(I[k])
leg_labels[k].append(L[k])
leg_items = [[Rect((0,0),1,1,fc='white',ec='white') for x in leg_ids]]+leg_items
leg_labels = [leg_ids]+leg_labels
ncols = len(leg_ids)
ncols = maxLen+1
leg_items = [a for b in leg_items for a in b]
leg_labels = [a for b in leg_labels for a in b]
#hp,cs = 0.1, 0.1
hp,cs = 0.13, 0.3
plt.legend(leg_items,leg_labels,ncol=ncols,bbox_to_anchor=(1.2,1.7),loc='upper left',handletextpad=hp,columnspacing=cs,fontsize=12)
plt.subplots_adjust(left=0.02, bottom=0.02, right=0.98, top=0.82) # ,wspace=0.25,hspace=1.05)
plt.savefig(self.mname+'.png',dpi=500)
if options.show:
plt.show()
if self.options.verbose: sys.stderr.write('\n')
def make_minor_boxes(self, major_id, opts, vals):
#self.ax.bar(xLoc+xOffset,X_key[X][s],width=1,bottom=0,color=a,ec=b,hatch='*')
#items.append(Rect((0,0),1,1,fc=a,ec=b,hatch='*'))
major_clrs, minor_clrs, global_sigs, local_sigs, jitters = [
self.get_color(major_id) for X in opts
], [self.get_color(X)
for X in opts], [1 for X in opts], [1 for X in opts], []
self.pos, self.width = [
self.xOffset + 0.20 * x for x in range(len(opts))
], 0.16
all_vals = [a for b in vals for a in b]
try:
valMin, valMean, valMax, valMeans = min(all_vals), np.mean(
all_vals), max(all_vals) * 1.25, [np.mean(v) for v in vals]
except ValueError:
valMin, valMean, valMax, valMeans = 0, 0, 0, [0 for v in vals]
for i in range(len(vals)):
other_vals = [
a for b in [vals[k] for k in range(len(vals)) if k != i]
for a in b
]
pv = stats.ttest_ind(vals[i], other_vals)[-1]
if valMeans[i] > valMean and pv < global_sigs[i]:
global_sigs[i] = pv
jitters.append([[
self.pos[i] + np.random.normal(0, 0.02)
for m in range(len(vals[i]))
], [v * np.random.normal(1, 0.005) for v in vals[i]]])
for j in range(i + 1, len(vals)):
pv = stats.ttest_ind(vals[i], vals[j])[-1]
if valMeans[i] > valMeans[j] and pv < local_sigs[i]:
local_sigs[i] = pv
elif valMeans[i] < valMeans[j] and pv < local_sigs[j]:
local_sigs[j] = pv
self.major_legend_items[major_id] = Rect(
(0, 0), 1, 1, fc=major_clrs[0]) #,ec=clr,linewidth=2,hatch='+')
self.fill_boxes(opts,
vals,
minor_clrs,
major_clrs,
JITTERS=jitters,
SIG_STARS=(global_sigs, local_sigs))
if valMax > self.yTop: self.yTop = valMax
self.xOffset = self.pos[-1] + 2 * self.width
self.axes[-1].text(np.mean(self.pos),
0 - (self.yTop) * 0.1,
major_id.split(';')[-1],
fontsize=16,
fontweight='bold',
horizontalalignment='center',
verticalalignment='top')
def print_clusters(pp, clusters, tuples=[], title=''):
fig = plt.figure(figsize=(8, 8))
sub = fig.add_subplot(111)
clusters.sort(key=lambda c: c.error)
for cluster in clusters:
x, y = tuple(map(list, zip(*cluster.bbox)))[:2]
x[0] = max(0, x[0])
x[1] = min(100, x[1])
y[0] = max(0, y[0])
y[1] = min(100, y[1])
c = cm.jet(cluster.error)
r = Rect((x[0], y[0]), x[1]-x[0], y[1]-y[0], alpha=max(0.1,cluster.error), ec=c, fill=False, lw=1.5)
sub.add_patch(r)
if tuples:
cols = zip(*tuples)
xs, ys, cs = cols[0], cols[1], cols[-1]
sub.scatter(ys, xs, c=cs, alpha=0.5, lw=0)
sub.set_ylim(-5, 105)
sub.set_xlim(-5, 105)
sub.set_title(title)
plt.savefig(pp, format='pdf')
def __init__(self, f_data, f_key, options):
self.options = options
self.colors = [
'darkblue', 'g', 'brown', 'r', 'purple', 'gray', 'orange', 'k',
'cornflowerblue', 'magenta', 'cyan', 'lime', 'yellow', 'pink',
'blue'
]
self.tFS = 25
self.log = True
self.color_key = {
'PANTHER': 'darkblue',
'GORILLA': 'green',
'ELEPHANT': 'brown',
'HUMAN': 'red',
'POLAR': 'purple',
'DOLPHIN': 'gray',
'TIGER': 'orange',
'ORCA': 'k'
}
self.feats = []
self.vals = []
self.log_vals = []
self.f_name = f_data
for line in open(f_data):
line = line.split()
if line[0] == '---': self.samples = line[1::]
else:
self.feats.append(line[0])
self.vals.append([float(x) for x in line[1::]])
self.log_vals.append(
[math.log(v + 1.0) for v in self.vals[-1]])
self.key = {}
for line in open(f_key):
line = line.split()
if line[0] == '---': continue
self.key[line[0]] = line[1]
try:
self.sample_color = {
k: self.color_key[self.key[k]]
for k in self.key.keys()
}
except KeyError:
self.color_key = {}
for i, k in enumerate(list(set(self.key.values()))):
self.color_key[k] = self.colors[i]
self.sample_color = {
k: self.color_key[self.key[k]]
for k in self.key.keys()
}
self.ncol = len(list(set(self.key.values())))
self.labels, self.items = [], []
for xI, yI in self.color_key.items():
if xI == 'NA': continue
self.labels.append(xI)
self.items.append(Rect((0, 0), 1, 1, fc=yI))
def add_pv_bar(self, pvs, sim_pvs=[]):
self.set_pv_key(pvs)
bar_cnts = []
pvk = sorted(self.pv_key.keys())
for sp in [pvs] + sim_pvs:
sp.sort()
pcnts, k, p = [0 for i in pvk], 0, 0
while True:
while k < len(pvk) and sp[p] > pvk[k]:
k += 1
while k < len(pvk) and p < len(sp) and sp[p] <= pvk[k]:
for j in range(k, len(pvk)):
pcnts[j] += 1
p += 1
if k == len(pvk) or p == len(sp): break
bar_cnts.append(pcnts)
myticks, myticklabels = [], []
r_width = 0.2
s_width = (1.0 - r_width) / len(bar_cnts)
for i, k in enumerate(pvk[-1::-1]):
myticks.append(i)
myticklabels.append(' < ' + str(k))
idx = -1 - i
self.ax.bar(i, bar_cnts[0][idx], width=r_width, color='blue')
iLoc = i + r_width
bh = sorted([b[idx] for b in bar_cnts[1::]], reverse=True)
for b in bh:
self.ax.bar(iLoc, b, width=s_width, color='red')
iLoc += s_width
self.ax.set_xticks(myticks)
self.ax.set_xticklabels(myticklabels, horizontalalignment='left')
items = [Rect((0, 0), 1, 1, fc='b'), Rect((0, 0), 1, 1, fc='r')]
self.ax.legend(items, ['Results', 'Simulation'],
loc='upper right',
bbox_to_anchor=(1, 1),
fontsize=10)
return self
def add_legend(self, labels, colors):
labs, items = [], []
for a, b in zip(labels, colors):
labs.append(a)
items.append(Rect((0, 0), 1, 1, fc=b))
self.ax_key[(0, self.yLen - 1)].legend(items,
labs,
ncol=1,
loc='upper right',
bbox_to_anchor=(1.2, 1.0),
fontsize=10)
def add_legend(self, labels, colors):
labs, items = [], []
for a, b in zip(labels, colors):
labs.append(a)
items.append(Rect((0, 0), 1, 1, fc=b))
plt.legend(items,
labs,
loc='upper left',
ncol=len(labs) / 3,
bbox_to_anchor=(-0.1, 1.16),
fontsize=10)
def visualize_detection(ax, target):
boxes = target['boxes']
labels = target['labels']
colors = plt.cm.tab20(np.linspace(0, 1, 20))
colors = colors[labels % 20]
colors = [mpl.colors.rgb2hex(c[:3]) for c in colors]
for (x0, y0, x1, y1), color in zip(boxes, colors):
x, y, w, h = x0, y0, x1 - x0, y1 - y0
ax.add_patch(Rect((x, y), w, h, ec=color, fc='none', lw=2))
return ax
def add_rs_bars(self, ax, p_key, pvs, sim_pvs=[]):
r_width = 0.2
s_width = (1.0 - r_width) / len(sim_pvs)
myticks, myticklabels = [], []
for i, k in enumerate(p_key):
myticks.append(i)
myticklabels.append(' >= ' + str(k))
ax.bar(i, pvs[i], width=r_width, color='blue')
iLoc = i + r_width
for b in sorted([bh[i] for bh in sim_pvs], reverse=True):
ax.bar(iLoc, b, width=s_width, color='red')
iLoc += s_width
ax.set_xticks(myticks)
ax.set_xticklabels(myticklabels, horizontalalignment='left')
ax.set_title('Variance Explained')
items = [Rect((0, 0), 1, 1, fc='b'), Rect((0, 0), 1, 1, fc='r')]
ax.legend(items, ['Results', 'Simulation'],
loc='upper right',
bbox_to_anchor=(1, 1),
fontsize=10)
return self
def prepare_labels(self):
try:
self.sample_color = {k: self.color_key[self.sample_key[k]] for k in self.sample_key.keys()}
except KeyError:
self.color_key = {}
for i,k in enumerate(list(set(self.key.values()))):
self.color_key[k] = self.colors[i]
self.sample_color = {k: self.color_key[self.sample_key[k]] for k in self.sample_key.keys()}
self.ncol = len(list(set(self.key.values())))
self.labels,self.items = [],[]
for xI,yI in self.color_key.items():
if xI == 'NA': continue
self.labels.append(xI)
self.items.append(Rect((0,0),1,1,fc=yI))
def sample_data(self, num, randvar=None):
self.num = num
SAMPLES = num
animals = {}
for x in [
'HUMANS', 'ELEPHANTS', 'GORILLAS', 'PANTHERS', 'POLAR_BEARS',
'ORCAS', 'DOLPHINS', 'TIGERS'
]:
if randvar:
myNum = int(np.random.normal(num, num / 8.0))
if myNum < 10: myNum = 10
animals[x] = [
sample_creature(self.stats[x], self.er, self.dp,
self.noise) for i in range(myNum)
]
else:
animals[x] = [
sample_creature(self.stats[x], self.er, self.dp,
self.noise) for i in range(SAMPLES)
]
self.key = animals
self.labels, self.items = [], []
for xI, yI in self.color_key.items():
if len(self.key[xI]) < 100:
self.labels.append(xI + ' (' + str(len(self.key[xI])) + ' )')
else:
self.labels.append(xI + ' (' + str(len(self.key[xI])) + ')')
self.items.append(Rect((0, 0), 1, 1, fc=yI))
self.merge_data()
def plot_clusters(self, clusters, cols=None, color=None, alpha=None):
self.clusters = clusters
errors = [c.error for c in clusters]
errors = np.array(errors)
mean, std = np.mean(errors), np.std(errors)
if std == 0: std = 1
errors = (errors - mean) / std
if not cols:
cols = [0, 1]
if isinstance(cols[0], basestring):
cols = map(clusters.cols.index, cols)
for idx, cluster in enumerate(clusters):
tup = tuple(map(list, zip(*cluster.bbox)))
x, y = tup[cols[0]], tup[cols[1]]
x, y = self.transform_box(x, y)
if not self.xbound:
self.xbound = [x[0], x[1]]
self.ybound = [y[0], y[1]]
else:
self.xbound = r_union(self.xbound, x)
self.ybound = r_union(self.ybound, y)
a = alpha or min(1, max(0.1, errors[idx]))
c = color or cm.jet(errors[idx])
r = Rect((x[0], y[0]),
x[1] - x[0],
y[1] - y[0],
alpha=a,
ec=c,
fill=False,
lw=1.5)
self.sub.add_patch(r)
self.set_lims()
def recurse_labels(ax,
mtree,
xlims,
ymax,
all_size,
cur_j=0,
clr_mtype=False,
add_lbls=True,
mut_clr=None):
all_wdth = len(MuType(mtree.allkey()).leaves())
cur_x = xlims[0]
muts_dict = dict(mtree)
if mut_clr is None:
mut_clr = variant_clrs['Point']
for lbl in sort_levels(list(muts_dict.keys()), mtree.mut_level):
muts = muts_dict[lbl]
if isinstance(muts, MuTree):
lf_wdth = len(MuType(muts.allkey()).leaves())
else:
lf_wdth = 1
lbl_prop = (lf_wdth / all_wdth) * (xlims[1] - xlims[0])
ax.plot([cur_x + lbl_prop / 2, (xlims[0] + xlims[1]) / 2],
[ymax - 0.18 - cur_j, ymax + 0.19 - cur_j],
c='black',
linewidth=0.9,
solid_capstyle='round')
if add_lbls:
mut_lbl = clean_label(lbl, mtree.mut_level)
if (lbl_prop / all_size) > 1 / 41 and len(tuple(mtree)) > 1:
if len(muts) == 1:
mut_lbl = "{}\n(1 sample)".format(mut_lbl)
else:
mut_lbl = "{}\n({} samps)".format(mut_lbl, len(muts))
if (lbl_prop / all_size) <= 1 / 19:
use_rot = 90
else:
use_rot = 0
ax.text(cur_x + lbl_prop / 2,
ymax - 0.5 - cur_j,
mut_lbl,
size=14 - 2.9 * cur_j,
ha='center',
va='center',
rotation=use_rot)
# colour this branch and all subtypes
if clr_mtype is None or clr_mtype is False:
use_clr = mut_clr
eg_clr = mut_clr
sub_mtype = clr_mtype
# do not colour this branch or any of its subtypes
elif clr_mtype.is_empty():
use_clr = variant_clrs['WT']
eg_clr = variant_clrs['WT']
sub_mtype = clr_mtype
# otherwise, check to see which subtypes need to be coloured
else:
sub_dict = dict(clr_mtype.subtype_iter())
if lbl not in sub_dict:
use_clr = variant_clrs['WT']
eg_clr = variant_clrs['WT']
sub_mtype = MuType({})
elif (isinstance(mtree[lbl], dict) or sub_dict[lbl] is None
or (sub_dict[lbl] == MuType(mtree[lbl].allkey()))):
use_clr = mut_clr
eg_clr = mut_clr
sub_mtype = None
else:
use_clr = 'none'
eg_clr = mut_clr
sub_mtype = sub_dict[lbl]
if clr_mtype is False:
sub_lbls = add_lbls
else:
sub_lbls = False
ax.add_patch(
Rect((cur_x + lbl_prop * 0.12, ymax - 0.8 - cur_j),
lbl_prop * 0.76,
0.6,
facecolor=use_clr,
edgecolor=eg_clr,
alpha=0.41,
linewidth=1.3))
if isinstance(muts, MuTree):
ax = recurse_labels(
ax, muts, (cur_x + lbl_prop * 0.06, cur_x + lbl_prop * 0.94),
ymax, all_size, cur_j + 1, sub_mtype, sub_lbls, mut_clr)
cur_x += lbl_prop
return ax
def plot_custom_pts(self,
pts,
key,
cnts=None,
cnt_idx=0,
genes=[],
neg_genes=[],
color=None,
VERSION='LOC',
TITLE='HI',
BAR_MSG=None,
AXES=[]):
if len(AXES) == 0:
self.axes.append(
plt.subplot2grid((self.xLen, self.yLen),
(self.xLoc, self.yLoc),
rowspan=1,
colspan=1))
else:
xl, yl, rs, cs = AXES
self.axes.append(
plt.subplot2grid((self.xLen, self.yLen), (xl, yl),
rowspan=rs,
colspan=cs))
cell_key = CellKey(key, self.options)
p_data, p_genes, p_names, label_key = [], [], [], {}
if len(genes) > 0:
TITLE = ",".join(genes)
if len(neg_genes) > 0:
TITLE += '\n' + ",".join(neg_genes)
genes += neg_genes
p_genes = [[] for p in genes]
for s in pts:
x, y = pts[s]
try:
cell = cell_key.identify(s)
except KeyError:
continue
if cell.type == 'NA': continue
if VERSION == 'FIRING':
clr, clr_label = cell.fire_color, cell.fs
else:
clr, clr_label = cell.type_color, cell.type
if cell.type == 'MINI': cell.type = 'canonical'
#print s,x,y,cell.type, cell.type2, cell.marker
#if cell.type == 'MZ' and cell.type2 == 'MZ': continue
if clr_label not in label_key:
self.color_key[clr_label] = clr
label_key[clr_label] = Rect((0, 0), 1, 1, fc=clr)
if cnts == None:
if cell.marker == 'o': continue
if VERSION == 'FIRING':
self.axes[-1].scatter(x,
y,
marker=cell.marker,
c=cell.fire_color,
s=70,
alpha=1)
else:
self.axes[-1].scatter(x,
y,
marker=cell.marker,
c=clr,
s=70,
alpha=1)
else:
for j in range(len(genes)):
p_genes[j].append(log(cnts[genes[j]][s] + 1, 2))
#print j,p_genes[j],cnts[genes[j]]
p_data.append([
np.array([log(cnts[g][s] + 1, 2) for g in genes]), x, y,
clr, cell
])
if cnts != None:
#print len(p_data)
#print len(p_data[0])
#print p_data[0]
#sys.exit()
if len(genes) > 0:
scaler = MinMaxScaler()
t_data = scaler.fit_transform(np.array([p[0] for p in p_data]))
for j in range(len(p_data)):
pos_cont = sum([
t_data[j][x] for x in range(len(genes))
if genes[x] not in neg_genes
])
neg_cont = sum([
t_data[j][x] for x in range(len(genes))
if genes[x] in neg_genes
])
#for x in range(len(genes)):
# print x,genes[x],t_data[j][x],p_data[j][0]
p_data[j][0] = pos_cont - neg_cont
p_data.sort(reverse=True)
p_data.sort() #reverse=True)
X = np.array([p[1] for p in p_data])
Y = np.array([p[2] for p in p_data])
I = np.array([p[0] for p in p_data])
#print iI
#z = gaussian_kde(I)(I)
Xs = scaler.fit_transform(X.reshape(-1, 1))
Ys = scaler.fit_transform(Y.reshape(-1, 1))
scatterplot = self.axes[-1].scatter(X,
Y,
c=I,
cmap='seismic',
s=40,
edgecolor='')
plt.colorbar(scatterplot,
ax=self.axes[-1],
orientation='horizontal',
fraction=0.04,
pad=0.1,
ticks=[])
self.axes[-1].scatter(X,
Y,
c=I,
cmap='seismic',
s=40,
edgecolor='')
if BAR_MSG == 'RELN':
self.axes[-1].text(0.6,
-7.5,
' CR Markers\n(RELN,CXCR4,PCP4)',
fontsize=11,
verticalalignment='bottom')
elif BAR_MSG[0:3] == 'CHL':
self.axes[-1].text(0.3,
-7.5,
' Choride Ratio\nSLC12A2:SLC12A5',
fontsize=11,
verticalalignment='bottom')
elif BAR_MSG[0:3] == 'GLU':
self.axes[-1].text(
0.0,
-7.5,
' Glutamatergic:GABA \n(SLC17A7,SLC1A2,GRIK3:ABAT,GAD)',
fontsize=9,
verticalalignment='bottom')
elif BAR_MSG[0:3] == 'VEN':
self.axes[-1].text(0.3,
-7.5,
' VEN Markers\n(FEZF2,BCL11B)',
fontsize=11,
verticalalignment='bottom')
else:
self.axes[-1].text(0.3,
-7.5,
BAR_MSG,
fontsize=11,
verticalalignment='bottom')
#self.axes[-1].scatter(X,Y,c=z,cmap='jet',s=50,edgecolor='')
#self.axes[-1].scatter(X,Y,c=C,s=100,alpha=0.5)
#self.axes[-1].scatter(X,Y,c=I,cmap='hot',s=50,edgecolor='')
#labels = label_key.keys()
#items = [label_key[x] for x in labels]
#nc,fs,cs,hp,bb1,bb2 = 2,12 ,0.5,0.5,-0.01,1.1
#leg = self.axes[-1].legend(items,labels,title=TITLE,handletextpad=hp,columnspacing=cs, fontsize=fs,ncol=nc,bbox_to_anchor=(bb1,bb2),loc='upper center')
else:
nc, fs, cs, hp, bb1, bb2 = 2, 9.0, 0.62, 0.62, 0.465, 1.0
PROX = [
Line2D([], [],
color=self.color_key[k],
marker='s',
mec='k',
markeredgewidth=0.2,
linestyle='None',
markersize=10,
label=k) for k in label_key if k != 'NA'
]
leg = self.axes[-1].legend(handles=PROX + self.loc_labs,
title=TITLE,
handletextpad=hp,
columnspacing=cs,
fontsize=fs,
ncol=nc,
bbox_to_anchor=(bb1, bb2),
loc='upper center')
self.axes[-1].axis('off')
self.yLoc += 1
if self.yLoc == self.yLen:
self.yLoc = 0
self.xLoc += 1
def add_multi_box(self, h_data, TITLE, BW=False):
iKey = dd(int)
data = sorted([[iKey[d[0]], d[0], d[1]] for d in h_data.items()])
data = [
opt for opt in data
if data[1] not in ['NA', 'UNK', 'UNKNOWN', 'UNCLEAR', 'NON']
]
opts, vals = [], []
for d in data:
d[2] = [float(dx) for dx in d[2] if dx != 'NA']
if d[1] in [
'NA', 'UNK', 'UNKNOWN', 'UNAVAIL', 'UNCLEAR', 'DOUBLETS'
]:
continue
elif d[1] == 'GT_18':
opts.append('>18')
elif d[1] == 'SUB_16':
opts.append('<16')
elif len(d[1].split('_')) > 1:
continue
else:
opts.append(d[1])
vals.append(d[2])
colors = [self.get_color(X) for X in opts]
colors = ['pink', 'orange', 'purple', 'orange']
if BW: colors = ['k' for X in opts]
if opts == ['SVZ', '*SVZ*', 'IZ', '*IZ*']:
opts = ['$SVZ$', '$SVZ_{novel}$', '$IZ$', '$IZ_{novel}$']
if opts == ['CRMZ', 'MEGA', 'MINI']:
opts = ['CAJAL-RETZIUS', 'NOVEL', 'CANONICAL']
items, labels = [], []
if type(colors[0]) == tuple:
colors, alt_colors = [c[0] for c in colors], [c[1] for c in colors]
else:
alt_colors = [c for c in colors]
R, pv = 'NA', 'NA'
pos, width = [self.xOffset + 0.20 * x for x in range(len(opts))], 0.16
self.bp = self.ax.boxplot(vals,
positions=pos,
widths=width,
patch_artist=True,
showmeans=True,
whis=1)
clevels = np.linspace(0., 3., len(opts) + 1)
self.xOffset = pos[-1] + 2 * width
xMid = np.mean(pos)
xticks = []
means = sorted([(np.mean(v), opt) for (v, opt) in zip(vals, opts)
if opt[0:2] == 'si'])
medians = sorted([(np.percentile(v, 60), opt)
for (v, opt) in zip(vals, opts) if opt[0:2] == 'si'])
sorts = [(sorted(v, reverse=True), opt)
for (v, opt) in zip(vals, opts) if opt[0:2] == 'si']
for i, opt in enumerate(opts):
clr, ac = colors[i], alt_colors[i]
self.bp['boxes'][i].set_edgecolor(clr)
self.bp['boxes'][i].set_linewidth(1)
plt.setp(self.bp['medians'][i], color=clr, linewidth=3)
plt.setp(self.bp['means'][i],
marker='h',
markersize=9,
markerfacecolor=clr)
plt.setp(self.bp['caps'][(i * 2) + 1], color=clr, linewidth=1)
plt.setp(self.bp['caps'][(i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][i * 2], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][1 + (i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['fliers'][i],
markerfacecolor=clr,
markeredgecolor=clr,
marker='s',
markersize=2.0)
if clr == ac: self.items[opt] = Rect((0, 0), 1, 1, fc=clr)
else: self.items[opt] = Rect((0, 0), 1, 1, fc=a, ec=b, hatch='*')
clevels = np.linspace(0., 1., len(vals))
xJitters = [
np.random.normal(pos[i], 0.01, len(vals[i]))
for i in range(len(vals))
]
yJitters = [[
vals[i][j] * np.random.normal(1, 0.005, len(vals[i]))[j]
for j in range(len(vals[i]))
] for i in range(len(vals))]
for xJ, yJ, clevel in zip(xJitters, yJitters, colors):
plt.scatter(xJ, yJ, c=clevel, alpha=0.7, s=4, zorder=9)
for patch, clevel in zip(self.bp['boxes'], colors):
patch.set_facecolor(clevel) #cm.prism(clevel))
patch.set_alpha(0.5)
yMin, yMax = self.ax.get_ylim()
yStep = (yMax - yMin) / 30.0
# for j,xpos in enumerate(self.ax.get_xticks()):
# self.ax.text(xpos,yMin-yStep,opts[j].split('~')[-1],fontweight='bold',rotation=60,verticalalignment='center',horizontalalignment='center')
#self.ax.text(xpos,yMin-yStep,opts[j].split('~')[-1],fontsize=25,fontweight='bold',verticalalignment='bottom',horizontalalignment='center')
# self.ax.axis('off')
title_key = {'LOC': 'LAYER', 'FS': 'FIRING STYLE'}
if TITLE:
self.ax.text(xMid,
yMax,
TITLE.split(';')[-1],
fontsize=20,
fontweight='bold')
return
if TITLE:
if TITLE in title_key: TITLE = title_key[TITLE]
def EEG_Comb(t,
data,
names,
network,
t0=None,
t1=None,
amplitude=0.5,
iis=None,
**spec):
## t0, t1 unit: second if is type float
import Utility as utl
from bisect import bisect
if t0 is None:
t0i = 0
t0 = t[0]
else:
t0i = bisect(t, float(t0))
if t1 is None:
t1i = len(t)
t1 = t[-1]
else:
t1i = bisect(t, float(t1))
t_plot = utl.DataSegment(t, t0i, t1i)
d_plot = utl.DataSegment(data, t0i, t1i)
t0_nest = utl.Time(seconds=float(t0) - t[0])
t1_nest = utl.Time(seconds=float(t1) - t[0])
clks, phase = network.get_WSN_spike_phases(t0=t0_nest,
t1=t1_nest,
group=True,
inverse=False)
clks_sec = clks / 1000.0 + t[0]
#N_ch,N_sig,N_delay,N_clk = phase.shape
#N_p_ch = N_sig * N_delay
#y_values = np.arange(0, N_ch, 1.0/N_sig, dtype=float)[::-1]
#phase = np.reshape(phase, (N_ch*N_sig, N_delay, N_clk))
#np.place(phase,np.isnan(phase),np.inf)
#phase = np.amin(phase, -2)
#np.place(phase,np.isinf(phase),np.nan)
N_ch, N_delay, N_clk = phase.shape
y_values = np.arange(0, N_ch, 1.0 / N_delay, dtype=float)[::-1]
y_ticks = np.arange(0, N_ch, dtype=float) + 0.5
phase = np.reshape(phase, (N_ch * N_delay, N_clk))
if 'font' in spec:
font = spec['font']
else:
font = default_plot_font
if 'figsize' in spec:
figsize = spec['figsize']
else:
figsize = [3.45 * 1.5, 5.5 * 1.5]
if 'dpi' in spec:
dpi = spec['dpi']
else:
dpi = 100
if 'plt_pos' in spec:
plt_pos = spec['plt_pos']
else:
plt_pos = {
'left': 0.1,
'right': 0.8,
'bottom': 0.1,
'top': 0.9,
'hspace': 0.2
}
if 'cbar_pos' in spec:
cbar_pos = spec['cbar_pos']
else:
cbar_pos = [0.85, 0.1, 0.03, 0.35]
if 'c_ticks' in spec:
c_ticks = spec['c_ticks']
else:
c_ticks = np.arange(0, 100.1, 20)
import matplotlib
matplotlib.rc('font', **font)
import matplotlib.pyplot as plt
fig, axes = plt.subplots(nrows=2, ncols=1, figsize=figsize, dpi=dpi)
PlotEEG(t_plot,
d_plot,
channels=names,
amplitude=amplitude,
figaxes=axes[0])
pShapes = pcolor(clks_sec,
y_values,
phase,
x_label='',
x_range=[float(t0), float(t1)],
y_label='',
y_ticks=y_ticks,
y_ticklabels=names[::-1],
Ax=axes[1])
if iis is not None:
from matplotlib.patches import Rectangle as Rect
for iis_t0, iis_t1 in iis:
if (iis_t0 >= t0 and iis_t0 <= t1) or (iis_t1 >= t0
and iis_t1 <= t1):
rec_width = float(iis_t1) - float(iis_t0)
for ax in axes:
ylim = ax.get_ylim()
rec_loc = (float(iis_t0), ylim[0])
rec_height = ylim[1] - ylim[0]
ax.add_patch(
Rect(rec_loc,
rec_width,
rec_height,
edgecolor='none',
facecolor='red',
alpha=0.3))
from Utility import Time
def second_to_timestr(seconds, loc):
tmptime = Time(seconds)
return repr(tmptime)
from matplotlib.ticker import FuncFormatter
axes[1].get_xaxis().set_major_formatter(FuncFormatter(second_to_timestr))
fig.subplots_adjust(**plt_pos)
cbar_ax = fig.add_axes(cbar_pos)
cb = fig.colorbar(pShapes, cax=cbar_ax)
cb.set_ticks(c_ticks)
plt.show()
return fig
def add_data(self, gkey, ax, KEY=False):
opts = gkey.keys()
vals = gkey.values()
colors = [
'DarkSalmon', 'DarkSlateGray', 'DeepPink', 'DarkTurquoise',
'ForestGreen', 'Crimson', 'Plum', 'Chocolate', 'FireBrick'
]
colors = VWCOLORS
axillary = [
'DarkSalmon', 'DarkSlateGray', 'DeepPink', 'DarkTurquoise',
'ForestGreen', 'Crimson', 'Plum', 'Chocolate', 'FireBrick'
]
if KEY:
if len(opts) == 2: colors = ['magenta', 'lime']
elif len(opts) == 3: colors = ['green', 'cyan', 'red']
else:
colors = [
'olive', 'DarkSalmon', 'DarkSlateGray', 'Plum',
'Chocolate', 'FireBrick'
] + VWCOLORS
try:
color_key = {opts[j]: colors[j] for j in range(len(opts))}
except IndexError:
print "okay this is error", opts, len(opts), len(colors)
sys.exit()
vals = [gkey[opt] if opt in gkey else [] for opt in opts]
colors = [color_key[opt] if opt in color_key else 'k' for opt in opts]
labels, items = [], []
for v, c in color_key.items():
labels.append(v)
items.append(Rect((0, 0), 1, 1, fc=c))
pos, width = [self.xOffset + 0.10 * x for x in range(len(opts))], 0.075
self.bp = ax.boxplot(vals,
positions=pos,
widths=width,
patch_artist=True,
showmeans=True,
whis=1)
clevels = np.linspace(0., 3., len(opts) + 1)
xticks = []
means = sorted([(np.mean(v), opt) for (v, opt) in zip(vals, opts)
if opt[0:2] == 'si'])
medians = sorted([(np.percentile(v, 60), opt)
for (v, opt) in zip(vals, opts) if opt[0:2] == 'si'])
sorts = [(sorted(v, reverse=True), opt)
for (v, opt) in zip(vals, opts) if opt[0:2] == 'si']
for i, opt in enumerate(opts):
if opt == '': continue
clr = color_key[opt]
self.bp['boxes'][i].set_edgecolor(clr)
self.bp['boxes'][i].set_linewidth(1)
plt.setp(self.bp['medians'][i], color=clr, linewidth=3)
plt.setp(self.bp['means'][i],
marker='h',
markersize=9,
markerfacecolor=clr)
plt.setp(self.bp['caps'][(i * 2) + 1], color=clr, linewidth=1)
plt.setp(self.bp['caps'][(i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][i * 2], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][1 + (i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['fliers'][i],
markerfacecolor=clr,
markeredgecolor=clr,
marker='s',
markersize=2.0)
clevels = np.linspace(0., 1., len(vals))
xJitters = [
np.random.normal(pos[i], 0.01, len(vals[i]))
for i in range(len(vals))
]
yJitters = [[
vals[i][j] * np.random.normal(1, 0.005, len(vals[i]))[j]
for j in range(len(vals[i]))
] for i in range(len(vals))]
for xJ, yJ, clevel in zip(xJitters, yJitters, colors):
ax.scatter(xJ, yJ, c=clevel, alpha=0.7, s=4, zorder=9)
for patch, clevel in zip(self.bp['boxes'], colors):
patch.set_facecolor(clevel) #cm.prism(clevel))
patch.set_alpha(0.2)
ax.set_xticklabels([opt.split('~')[-1] for opt in opts],
fontsize=10,
rotation=-30)
ax.set_xlim([pos[0] - 0.2, pos[-1] + 0.2])
return labels, items
def add_hist(self, h_data, TITLE=None, DNAME=None, MINSIZE=30):
self.ax = plt.subplot2grid((self.xLen, self.yLen),
(self.xLoc, self.yLoc),
rowspan=1,
colspan=1)
if type(h_data) == list:
y = sorted(h_data)
if len(self.ax.hist(y, bins='auto')[0]) < 2:
self.ax.clear()
self.ax.hist(np.hstack(y), bins=min(4, len(cc(y))))
else:
x, y = h_data.keys(), sorted(h_data.values())
X_key = dd(lambda: dd(int))
R, pv = 'NA', 'NA'
yOffset = 0
xOffset = 0
xLoc = 0
xticks = []
if DNAME == 'REL':
for X, Y in h_data.items():
if len(Y) < MINSIZE: continue
for y in Y:
yAdd = int(y)
if yAdd < -5: yAdd = -5
if yAdd > 5: yAdd = 5
X_key[X][yAdd] += 1
Y_key = dd(lambda: dd(int))
for xx in X_key:
xsum = sum(X_key[xx].values())
for yy in X_key[xx]:
Y_key[xx][yy] = X_key[xx][yy] / float(xsum)
try:
my_means = []
my_x = [int(xx) for xx in X_key.keys()]
for xx in X_key.keys():
x_len = sum(X_key[xx].values())
x_total = sum(
[int(a) * b for a, b in X_key[xx].items()])
my_means.append(x_total / x_len)
R, pv = stats.pearsonr(my_x, my_means)
except ValueError:
my_ex = []
X_key = Y_key
xJump = len(X_key)
hRange = [-3, -2, -1, 0, 1, 2, 3, 4, 5]
items, labels = [], []
color_idx = 0
for X in X_key:
xLoc = 0
labels.append(X)
clr = self.get_color(X)
if type(clr) == tuple:
a, b = clr
items.append(Rect((0, 0), 1, 1, fc=a, ec=b, hatch='*'))
for s in hRange:
self.ax.bar(xLoc + xOffset,
X_key[X][s],
width=1,
bottom=0,
color=a,
ec=b,
hatch='*')
if xOffset == 0:
xticks.append((xLoc + (xJump / 2), s))
xLoc += xJump
xOffset += 1
else:
items.append(Rect((0, 0), 1, 1, fc=clr))
for s in hRange:
self.ax.bar(xLoc + xOffset,
X_key[X][s],
width=1,
bottom=0,
color=clr)
if xOffset == 0:
xticks.append((xLoc + (xJump / 2), s))
xLoc += xJump
xOffset += 1
else:
for X, Y in h_data.items():
if len(Y) < MINSIZE: continue
for y in Y:
yAdd = y % 100
if yAdd >= 50: yAdd = 50
else: yAdd = 0
yRnd = (int(y / 100) * 100) + yAdd
if yRnd > 1000: yRnd = 1000
X_key[X][yRnd] += 1
Y_key = dd(lambda: dd(int))
for xx in X_key:
xsum = sum(X_key[xx].values())
for yy in X_key[xx]:
Y_key[xx][yy] = X_key[xx][yy] / float(xsum)
try:
my_means = []
my_x = [int(xx) for xx in X_key.keys()]
for xx in X_key.keys():
x_len = sum(X_key[xx].values())
x_total = sum(
[int(a) * b for a, b in X_key[xx].items()])
my_means.append(x_total / x_len)
R, pv = stats.pearsonr(my_x, my_means)
except ValueError:
my_ex = []
X_key = Y_key
xJump = len(X_key)
hRange = range(0, 1000, 50)
items, labels = [], []
color_idx = 0
for X in X_key:
xLoc = 0
labels.append(X)
clr = self.get_color(X)
if type(clr) == tuple:
a, b = clr
items.append(Rect((0, 0), 1, 1, fc=a, ec=b, hatch='*'))
for s in hRange:
self.ax.bar(xLoc + xOffset,
X_key[X][s],
width=1,
bottom=0,
color=a,
ec=b,
hatch='*')
if xOffset == 0:
xticks.append((xLoc + (xJump / 2), s))
xLoc += xJump
xOffset += 1
else:
items.append(Rect((0, 0), 1, 1, fc=clr))
for s in hRange:
self.ax.bar(xLoc + xOffset,
X_key[X][s],
width=1,
bottom=0,
color=clr)
if xOffset == 0:
xticks.append((xLoc + (xJump / 2), s))
xLoc += xJump
xOffset += 1
xt, xl = [xx[0] for xx in xticks], [str(xx[1]) for xx in xticks]
self.ax.set_xticks(xt)
self.ax.set_xticklabels(xl)
if R != 'NA':
rSS = str(round(R, 5))
pSS = str(round(pv, 8))
self.ax.set_title('CORR: ' + rSS + ' pv= ' + pSS)
if TITLE:
nc, fs, cs, hp, bb1, bb2 = 1, 12, 0.5, 0.5, 0.9, 1.1
if len(items) > 5:
if len(items) < 10:
nc, fs, cs, hp = 2, 10, 0.4, 0.4
elif len(items) < 15:
nc, fs, cs, hp, bb1, bb2 = 3, 10, 0.35, 0.35, 0.85, 1.1
elif len(items) < 25:
nc, fs, cs, hp, bb1, bb2 = 4, 9, 0.25, 0.25, 0.85, 1.1
elif len(items) < 35:
nc, fs, cs, hp, bb1, bb2 = 5, 8, 0.20, 0.20, 0.8, 1.1
else:
nc, fs, cs, hp, bb1, bb2 = 6, 7, 0.15, 0.15, 0.75, 1.1
leg = self.ax.legend(items,
labels,
title=TITLE,
handletextpad=hp,
columnspacing=cs,
fontsize=fs,
ncol=nc,
bbox_to_anchor=(bb1, bb2),
loc='upper center')
self.yLoc += 1
if self.yLoc == self.yLen:
self.xLoc += 1
self.yLoc = 0
def add_iso_boxes(self,box_data,TITLE=None,BW=False):
opts,counts,vals = [], [],[]
for obs,avg,name,cnts in box_data:
opts.append(name)
counts.append(cnts)
vals.append([log(c+1,2) for c in cnts])
colors = [self.get_color(X) for X in opts]
# colors = ['pink','orange','purple','orange']
if BW: colors = ['k' for X in opts]
#if opts == ['SVZ', '*SVZ*', 'IZ', '*IZ*']: opts = ['$SVZ$','$SVZ_{novel}$','$IZ$','$IZ_{novel}$']
items,labels = [],[]
if type(colors[0]) == tuple:
colors,alt_colors = [c[0] for c in colors], [c[1] for c in colors]
else:
alt_colors = [c for c in colors]
R,pv = 'NA','NA'
pos, width = [self.xOffset+0.20*x for x in range(len(opts))], 0.16
self.bp = self.ax.boxplot(vals,positions=pos, widths=width, patch_artist=True, showmeans=True,whis=1)
clevels = np.linspace(0., 3., len(opts)+1)
self.xOffset = pos[-1]+2*width
xMid = np.mean(pos)
xticks = []
means = sorted([(np.mean(v),opt) for (v,opt) in zip(vals,opts) if opt[0:2] == 'si'])
medians = sorted([(np.percentile(v,60),opt) for (v,opt) in zip(vals,opts) if opt[0:2] == 'si'])
sorts = [(sorted(v,reverse=True),opt) for (v,opt) in zip(vals,opts) if opt[0:2] == 'si']
for i,opt in enumerate(opts):
clr,ac = colors[i], alt_colors[i]
self.bp['boxes'][i].set_edgecolor(clr)
self.bp['boxes'][i].set_linewidth(1)
plt.setp(self.bp['medians'][i], color=clr, linewidth=3)
plt.setp(self.bp['means'][i], marker='h',markersize=9,markerfacecolor=clr)
plt.setp(self.bp['caps'][(i*2)+1], color=clr,linewidth=1)
plt.setp(self.bp['caps'][(i*2)], color=clr,linewidth=1)
plt.setp(self.bp['whiskers'][i*2], color=clr,linewidth=1)
plt.setp(self.bp['whiskers'][1+(i*2)], color=clr,linewidth=1)
plt.setp(self.bp['fliers'][i], markerfacecolor=clr, markeredgecolor = clr, marker='s',markersize=2.0)
if clr == ac: self.items[opt] = Rect((0,0),1,1,fc=clr)
else: self.items[opt] = Rect((0,0),1,1,fc=a,ec=b,hatch='*')
clevels = np.linspace(0., 1., len(vals))
xJitters = [np.random.normal(pos[i],0.01,len(vals[i])) for i in range(len(vals))]
yJitters = [[vals[i][j]*np.random.normal(1,0.005,len(vals[i]))[j] for j in range(len(vals[i]))] for i in range(len(vals))]
for xJ, yJ, clevel in zip(xJitters, yJitters, colors):
plt.scatter(xJ, yJ, c=clevel, alpha=0.7,s=4,zorder=9)
for patch, clevel in zip(self.bp['boxes'], colors):
patch.set_facecolor(clevel) #cm.prism(clevel))
patch.set_alpha(0.5)
yMin,yMax = self.ax.get_ylim()
yStep = (yMax - yMin)/30.0
# for j,xpos in enumerate(self.ax.get_xticks()):
# self.ax.text(xpos,yMin-yStep,opts[j].split('~')[-1],fontweight='bold',rotation=60,verticalalignment='center',horizontalalignment='center')
#self.ax.text(xpos,yMin-yStep,opts[j].split('~')[-1],fontsize=25,fontweight='bold',verticalalignment='bottom',horizontalalignment='center')
# self.ax.set_xticks([])
if TITLE:
self.ax.text(xMid,yMax,TITLE.split(';')[-1],fontsize=20,fontweight='bold')
return
def add_boxes(self, h_data, TITLE=None, DNAME=None, MINSIZE=30, BW=True):
self.ax = plt.subplot2grid((self.xLen, self.yLen),
(self.xLoc, self.yLoc),
rowspan=1,
colspan=1)
self.xOffset = 0
#clr = self.get_color(X)
iKey = dd(int)
iKey['16'], iKey['17'], iKey['18'], iKey['GT_18'] = 1, 2, 3, 4
iKey['IZ'], iKey['SP'], iKey['CP'], iKey['CR'], iKey[
'MZ'] = 1, 2, 3, 4, 5
iKey['SINGLE'], iKey['BRIEF'], iKey['REPETITIVE'] = 1, 2, 3
iKey['*SVZ*'], iKey['IZ'], iKey['*IZ*'] = 0.5, 1, 2
data = sorted([[iKey[d[0]], d[0], d[1]] for d in h_data.items()])
data = [
opt for opt in data
if data[1] not in ['NA', 'UNK', 'UNKNOWN', 'UNCLEAR', 'NON']
]
opts, vals = [], []
for d in data:
d[2] = [float(dx) for dx in d[2] if dx != 'NA']
if d[1] in [
'NA', 'UNK', 'UNKNOWN', 'UNAVAIL', 'UNCLEAR', 'DOUBLETS'
]:
continue
elif d[1] == 'GT_18':
opts.append('>18')
elif d[1] == 'SUB_16':
opts.append('<16')
elif len(d[1].split('_')) > 1:
continue
else:
opts.append(d[1])
vals.append(d[2])
colors = [self.get_color(X) for X in opts]
colors = ['pink', 'orange', 'purple', 'orange']
if BW: colors = ['k' for X in opts]
if opts == ['SVZ', '*SVZ*', 'IZ', '*IZ*']:
opts = ['$SVZ$', '$SVZ_{novel}$', '$IZ$', '$IZ_{novel}$']
items, labels = [], []
if type(colors[0]) == tuple:
colors, alt_colors = [c[0] for c in colors], [c[1] for c in colors]
else:
alt_colors = [c for c in colors]
R, pv = 'NA', 'NA'
pos, width = [self.xOffset + 0.20 * x for x in range(len(opts))], 0.16
self.bp = self.ax.boxplot(vals,
positions=pos,
widths=width,
patch_artist=True,
showmeans=True,
whis=1)
clevels = np.linspace(0., 3., len(opts) + 1)
xticks = []
means = sorted([(np.mean(v), opt) for (v, opt) in zip(vals, opts)
if opt[0:2] == 'si'])
medians = sorted([(np.percentile(v, 60), opt)
for (v, opt) in zip(vals, opts) if opt[0:2] == 'si'])
sorts = [(sorted(v, reverse=True), opt)
for (v, opt) in zip(vals, opts) if opt[0:2] == 'si']
for i, opt in enumerate(opts):
clr, ac = colors[i], alt_colors[i]
self.bp['boxes'][i].set_edgecolor(clr)
self.bp['boxes'][i].set_linewidth(1)
plt.setp(self.bp['medians'][i], color=clr, linewidth=3)
plt.setp(self.bp['means'][i],
marker='h',
markersize=9,
markerfacecolor=clr)
plt.setp(self.bp['caps'][(i * 2) + 1], color=clr, linewidth=1)
plt.setp(self.bp['caps'][(i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][i * 2], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][1 + (i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['fliers'][i],
markerfacecolor=clr,
markeredgecolor=clr,
marker='s',
markersize=2.0)
if clr == ac: items.append(Rect((0, 0), 1, 1, fc=clr))
else: items.append(Rect((0, 0), 1, 1, fc=a, ec=b, hatch='*'))
labels.append(opt)
clevels = np.linspace(0., 1., len(vals))
xJitters = [
np.random.normal(pos[i], 0.01, len(vals[i]))
for i in range(len(vals))
]
yJitters = [[
vals[i][j] * np.random.normal(1, 0.005, len(vals[i]))[j]
for j in range(len(vals[i]))
] for i in range(len(vals))]
for xJ, yJ, clevel in zip(xJitters, yJitters, colors):
plt.scatter(xJ, yJ, c=clevel, alpha=0.7, s=4, zorder=9)
for patch, clevel in zip(self.bp['boxes'], colors):
patch.set_facecolor(clevel) #cm.prism(clevel))
patch.set_alpha(0.5)
yMin, yMax = self.ax.get_ylim()
yStep = (yMax - yMin) / 30.0
# self.ax.set_xticklabels([opt.split('~')[-1] for opt in opts],fontsize=10,fontweight='bold',rotation=-30)
for j, xpos in enumerate(self.ax.get_xticks()):
#self.ax.text(xpos,yMin-yStep,opts[j].split('~')[-1],fontweight='bold',rotation=60,verticalalignment='top',horizontalalignment='center')
self.ax.text(xpos,
yMin + yStep,
opts[j].split('~')[-1],
fontsize=25,
fontweight='bold',
verticalalignment='bottom',
horizontalalignment='center')
self.ax.set_xlim([pos[0] - 0.2, pos[-1] + 0.2])
# self.ax.plot((pos[0],pos[-1]),(yMin-yStep*2,yMin-2*yStep))
self.ax.axis('off')
title_key = {'LOC': 'LAYER', 'FS': 'FIRING STYLE'}
if TITLE:
if TITLE in title_key: TITLE = title_key[TITLE]
nc, fs, cs, hp, bb1, bb2 = 1, 12, 0.5, 0.5, 0.2, 1.1
if BW: self.ax.set_title(TITLE)
else:
if len(items) > 4:
if len(items) < 10:
nc, fs, cs, hp = 2, 10, 0.4, 0.4
elif len(items) < 15:
nc, fs, cs, hp, bb1, bb2 = 3, 10, 0.35, 0.35, 0.85, 1.1
elif len(items) < 25:
nc, fs, cs, hp, bb1, bb2 = 4, 9, 0.25, 0.25, 0.85, 1.1
elif len(items) < 35:
nc, fs, cs, hp, bb1, bb2 = 5, 8, 0.20, 0.20, 0.8, 1.1
else:
nc, fs, cs, hp, bb1, bb2 = 6, 7, 0.15, 0.15, 0.75, 1.1
leg = self.ax.legend(items,
labels,
title=TITLE,
handletextpad=hp,
columnspacing=cs,
fontsize=fs,
ncol=nc,
bbox_to_anchor=(bb1, bb2),
loc='upper center')
self.yLoc += 1
if self.yLoc == self.yLen:
self.xLoc += 1
self.yLoc = 0
def plot_tree_classif(pred_dict, phn_dict, auc_dict, use_lvls, cdata_dict,
args):
base_cdata = tuple(cdata_dict.values())[0]
base_lvls = 'Exon__Location__Protein'
lvls_k = tuple(use_lvls.split('__'))
base_cdata.add_mut_lvls(('Gene', ) + lvls_k)
use_mtree = base_cdata.mtrees[('Gene', ) + lvls_k][args.gene]
use_mtypes = {(src, clf): [MuType({('Gene', args.gene): pnt_mtype})]
for src, lvls, clf in pred_dict if lvls == use_lvls}
use_criter = {(src, clf): [(np.mean(phn_dict[src, base_lvls,
clf][mtypes[0]]),
auc_dict[src, base_lvls, clf].loc[mtypes[0]])]
for (src, clf), mtypes in use_mtypes.items()}
for src, clf in use_mtypes:
cur_mtypes = {
mtype.subtype_iter()[0][1]
for mtype, phn in phn_dict[src, use_lvls, clf].items()
if (not isinstance(mtype, RandomType) and mtype.subtype_iter()[0]
[1] != pnt_mtype and (mtype.subtype_iter()[0][1]
& copy_mtype).is_empty())
}
if len(cur_mtypes) >= 5:
use_mtypes[src, clf] += [
MuType({('Gene', args.gene): mtype})
for mtype in random.sample(sorted(cur_mtypes), k=5)
]
use_criter[src, clf] += [(np.mean(phn_dict[src, use_lvls,
clf][mtype]),
auc_dict[src, use_lvls, clf].loc[mtype])
for mtype in use_mtypes[src, clf][1:]]
use_src, use_clf = sorted(
use_criter.items(),
key=lambda x: np.prod(np.var(np.array(x[1]), axis=0)))[-1][0]
plt_mtypes = use_mtypes[use_src, use_clf]
fig = plt.figure(figsize=(1.1 + 3.1 * len(plt_mtypes), 12))
gs = gridspec.GridSpec(nrows=3,
ncols=len(plt_mtypes) + 1,
width_ratios=[1] + [3] * len(plt_mtypes),
height_ratios=[2, 1, 4])
lbl_ax = fig.add_subplot(gs[:, 0])
lbl_ax.axis('off')
tree_ax = fig.add_subplot(gs[0, 1:])
tree_ax.axis('off')
leaf_count = len(MuType(use_mtree.allkey()).leaves())
tree_ax.add_patch(
Rect((leaf_count * 0.03, len(lvls_k) + 0.17),
leaf_count * 0.23,
0.93,
facecolor=variant_clrs['WT'],
alpha=0.41,
clip_on=False,
linewidth=0))
tree_ax.text(
leaf_count * 0.15,
len(lvls_k) + 0.59,
"Wild-Type for\n{} Point Mutations\n({} samples)".format(
args.gene,
len(set(base_cdata.get_samples()) - set(use_mtree.get_samples()))),
size=19,
ha='center',
va='center',
weight='semibold')
tree_ax.add_patch(
Rect((leaf_count * 0.31, len(lvls_k) + 0.23),
leaf_count * 0.43,
0.79,
facecolor=variant_clrs['Point'],
alpha=0.41,
clip_on=False,
linewidth=0))
tree_ax.text(leaf_count / 2,
len(lvls_k) + 0.61,
"All {} Point Mutations\n({} samples)".format(
args.gene, len(use_mtree)),
size=19,
ha='center',
va='center',
weight='semibold')
tree_ax = recurse_labels(tree_ax,
use_mtree, (0, leaf_count),
len(lvls_k),
leaf_count,
clr_mtype=False,
add_lbls=True)
tree_ax.set_xlim(0, leaf_count * 1.03)
tree_ax.set_ylim(0, len(lvls_k) + 0.6)
for i, lvl in enumerate(lvls_k):
lbl_ax.text(1.31,
0.88 - i / 10.11,
clean_level(lvl),
size=19,
ha='right',
va='center')
for i, plt_mtype in enumerate(plt_mtypes):
mtype_ax = fig.add_subplot(gs[1, i + 1])
mtype_ax.axis('off')
if plt_mtype == MuType({('Gene', args.gene): pnt_mtype}):
tree_mtype = None
top_fc = variant_clrs['Point']
top_ec = 'none'
else:
tree_mtype = plt_mtype.subtype_iter()[0][1].subtype_iter()[0][1]
top_fc = 'none'
top_ec = variant_clrs['Point']
mtype_ax.add_patch(
Rect((leaf_count * 0.19, len(lvls_k) - 0.19),
leaf_count * 0.67,
0.31,
clip_on=False,
facecolor=top_fc,
edgecolor=top_ec,
alpha=0.41,
linewidth=2.7))
mtype_ax = recurse_labels(mtype_ax,
use_mtree, (0, leaf_count),
len(lvls_k) - 0.4,
leaf_count,
clr_mtype=tree_mtype,
add_lbls=False)
mtype_ax.set_xlim(0, leaf_count * 1.03)
mtype_ax.set_ylim(0, len(lvls_k) + 0.6)
if i == 0:
pred_vals = pred_dict[use_src, base_lvls, use_clf].loc[plt_mtype]
use_phn = phn_dict[src, base_lvls, clf][plt_mtype]
else:
pred_vals = pred_dict[use_src, use_lvls, use_clf].loc[plt_mtype]
use_phn = phn_dict[src, use_lvls, clf][plt_mtype]
pred_vals = pred_vals.apply(np.mean)
use_auc = use_criter[src, clf][i][1]
viol_ax = fig.add_subplot(gs[2, i + 1])
sns.violinplot(x=pred_vals[~use_phn].values,
ax=viol_ax,
palette=[variant_clrs['WT']],
orient='v',
linewidth=0,
cut=0,
width=0.67)
sns.violinplot(x=pred_vals[use_phn].values,
ax=viol_ax,
palette=[variant_clrs['Point']],
orient='v',
linewidth=0,
cut=0,
width=0.67)
viol_ax.text(0.5,
1.01,
'\n'.join(get_fancy_label(plt_mtype).split('\n')[1:]),
size=12,
ha='center',
va='top',
transform=viol_ax.transAxes)
viol_ax.text(1.07,
0.83,
'\n'.join([str(np.sum(use_phn)), "mutated", "samples"]),
size=13,
ha='right',
va='top',
c=variant_clrs['Point'],
weight='semibold',
transform=viol_ax.transAxes)
viol_ax.text(0.5,
-0.05,
"AUC: {:.3f}".format(use_auc),
size=21,
ha='center',
va='bottom',
transform=viol_ax.transAxes)
viol_ax.get_children()[0].set_alpha(0.41)
viol_ax.get_children()[2].set_alpha(0.41)
viol_ax.set_yticklabels([])
viol_ylims = viol_ax.get_ylim()
ylim_gap = (viol_ylims[1] - viol_ylims[0]) / 13
viol_ax.set_ylim([viol_ylims[0], viol_ylims[1] + ylim_gap])
# save the plot to file
fig.tight_layout(w_pad=1.9, h_pad=1.5)
fig.savefig(os.path.join(
plot_dir, args.cohort,
"{}_tree-classif__{}.svg".format(args.gene, use_lvls)),
bbox_inches='tight',
format='svg')
plt.close()
def plot_lollipop(cdata_dict, domain_dict, args):
fig, ax = plt.subplots(figsize=(9, 5))
use_cdata = tuple(cdata_dict.values())[0]
gn_annot = use_cdata.gene_annot[args.gene]
loc_lvls = 'Gene', 'Scale', 'Copy', 'Consequence', 'Position', 'HGVSp'
if loc_lvls not in use_cdata.mtrees:
use_cdata.add_mut_lvls(loc_lvls)
use_mtree = use_cdata.mtrees[loc_lvls][args.gene]
mut_count = len(use_mtree.get_samples())
var_count = len(use_mtree['Point'].get_samples())
pie_clrs = [
choose_subtype_colour(pnt_mtype),
choose_subtype_colour(pnt_mtype | gains_mtype),
choose_subtype_colour(pnt_mtype | dels_mtype)
]
pie_ax = ax.inset_axes(bounds=(0.05, 0.41, 0.27, 0.27))
pie_ax.pie(x=get_pie_counts(pnt_mtype, use_mtree),
colors=pie_clrs,
labels=[
'point muts\nwithout CNAs', 'point & any gain',
'point & any loss'
],
labeldistance=1.23,
explode=[0.19] * 3,
startangle=90,
autopct=lambda x: format(x * var_count / 100, '.0f'),
pctdistance=0.37,
wedgeprops=dict(alpha=0.67),
textprops=dict(size=9))
for pie_indx1, pie_indx2 in [(3, 5), (5, 7)]:
txt_arts1 = pie_ax.get_children()[pie_indx1]
txt_arts2 = pie_ax.get_children()[pie_indx2]
txt_x1, txt_y1 = txt_arts1.get_position()
txt_x2, txt_y2 = txt_arts2.get_position()
if abs(txt_x1 - txt_x2) < 1 and abs(txt_y1 - txt_y2) < 0.37:
if txt_y1 >= txt_y2:
txt_arts2.set_position((txt_x2, txt_y2 - 0.41))
else:
txt_arts1.set_position((txt_x1, txt_y1 - 0.41))
loc_dict = {
form: sorted(
[(int(loc.split('-')[0]), len(loc_muts.get_samples()))
for loc, loc_muts in form_muts if loc.split('-')[0].isnumeric()],
key=itemgetter(0))
for form, form_muts in use_mtree['Point']
}
# calculate the minimum and maximum amino acid positions for the
# mutations to be plotted, as well as the most samples at any hotspot
min_pos = min(pos for loc_counts in loc_dict.values()
for pos, _ in loc_counts if pos >= 0)
max_pos = max(pos for loc_counts in loc_dict.values()
for pos, _ in loc_counts)
min_pos -= (max_pos - min_pos) / 43
max_count = max(
max(count for loc_counts in loc_dict.values()
for _, count in loc_counts), 13)
pos_rng = max_pos - min_pos
lgnd_ptchs = []
for form, form_muts in use_mtree['Point']:
if loc_dict[form]:
form_mtype = MuType({
('Scale', 'Point'): {
('Consequence', form): None
}
})
lgnd_ptchs += [
Patch(color=form_clrs[form],
alpha=0.67,
label=get_fancy_label(form_mtype))
]
mrks, stms, basl = ax.stem(*zip(*loc_dict[form]),
use_line_collection=True)
plt.setp(mrks,
markersize=7,
markeredgecolor='black',
markerfacecolor=form_clrs[form],
zorder=5)
plt.setp(stms, linewidth=0.8, color='black', zorder=1)
plt.setp(basl, linewidth=1.1, color='black', zorder=2)
for loc, loc_muts in form_muts:
loc_size = len(loc_muts.get_samples())
if loc != '-' and loc_size >= 10:
loc_int = int(loc.split('-')[0])
loc_mtypes = sorted(
[
MuType({('HGVSp', lbl): None})
for lbl, _ in loc_muts
],
key=lambda mtype: len(mtype.get_samples(form_muts)
))[::-1]
mut_lbls = [
get_fancy_label(loc_mtype) for loc_mtype in loc_mtypes
]
root_strs = {
re.match('[A-Z][0-9]+', lbl).group()
for lbl in mut_lbls
}
if len(root_strs) > 1:
loc_lbl = '\n'.join(mut_lbls)
else:
lbl_root = tuple(root_strs)[0]
if max(len(lbl) - len(lbl_root)
for lbl in mut_lbls) > 4:
loc_lbl = '\n'.join([mut_lbls[0]] + [
''.join([
' ' * len(lbl_root) * 2,
lbl.split(lbl_root)[1]
]) for lbl in mut_lbls[1:]
])
else:
loc_lbl = "/".join([mut_lbls[0]] + [
lbl.split(lbl_root)[1] for lbl in mut_lbls[1:]
])
ax.text(loc_int + pos_rng / 115,
loc_size + max_count / 151,
loc_lbl,
size=8,
ha='left',
va='center')
loc_mtype = MuType({
('Scale', 'Point'): {
('Consequence', form): {
('Position', loc): None
}
}
})
pie_ax = ax.inset_axes(bounds=(loc_int - pos_rng / 7,
loc_size, max_pos / 5,
max_count / 5),
transform=ax.transData)
pie_ax.pie(x=get_pie_counts(loc_mtype, use_mtree),
colors=pie_clrs,
explode=[0.09] * 3,
startangle=90,
wedgeprops=dict(alpha=0.67))
use_tx = use_cdata._muts.loc[
(use_cdata._muts.Gene == args.gene)
& ~use_cdata._muts.Feature.isnull()].Feature.unique()
assert len(use_tx) == 1, (
"Multiple transcripts detected in {} for {} !".format(
args.cohort, args.gene))
# TODO: do domains need to be included in these plots?
use_tx = use_tx[0]
prot_patches = []
for i, (domn_lbl, domn_data) in enumerate(domain_dict.items()):
tx_annot = gn_annot['Transcripts'][use_tx]
gene_domns = domn_data[(domn_data.Gene == gn_annot['Ens'])
& (domn_data.Transcript == use_tx)]
min_pos = min(min_pos, gene_domns.DomainStart.min() - max_pos / 67)
for domn_id, domn_start, domn_end in zip(gene_domns.DomainID,
gene_domns.DomainStart,
gene_domns.DomainEnd):
prot_patches.append(
Rect((domn_start, -max_count * (0.15 + i * 0.11)),
domn_end - domn_start, max_count * 0.09))
ax.text((domn_start + domn_end) / 2,
-max_count * (0.11 + i * 0.11),
domn_id,
size=9,
ha='center',
va='center')
ax.add_collection(
PatchCollection(prot_patches, color='#D99100', alpha=0.4, linewidth=0))
for i, domn_nm in enumerate(domain_dict):
ax.text(min_pos,
-max_count * (0.07 + i * 0.13),
"{}\nDomains".format(domn_nm),
size=8,
ha='right',
va='top',
linespacing=0.71,
rotation=37)
ax.text(0.03,
0.98,
args.gene,
size=14,
fontweight='semibold',
ha='left',
va='center',
transform=ax.transAxes)
ax.text(0.03,
0.941,
"{} point mutants\n{} gain mutants\n{} loss mutants"
"\n{:.1%} of {} affected".format(
var_count,
len(gains_mtype.get_samples(use_mtree)),
len(dels_mtype.get_samples(use_mtree)),
mut_count / len(use_cdata.get_samples()),
get_cohort_label(args.cohort),
),
size=12,
ha='left',
va='top',
transform=ax.transAxes)
# add the legend for the colour used for each form of mutation
plt_lgnd = ax.legend(handles=lgnd_ptchs,
frameon=False,
fontsize=11,
ncol=2,
loc=1,
handletextpad=0.7,
bbox_to_anchor=(0.98, 1.02))
ax.add_artist(plt_lgnd)
ax.grid(linewidth=0.41, alpha=0.41)
ax.set_xlabel("Amino Acid Position", size=17, weight='semibold')
ax.set_ylabel(" # of Mutated Samples", size=17, weight='semibold')
ax.set_xlim(min_pos, max_pos * 1.02)
ax.set_ylim(-max_count * (0.05 + len(domain_dict) * 0.13),
max_count * 17 / 11)
ax.set_yticks([tck for tck in ax.get_yticks() if tck >= 0])
fig.savefig(os.path.join(plot_dir, args.cohort,
"{}_lollipop.svg".format(args.gene)),
bbox_inches='tight',
format='svg')
plt.close()
def plot_lollipop(cdata_dict, domain_dict, args):
fig, main_ax = plt.subplots(figsize=(11, 4))
base_cdata = tuple(cdata_dict.values())[0]
gn_annot = base_cdata.gene_annot[args.gene]
base_lvls = 'Gene', 'Scale', 'Form_base', 'Location', 'Protein',
domn_lvls = [('Gene', 'Scale', 'Transcript', '_'.join(['Domain', domn_nm]),
'Location') for domn_nm in domain_dict]
for lvls in [base_lvls] + domn_lvls:
if lvls not in base_cdata.mtrees:
base_cdata.add_mut_lvls(lvls)
loc_mtree = base_cdata.mtrees[base_lvls][args.gene]['Point']
#TODO: double-check mutations in cohorts such as METABRIC that are not
# numeric but maybe should be (i.e. in CDH1, MAP3K1)
loc_dict = {
form: sorted([(int(loc), len(loc_muts)) if loc.isnumeric() else
(-1, len(loc_muts)) for loc, loc_muts in form_muts],
key=itemgetter(0))
for form, form_muts in loc_mtree
}
# calculate the minimum and maximum amino acid positions for the
# mutations to be plotted, as well as the most samples at any hotspot
min_pos = min(pos for loc_counts in loc_dict.values()
for pos, _ in loc_counts if pos >= 0)
max_pos = max(pos for loc_counts in loc_dict.values()
for pos, _ in loc_counts)
max_count = max(count for loc_counts in loc_dict.values()
for _, count in loc_counts)
lgnd_ptchs = []
for form, form_muts in loc_mtree:
lgnd_ptchs += [
Patch(color=form_clrs[form],
alpha=0.53,
label=clean_label(form, 'Form_base'))
]
mrks, stms, basl = main_ax.stem(*zip(*loc_dict[form]),
use_line_collection=True)
plt.setp(mrks,
markersize=7,
markeredgecolor='black',
markerfacecolor=form_clrs[form],
zorder=5)
plt.setp(stms, linewidth=0.8, color='black', zorder=1)
plt.setp(basl, linewidth=1.1, color='black', zorder=2)
for loc, loc_muts in form_muts:
if len(loc_muts) >= 10:
mut_lbls = sorted(lbl for lbl, _ in loc_muts)
root_indx = re.match('p.[A-Z][0-9]+', mut_lbls[0]).span()[1]
lbl_root = mut_lbls[0][2:root_indx]
if max(len(lbl) - len(lbl_root) for lbl in mut_lbls) > 4:
loc_lbl = '\n'.join([mut_lbls[0][2:]] + [
''.join(
[' ' * len(lbl_root) * 2,
lbl.split(lbl_root)[1]]) for lbl in mut_lbls[1:]
])
else:
loc_lbl = "/".join(
[mut_lbls[0][2:]] +
[lbl.split(lbl_root)[1] for lbl in mut_lbls[1:]])
main_ax.text(int(loc) + (max_pos - min_pos) / 115,
len(loc_muts) + max_count / 151,
loc_lbl,
size=8,
ha='left',
va='center')
prot_patches = []
for i, lvls in enumerate(domn_lvls):
tx_mtree = base_cdata.mtrees[lvls][args.gene]['Point']
tx_id = tuple(tx_mtree)[0][0]
tx_annot = gn_annot['Transcripts'][tx_id]
domn_nm = lvls[3].split('_')[1]
domn_df = domain_dict[domn_nm]
gene_domns = domn_df[(domn_df.Gene == gn_annot['Ens'])
& (domn_df.Transcript == tx_id)]
for domn_id, domn_start, domn_end in zip(gene_domns.DomainID,
gene_domns.DomainStart,
gene_domns.DomainEnd):
prot_patches.append(
Rect((domn_start, -max_count * (0.3 + i * 0.13)),
domn_end - domn_start, max_count * 0.11))
main_ax.text((domn_start + domn_end) / 2,
-max_count * (0.25 + i * 0.13),
domn_id,
size=9,
ha='center',
va='center')
main_ax.add_collection(
PatchCollection(prot_patches, alpha=0.4, linewidth=0, color='#D99100'))
exn_patches = []
exn_pos = 1
for i, exn_annot in enumerate(tx_annot['Exons']):
exn_len = exn_annot['End'] - exn_annot['Start'] + 1
if 'UTRs' in tx_annot:
for utr_annot in tx_annot['UTRs']:
if (exn_annot['Start'] <= utr_annot['Start'] <=
exn_annot['End'] <= utr_annot['End']):
exn_len -= exn_annot['End'] - utr_annot['Start'] + 1
elif (exn_annot['Start'] <= utr_annot['Start'] <=
utr_annot['End'] <= exn_annot['End']):
exn_len -= utr_annot['End'] - utr_annot['Start'] + 1
if exn_len > 0 and exn_pos <= max_pos:
exn_len //= 3
if i == (len(tx_annot['Exons']) - 1):
if (exn_pos + exn_len) > max_pos:
exn_len = max_pos - exn_pos + 10
if (exn_pos + exn_len) >= min_pos:
exn_patches.append(
Rect((exn_pos, max_count * -0.15),
exn_len,
max_count * 0.11,
color='green'))
main_ax.text(max(exn_pos + exn_len / 2, min_pos + 5),
max_count * -0.1,
exn_annot['number'],
size=min(11, (531 * exn_len / max_pos)**0.6),
ha='center',
va='center')
exn_pos += exn_len
for i, domn_nm in enumerate(domain_dict):
main_ax.text(min_pos - exn_pos / 29,
-max_count * (0.23 + i * 0.15),
"{}\nDomains".format(domn_nm),
size=7,
ha='right',
va='top',
linespacing=0.65,
rotation=37)
# add the patches describing the boundaries of each exon and annotate them
main_ax.add_collection(
PatchCollection(exn_patches, alpha=0.4, linewidth=1.4,
color='#002C91'))
main_ax.text(min_pos - exn_pos / 29,
max_count * -0.08,
"{}\nExons".format(tx_annot['transcript_name']),
size=7,
ha='right',
va='top',
linespacing=0.65,
rotation=37)
if '_' in args.cohort:
coh_lbl = "{}({})".format(*args.cohort.split('_'))
else:
coh_lbl = str(args.cohort)
main_ax.text(0.03,
0.97,
"{} {}-mutated samples\n{:.1%} of {} affected".format(
len(loc_mtree),
args.gene,
len(loc_mtree) / len(base_cdata.get_samples()),
coh_lbl,
),
size=13,
ha='left',
va='top',
transform=main_ax.transAxes)
# add the legend for the colour used for each form of mutation
plt_lgnd = main_ax.legend(handles=lgnd_ptchs,
frameon=False,
fontsize=11,
ncol=3,
loc=1,
handletextpad=0.7,
bbox_to_anchor=(0.98, 1.02))
main_ax.add_artist(plt_lgnd)
main_ax.grid(linewidth=0.31)
main_ax.set_xlabel("Amino Acid Position", size=17, weight='semibold')
main_ax.set_ylabel(" # of Mutated Samples",
size=17,
weight='semibold')
main_ax.set_xlim(min_pos - exn_pos / 29, exn_pos * 1.02)
main_ax.set_ylim(-max_count * (0.03 + len(domain_dict) * 0.23),
max_count * 24 / 17)
main_ax.set_yticks([tck for tck in main_ax.get_yticks() if tck >= 0])
# save the plot to file
fig.savefig(os.path.join(plot_dir, args.cohort,
"{}_lollipop.svg".format(args.gene)),
bbox_inches='tight',
format='svg')
plt.close()
def fill_boxes(self,
opts,
vals,
colors,
alt_colors,
JITTERS=None,
SIG_STARS=None):
G_THRES, L_THRES = 0.05, 0.05
self.bp = self.axes[-1].boxplot(vals,
positions=self.pos,
widths=self.width,
patch_artist=True,
showmeans=True,
whis=0.7)
for i, opt in enumerate(opts):
clr, ac = colors[i], alt_colors[i]
self.bp['boxes'][i].set_edgecolor(clr)
self.bp['boxes'][i].set_linewidth(1)
plt.setp(self.bp['medians'][i], color=clr, linewidth=3)
plt.setp(self.bp['means'][i],
marker='h',
markersize=9,
markerfacecolor=clr,
markeredgecolor=ac)
plt.setp(self.bp['caps'][(i * 2) + 1], color=clr, linewidth=1)
plt.setp(self.bp['caps'][(i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][i * 2], color=clr, linewidth=1)
plt.setp(self.bp['whiskers'][1 + (i * 2)], color=clr, linewidth=1)
plt.setp(self.bp['fliers'][i],
markerfacecolor=clr,
markeredgecolor=ac,
markeredgewidth=2,
marker='s',
markersize=2.0)
for box, opt, clr, ac in zip(self.bp['boxes'], opts, colors,
alt_colors):
if clr == ac:
box.set_facecolor(clr)
self.major_legend_items[opt] = Rect((0, 0), 1, 1, fc=clr)
else:
box.set(ec=clr, facecolor=ac, linewidth=4, hatch='+')
self.minor_legend_items[opt] = Rect((0, 0),
1,
1,
fc='white',
ec=clr,
linewidth=2,
hatch='+')
box.set_alpha(0.5)
if JITTERS != None:
for i, (xJ, yJ) in enumerate(JITTERS):
plt.scatter(xJ, yJ, c=colors[i], alpha=0.7, s=4, zorder=9)
if SIG_STARS != None:
for i, (g, l) in enumerate(zip(SIG_STARS[0], SIG_STARS[1])):
if l < L_THRES:
plt.scatter(self.pos[i],
max(vals[i]) * 1.1,
s=pv_2_size(l),
marker='*',
color='silver',
edgecolor='black',
linewidth=1)
if g < G_THRES:
plt.scatter(self.pos[i], (max(vals[i]) * 1.1) + 2,
s=1.5 * pv_2_size(l),
marker='*',
color='gold',
edgecolor='black',
linewidth=1)
def __init__(self, args):
self.tFS = 25
self.log = False
self.color_key = {
'PANTHERS': 'darkblue',
'GORILLAS': 'green',
'ELEPHANTS': 'brown',
'HUMANS': 'red',
'POLAR_BEARS': 'purple'
}
self.coord_key = {'weight': 0, 'brain': 1, 'speed': 2, 'shade': 3}
self.label_key = {
'weight': 'Body Mass (g)',
'brain': 'Brain Mass (g)',
'speed': 'Top Speed (m/s)',
'shade': 'Grayscale Shade'
}
self.color_key['ORCAS'] = 'k'
self.color_key['DOLPHINS'] = 'gray'
self.color_key['TIGERS'] = 'orange'
self.noise = []
self.noise.append([5, [(100, 50), (1000, 500)]])
self.noise.append([5, [('U', 0, 5000)]])
self.noise.append([10, [(100, 25)]])
self.noise.append([10, [(1000, 250)]])
# self.noise.append([5,[(25,25)]])
# self.noise.append([5,[(250,250)]])
# self.noise.append([5,[(2500,2500)]])
# self.noise.append([5,[(100,200)]])
# self.noise.append([5,[(1000,2000)]])
# self.noise.append([1,[(1000,2000)]])
# self.noise.append([5,[(1000,1000)]])
# self.noise.append([10,[(1,250)]])
# self.noise.append([5,[(50,25)]])
# self.noise.append([10,[(1,250)]])
# self.noise.append([1,[(1000,750)]])
# self.noise.append([5, [(50,10),(500,100)]])
# self.noise += [[5,[(20000,10000)]]]
# self.noise = [[5,[(2000,1000)]]]
# self.noise = [[5,[(2000,1000)]]]
self.er, self.dp = 0.05, 0.025
# self.er,self.dp = 0.1,0.000001
self.ncol = len(self.color_key)
self.labels, self.items = [], []
for xI, yI in self.color_key.items():
self.labels.append(xI)
self.items.append(Rect((0, 0), 1, 1, fc=yI))
self.stats = {}
self.stats['HUMANS'] = [70, 20, 1400, 30, 8.5, 1.0, 50, 20]
self.stats['ELEPHANTS'] = [3000, 100, 500, 20, 7.0, 1.0, 15, 8]
self.stats['GORILLAS'] = [200, 50, 500, 20, 9.75, 0.75, 88, 2]
self.stats['PANTHERS'] = [60, 14, 150, 20, 18.5, 1, 90, 2]
self.stats['POLAR_BEARS'] = [400, 75, 480, 20, 10.25, 0.75, 8, 2]
self.stats['ORCAS'] = [10000, 1000, 5000, 20, 0.5, 0.01, 90, 2]
self.stats['DOLPHINS'] = [250, 25, 1600, 100, 0.5, 0.1, 85, 10]
self.stats['TIGERS'] = [300, 50, 250, 20, 15.5, 1, 60, 20]
self.tXLoc, self.tYLoc = 0.94, 1.035
ds = [
'mass(kg)', '', 'brain_size(g)', '', 'Top_Land_Speed(m/s)', '',
'Grayscale_Shade(%)'
]
def plot_mutation_lollipop(cdata, domain_data, args):
fig, main_ax = plt.subplots(figsize=(10, 4))
# get tree of point mutations and the samples carrying
# each of the major mutation types
pnt_muts = cdata.train_mut['Point']
samp_dict = {
lbl: mtype.get_samples(cdata.train_mut)
for lbl, mtype in variant_mtypes
}
# get the number of samples with a point mutation at each amino acid
mut_counts = sorted([(int(loc), len(muts))
for loc, muts in pnt_muts if loc != '.'],
key=itemgetter(0))
# create the mutation count lollipops and set the aesthetics of
# the individual lollipop elements
mrks, stms, basl = main_ax.stem(*zip(*mut_counts))
plt.setp(mrks, markersize=5, markeredgecolor='black', zorder=5)
plt.setp(stms, linewidth=0.8, color='black', zorder=1)
plt.setp(basl, linewidth=1.1, color='black', zorder=2)
# for each amino acid location with at least twenty mutated samples, get
# the list of specific amino acid substitutions present at that location
for loc, mut_count in mut_counts:
if mut_count >= 20:
mut_lbls = sorted(lbl for lbl, _ in pnt_muts[str(loc)])
lbl_root = mut_lbls[0][2:-1]
# create a label on the plot next to the head of the lollipop for
# this point mutation location listing the aa substitutions
main_ax.text(loc + mut_counts[-1][0] / 109,
mut_count,
lbl_root +
"/".join(lbl.split(lbl_root)[-1] for lbl in mut_lbls),
size=11,
ha='left',
va='bottom')
# create a plotting space for a Venn diagram showing the overlap
# between these point mutations and copy number alterations
pie_ax = inset_axes(main_ax,
width=0.57,
height=0.57,
bbox_to_anchor=(loc, mut_count),
bbox_transform=main_ax.transData,
loc=4,
axes_kwargs=dict(aspect='equal'),
borderpad=0)
# get the number of samples with a point mutation at this location
# that also have a gain or loss alteration, or neither
loc_samps = pnt_muts[str(loc)].get_samples()
loc_ovlps = [
len(loc_samps & samp_dict[lbl]) if lbl != 'Point' else
len(loc_samps - samp_dict['Gain'] - samp_dict['Loss'])
for lbl, _ in variant_mtypes
]
# get the 2x2 tables of sample overlap for those with point
# mutations and those with gain or loss alterations
loc_croxs = [[[loc_ovlp, len(loc_samps - samp_dict[lbl])],
[
len(samp_dict[lbl] - loc_samps),
len(cdata.samples - loc_samps - samp_dict[lbl])
]] if lbl != 'Point' else None
for loc_ovlp, (lbl,
_) in zip(loc_ovlps, variant_mtypes)]
# test for statistically significant co-occurence or mutual
# exclusivity between these point mutations and alterations
loc_tests = [(fisher_exact(loc_crox, alternative='less')[1],
fisher_exact(loc_crox, alternative='greater')[1])
if lbl != 'Point' else None
for loc_crox, (lbl,
_) in zip(loc_croxs, variant_mtypes)]
# create labels for sample ounts and significance for the overlap
# Venn diagram
loc_lbls = [str(loc_ovlp) for loc_ovlp in loc_ovlps]
for i, (loc_test,
(lbl, _)) in enumerate(zip(loc_tests, variant_mtypes)):
if lbl != 'Point':
if loc_test[0] < 0.05:
loc_lbls[i] += '(-)'
if loc_test[0] < 0.001:
loc_lbls[i] += '**'
else:
loc_lbls[i] += '*'
if loc_test[1] < 0.05:
loc_lbls[i] += '(+)'
if loc_test[1] < 0.001:
loc_lbls[i] += '**'
else:
loc_lbls[i] += '*'
# plot the overlap Venn diagram next to the head of the lollipop
pie_ptchs, pie_txts = pie_ax.pie(
x=loc_ovlps,
labels=loc_lbls,
explode=[0.13, 0, 0.13],
colors=[
variant_clrs[lbl]
if lbl == 'Point' else mcomb_clrs["Point+{}".format(lbl)]
for lbl, _ in variant_mtypes
],
labeldistance=0.47,
wedgeprops=dict(alpha=0.71))
# adjust the properties of the Venn diagram's text annotation
for i in range(len(pie_txts)):
pie_txts[i].set_fontsize(7)
pie_txts[i].set_horizontalalignment('center')
gn_annot = cdata.gene_annot[args.gene]
main_tx = {
tx_id
for tx_id, tx_annot in gn_annot['Transcripts'].items()
if tx_annot['transcript_name'] == '{}-001'.format(args.gene)
}
prot_patches = []
max_count = max(count for _, count in mut_counts)
gene_doms = domain_data[(domain_data['Gene'] == gn_annot['Ens'])
& (domain_data['Transcript'].isin(main_tx))]
for dom_id, dom_start, dom_end in zip(gene_doms.DomainID,
gene_doms.DomainStart,
gene_doms.DomainEnd):
prot_patches.append(
Rect((dom_start, max_count * -0.12), dom_end - dom_start,
max_count * 0.08))
main_ax.text((dom_start + dom_end) / 2,
max_count * -0.086,
dom_id,
size=9,
ha='center',
va='center')
main_ax.add_collection(
PatchCollection(prot_patches, alpha=0.4, linewidth=0, color='#D99100'))
exn_patches = []
exn_pos = 1
for i, exn_annot in enumerate(gn_annot['Exons']):
exn_len = exn_annot['End'] - exn_annot['Start']
if 'UTR' in exn_annot:
for utr_annot in exn_annot['UTR']:
exn_len -= utr_annot['End'] - utr_annot['Start']
if exn_len > 0 and exn_pos <= mut_counts[-1][0]:
exn_len /= 3
exn_patches.append(
Rect((exn_pos, max_count * -0.23),
exn_len,
max_count * 0.08,
color='green'))
main_ax.text(exn_pos + exn_len / 2,
max_count * -0.196,
"{}/{}".format(i + 1, len(gn_annot['Exons'])),
size=min(11,
(531 * exn_len / mut_counts[-1][0])**0.6),
ha='center',
va='center')
exn_pos += exn_len
main_ax.add_collection(
PatchCollection(exn_patches, alpha=0.4, linewidth=1.4,
color='#002C91'))
main_ax.text(exn_pos / -391,
max_count * -0.05,
"{}\nDomains".format(args.domains),
size=7,
ha='right',
va='top',
linespacing=0.65,
rotation=37)
main_ax.text(exn_pos / -391,
max_count * -0.16,
"{}-001\nExons".format(args.gene),
size=7,
ha='right',
va='top',
linespacing=0.65,
rotation=37)
main_ax.text(0.02,
0.34,
"{} {}-mutated samples\n{:.1%} of {} cohort affected".format(
len(pnt_muts),
args.gene,
len(pnt_muts) / len(cdata.samples),
args.cohort,
),
size=9,
va='bottom',
transform=main_ax.transAxes)
main_ax.set_xlabel("Amino Acid Position", size=15, weight='semibold')
main_ax.set_ylabel("# of Mutated Samples", size=15, weight='semibold')
main_ax.grid(linewidth=0.31)
main_ax.set_xlim(exn_pos / -519, exn_pos * 1.01)
main_ax.set_ylim(max_count / -3.6, max_count * 1.21)
main_ax.set_yticks([tck for tck in main_ax.get_yticks() if tck >= 0])
venn_ax = inset_axes(main_ax,
width=2.19,
height=1.31,
loc=3,
bbox_to_anchor=(mut_counts[-1][0] / 103,
max_count * 0.67),
bbox_transform=main_ax.transData,
borderpad=0)
v_plot = venn3([samp_dict[lbl] for lbl, _ in variant_mtypes[::-1]],
["Gains", "Point\nMutations", "Losses"],
[variant_clrs[lbl] for lbl, _ in variant_mtypes[::-1]],
alpha=0.71,
ax=venn_ax)
for i in range(len(v_plot.set_labels)):
if v_plot.set_labels[i] is not None:
v_plot.set_labels[i].set_fontsize(11)
for i in range(len(v_plot.subset_labels)):
if v_plot.subset_labels[i] is not None:
v_plot.subset_labels[i].set_fontsize(10)
# save the plot to file
fig.savefig(os.path.join(
plot_dir,
"mut-lollipop_{}__{}_domains-{}.svg".format(args.cohort, args.gene,
args.domains)),
dpi=350,
bbox_inches='tight',
format='svg')
plt.close()
def add_data(self, dim_members, dim_run, key={}):
pts = dim_run['pts']
axes_labels = dim_run['axes']
x_comp, y_comp = self.xLoc, self.yLoc + 1
legend = {}
while True:
self.axes.append(
plt.subplot2grid((self.xLen, self.yLen),
(self.xLoc, self.yLoc),
rowspan=1,
colspan=1))
pt_id, pt_color, pt_mark = '', 'w', 'o'
for p, m in zip(pts, dim_members):
if 'color' in m.notes:
pt_id = m.notes['labels']
pt_color = m.notes['color']
if pt_id != 'NA':
if 'size' in key:
self.axes[-1].scatter(p[x_comp],
p[y_comp],
color=pt_color,
alpha=0.6,
s=key['size'])
elif m.notes['size']:
self.axes[-1].scatter(p[x_comp],
p[y_comp],
color=pt_color,
s=m.notes['size'],
alpha=0.6)
else:
self.axes[-1].scatter(p[x_comp],
p[y_comp],
color=pt_color,
alpha=0.85)
if pt_id: legend[pt_id[0]] = (pt_color, pt_mark)
if self.options.plotnames:
if 'color' in m.notes:
self.axes[-1].text(p[x_comp],
p[y_comp],
m.name.split(';')[-1],
color='white')
else:
self.axes[-1].text(p[x_comp],
p[y_comp],
m.name.split(';')[-1],
color='cyan')
if 'zoom' in key:
pX = sorted([p[x_comp] for p in pts])
pY = sorted([p[y_comp] for p in pts])
xQ1, xQ3 = int(len(pX) * 0.25), int(len(pX) * 0.75)
yQ1, yQ3 = int(len(pY) * 0.25), int(len(pY) * 0.75)
iqX = pX[xQ3] - pX[xQ1]
iqY = pY[yQ3] - pY[yQ1]
self.axes[-1].set_xlim(pX[xQ1] - (iqX * 2.5),
pX[xQ3] + (iqX * 2.5))
self.axes[-1].set_ylim(pY[yQ1] - (iqY * 2.5),
pY[yQ3] + (iqY * 2.5))
self.axes[-1].set_xlabel(axes_labels[x_comp], fontweight='bold')
self.axes[-1].set_ylabel(axes_labels[y_comp], fontweight='bold')
self.axes[-1].set_xticks([])
self.axes[-1].set_yticks([])
if self.yLoc + 1 == self.yLen:
self.xLoc, self.yLoc = self.xLoc + 1, 0
else:
self.yLoc += 1
if y_comp == x_comp + 1: y_comp += 1
else: x_comp += 1
# x_comp, y_comp = x_comp +2, y_comp +2
if self.xLoc == self.xLen:
break
if 'title' in key:
plt.suptitle(key['title'], fontsize=20, fontweight='bold')
items, labels = [
Rect((0, 0), 1, 1, fc=b[0]) for a, b in legend.items() if a != 'NA'
], [a for a in legend.keys() if a != 'NA']
self.axes[0].legend(items,
labels,
loc='upper left',
ncol=len(legend.values()),
bbox_to_anchor=(1.7, 1.1))
if 'out' in key: self.fig.savefig(key['out'], dpi=300)
if self.options.show: plt.show()
return self