def plotFigure(score_data=None,release=True):
pl.ion()
font_size = 13
mpl.rcParams['font.family']="sans-serif"
mpl.rcParams['font.sans-serif']="Arial"
mpl.rcParams['font.size']=font_size
mpl.rcParams['font.weight']='medium'
mpl.rcParams['figure.dpi'] = 150
mpl.rcParams['figure.facecolor'] = 'white'
mpl.rcParams['lines.linewidth'] = 2
mpl.rcParams['axes.facecolor'] = 'white'
mpl.rcParams['patch.edgecolor'] = 'black'
color_t = ['#F7977A','#FDC68A','#A2D39C','#6ECFF6','#8493CA','#BC8DBF','#F6989D',
'#FFF79A','#998675','#A4A4A4','#5AFF00','#29A3A3','#F53DD6','#F2800D','#3399FF']
#load data
val = utils.Validation()
datasets = ['humvar','exovar','varibench_selected','predictSNP_selected','swissvar_selected']
fig_names = ['FigureS14','FigureS15','FigureS16','FigureS17','FigureS18']
for i,dataset in enumerate(datasets):
score_data.selectDataset(dataset)
labels = score_data.getTrueLabels()
print "\tCreating ROC and ROC-PR curves for dataset: " + dataset + " (" + fig_names[i] + ")"
legend = []
pl.figure(figsize=(12,6))
fig = pl.subplot(121)
pl.grid(True)
fig.set_xlim([0,1])
fig.set_ylim([0,1])
spines_to_remove = ['top','right']
ax = fig.get_axes()
for spine in spines_to_remove:
ax.spines[spine].set_visible(False)
ax.xaxis.set_ticks_position('bottom')
ax.yaxis.set_ticks_position('left')
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('fathmm_w')),
name="FatHMM-W",
legend=legend,
color=color_t[5])
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('logit_condel_new')),
name="Logit+",
legend=legend,
color=color_t[14])
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('condel_new')),
name="Condel+",
legend=legend,
color=color_t[13])
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('logit_condel_old')),
name="Logit",
legend=legend,
color=color_t[12])
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('condel_old')),
name="Condel",
legend=legend,
color=color_t[11])
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('polyphen2')),
name="PolyPhen-2",
legend=legend,
color=color_t[1])
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('mutationassessor')),
name="MutationAssessor",
legend=legend,
color=color_t[8])
legend = plot_roc(results=val.getROCStats(labels,score_data.getScores('sift')),
name="SIFT",
legend=legend,
color=color_t[2])
leg = ax.legend(legend,'lower right',numpoints=1,prop={'size':12},fancybox=True)
leg.get_frame().set_alpha(0.5)
fig.set_xlabel("False Positive Rate or (1-Specificity)")
fig.set_ylabel("True Positive Rate or (Sensitivity)")
fig.plot([0,1],[0,1],'--',color='#ACACAC')
fig.text(-0.1,1.05,"a",fontsize=14,fontweight='bold',va='top',transform=fig.transAxes)
fig = pl.subplot(122)
legend = []
pl.grid(True)
fig.set_xlim([0,1])
fig.set_ylim([0,1])
spines_to_remove = ['top','right']
ax = fig.get_axes()
for spine in spines_to_remove:
ax.spines[spine].set_visible(False)
ax.xaxis.set_ticks_position('bottom')
ax.yaxis.set_ticks_position('left')
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('fathmm_w')),
name="FatHMM-W",
legend=legend,
color=color_t[5])
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('logit_condel_new')),
name="Logit+",
legend=legend,
color=color_t[14])
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('condel_new')),
name="Condel+",
legend=legend,
color=color_t[13])
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('logit_condel_old')),
name="Logit",
legend=legend,
color=color_t[12])
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('condel_old')),
name="Condel",
legend=legend,
color=color_t[11])
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('polyphen2')),
name="PolyPhen-2",
legend=legend,
color=color_t[1])
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('mutationassessor')),
name="MutationAssessor",
legend=legend,
color=color_t[8])
legend = plot_roc_pr(results=val.getROCStats(labels,score_data.getScores('sift')),
name="SIFT",
legend=legend,
color=color_t[2])
leg = ax.legend(legend,'lower right',numpoints=1,prop={'size':12},fancybox=True)
leg.get_frame().set_alpha(0.5)
fig.set_xlabel("Recall")
fig.set_ylabel("Precision")
fig.text(-0.1,1.05,"b",fontsize=14,fontweight='bold',va='top',transform=fig.transAxes)
pl.subplots_adjust(left=0.05,bottom=0.08,right=0.99,top=0.93,wspace=0.11)
if release:
pl.savefig(os.path.abspath('Output/Supplementary/' + fig_names[i] + '.pdf'))
else:
pl.savefig(os.path.abspath('Output/Supplementary/' + fig_names[i] + '.pdf'))
pl.savefig(os.path.abspath('Output/Supplementary/' + fig_names[i] + '.tiff'),dpi=300)
pl.savefig(os.path.abspath('Output/Supplementary/' + fig_names[i] + '.jpg'))
pl.close()
def plotFigure(score_data=None, release=True):
pl.ion()
font_size = 10
mpl.rcParams['font.family'] = "sans-serif"
mpl.rcParams['font.sans-serif'] = "Arial"
mpl.rcParams['font.size'] = font_size
mpl.rcParams['font.weight'] = 'medium'
mpl.rcParams['figure.dpi'] = 150
mpl.rcParams['figure.facecolor'] = 'white'
mpl.rcParams['lines.linewidth'] = 2
mpl.rcParams['axes.facecolor'] = 'white'
mpl.rcParams['patch.edgecolor'] = 'black'
color_t = [
'#F7977A', '#FDC68A', '#A2D39C', '#6ECFF6', '#8493CA', '#BC8DBF',
'#F6989D', '#FFF79A', '#998675', '#A4A4A4'
]
val = utils.Validation()
categories = ['cat1', 'cat2', 'cat3', 'cat4', 'cat5']
n = len(categories)
mt = sp.zeros(n)
pp2 = sp.zeros(n)
mass = sp.zeros(n)
cadd = sp.zeros(n)
sift = sp.zeros(n)
lrt = sp.zeros(n)
fathmmu = sp.zeros(n)
fathmmw = sp.zeros(n)
gerp = sp.zeros(n)
phylop = sp.zeros(n)
score_data.selectDataset('varibench_selected')
score_data.loadCategories()
for i, cat in enumerate(categories):
if cat == "all":
labels = score_data.getTrueLabels()
mt[i] = val.getROCStats(
labels, score_data.getScores('mutationtaster'))['auc']
pp2[i] = val.getROCStats(labels,
score_data.getScores('polyphen2'))['auc']
mass[i] = val.getROCStats(
labels, score_data.getScores('mutationassessor'))['auc']
cadd[i] = val.getROCStats(labels,
score_data.getScores('CADD'))['auc']
sift[i] = val.getROCStats(labels,
score_data.getScores('sift'))['auc']
fathmmu[i] = val.getROCStats(
labels, score_data.getScores('fathmm_u'))['auc']
fathmmw[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
gerp[i] = val.getROCStats(labels,
score_data.getScores('gerp++'))['auc']
phylop[i] = val.getROCStats(labels,
score_data.getScores('phylop'))['auc']
lrt[i] = val.getROCStats(labels,
score_data.getScores('lrt'))['auc']
else:
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='mutationtaster')
mt[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='polyphen2')
pp2[i] = val.getROCStats(labels, scores)['auc']
[labels, scores
] = score_data.getData4Categorie(category=cat,
tool_name='mutationassessor')
mass[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='CADD')
cadd[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='sift')
sift[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='fathmm_u')
fathmmu[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='fathmm_w')
fathmmw[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='gerp++')
gerp[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='phylop')
phylop[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='lrt')
lrt[i] = val.getROCStats(labels, scores)['auc']
pl.figure(figsize=(12, 8))
fig = pl.subplot(311)
font = FontProperties()
font.set_weight('bold')
pl.plot(fathmmw, 'o-', color=color_t[0])
pl.plot(fathmmu, 'h-', color=color_t[1])
pl.plot(mt, 'x-', color=color_t[2])
pl.plot(mass, '<-', color=color_t[8])
pl.plot(pp2, '>-', color=color_t[3])
pl.plot(cadd, 'D-', color=color_t[9])
pl.plot(sift, '1-', color=color_t[4])
pl.plot(lrt, '2-', color=color_t[5])
pl.plot(gerp, '3-', color=color_t[6])
pl.plot(phylop, '4-', color=color_t[7])
pl.ylim(0.53, 0.7)
pl.xlim(-.1, 4.1)
pl.yticks([0.53, 0.6, 0.7])
pl.grid(axis='y')
pl.ylabel("AUC")
pl.xticks(sp.arange(5), [
']0.0,1.0[',
'[0.1,0.9]',
'[0.2,0.8]',
'[0.3,0.7]',
'[0.4,0.6]',
],
fontsize=font_size,
rotation=90)
leg = pl.legend([
'FatHMM-W', 'FatHMM-U', 'MT2', 'MASS', 'PP2', 'CADD', 'SIFT', 'LRT',
'GERP++', 'pyhloP'
],
loc='upper right',
fancybox=True,
ncol=5,
prop={'size': 10},
numpoints=1)
leg.get_frame().set_alpha(0.2)
leg.get_frame().set_edgecolor("none")
fig.text(-0.05,
1.02,
"a",
fontsize=15,
fontweight="bold",
va="top",
transform=fig.transAxes)
pl.title("VariBenchSelected")
remove_border()
mt = sp.zeros(n)
pp2 = sp.zeros(n)
mass = sp.zeros(n)
cadd = sp.zeros(n)
sift = sp.zeros(n)
lrt = sp.zeros(n)
fathmmu = sp.zeros(n)
fathmmw = sp.zeros(n)
gerp = sp.zeros(n)
phylop = sp.zeros(n)
score_data.selectDataset('predictSNP_selected')
score_data.loadCategories()
for i, cat in enumerate(categories):
if cat == "all":
labels = score_data.getTrueLabels()
mt[i] = val.getROCStats(
labels, score_data.getScores('mutationtaster'))['auc']
pp2[i] = val.getROCStats(labels,
score_data.getScores('polyphen2'))['auc']
mass[i] = val.getROCStats(
labels, score_data.getScores('mutationassessor'))['auc']
cadd[i] = val.getROCStats(labels,
score_data.getScores('CADD'))['auc']
sift[i] = val.getROCStats(labels,
score_data.getScores('sift'))['auc']
fathmmu[i] = val.getROCStats(
labels, score_data.getScores('fathmm_u'))['auc']
fathmmw[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
gerp[i] = val.getROCStats(labels,
score_data.getScores('gerp++'))['auc']
phylop[i] = val.getROCStats(labels,
score_data.getScores('phylop'))['auc']
lrt[i] = val.getROCStats(labels,
score_data.getScores('lrt'))['auc']
else:
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='mutationtaster')
mt[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='polyphen2')
pp2[i] = val.getROCStats(labels, scores)['auc']
[labels, scores
] = score_data.getData4Categorie(category=cat,
tool_name='mutationassessor')
mass[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='CADD')
cadd[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='sift')
sift[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='fathmm_u')
fathmmu[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='fathmm_w')
fathmmw[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='gerp++')
gerp[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='phylop')
phylop[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='lrt')
lrt[i] = val.getROCStats(labels, scores)['auc']
fig = pl.subplot(312)
font = FontProperties()
font.set_weight('bold')
pl.plot(fathmmw, 'o-', color=color_t[0])
pl.plot(fathmmu, 'h-', color=color_t[1])
pl.plot(mt, 'x-', color=color_t[2])
pl.plot(mass, '<-', color=color_t[8])
pl.plot(pp2, '>-', color=color_t[3])
pl.plot(cadd, 'D-', color=color_t[9])
pl.plot(sift, '1-', color=color_t[4])
pl.plot(lrt, '2-', color=color_t[5])
pl.plot(gerp, '3-', color=color_t[6])
pl.plot(phylop, '4-', color=color_t[7])
pl.ylim(0.55, 0.8)
pl.xlim(-.1, 4.1)
pl.yticks([0.55, 0.6, 0.7, 0.8])
pl.grid(axis='y')
pl.ylabel("AUC")
pl.xticks(sp.arange(5), [
']0.0,1.0[',
'[0.1,0.9]',
'[0.2,0.8]',
'[0.3,0.7]',
'[0.4,0.6]',
],
fontsize=font_size,
rotation=90)
fig.text(-0.05,
1.02,
"b",
fontsize=15,
fontweight="bold",
va="top",
transform=fig.transAxes)
pl.title("predictSNPSelected")
remove_border()
mt = sp.zeros(n)
pp2 = sp.zeros(n)
mass = sp.zeros(n)
cadd = sp.zeros(n)
sift = sp.zeros(n)
lrt = sp.zeros(n)
fathmmu = sp.zeros(n)
fathmmw = sp.zeros(n)
gerp = sp.zeros(n)
phylop = sp.zeros(n)
score_data.selectDataset('swissvar_selected')
score_data.loadCategories()
for i, cat in enumerate(categories):
if cat == "all":
labels = score_data.getTrueLabels()
mt[i] = val.getROCStats(
labels, score_data.getScores('mutationtaster'))['auc']
pp2[i] = val.getROCStats(labels,
score_data.getScores('polyphen2'))['auc']
mass[i] = val.getROCStats(
labels, score_data.getScores('mutationassessor'))['auc']
cadd[i] = val.getROCStats(labels,
score_data.getScores('CADD'))['auc']
sift[i] = val.getROCStats(labels,
score_data.getScores('sift'))['auc']
fathmmu[i] = val.getROCStats(
labels, score_data.getScores('fathmm_u'))['auc']
fathmmw[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
gerp[i] = val.getROCStats(labels,
score_data.getScores('gerp++'))['auc']
phylop[i] = val.getROCStats(labels,
score_data.getScores('phylop'))['auc']
lrt[i] = val.getROCStats(labels,
score_data.getScores('lrt'))['auc']
else:
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='mutationtaster')
mt[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='polyphen2')
pp2[i] = val.getROCStats(labels, scores)['auc']
[labels, scores
] = score_data.getData4Categorie(category=cat,
tool_name='mutationassessor')
mass[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='CADD')
cadd[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='sift')
sift[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='fathmm_u')
fathmmu[i] = val.getROCStats(labels, scores)['auc']
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='fathmm_w')
fathmmw[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='gerp++')
gerp[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='phylop')
phylop[i] = val.getROCStats(labels, scores)['auc']
[labels, scores] = score_data.getData4Categorie(category=cat,
tool_name='lrt')
lrt[i] = val.getROCStats(labels, scores)['auc']
fig = pl.subplot(313)
font = FontProperties()
font.set_weight('bold')
pl.plot(fathmmw, 'o-', color=color_t[0])
pl.plot(fathmmu, 'h-', color=color_t[1])
pl.plot(mt, 'x-', color=color_t[2])
pl.plot(mass, '<-', color=color_t[8])
pl.plot(pp2, '>-', color=color_t[3])
pl.plot(cadd, 'D-', color=color_t[9])
pl.plot(sift, '1-', color=color_t[4])
pl.plot(lrt, '2-', color=color_t[5])
pl.plot(gerp, '3-', color=color_t[6])
pl.plot(phylop, '4-', color=color_t[7])
pl.ylim(0.55, 0.73)
pl.xlim(-.1, 4.1)
pl.yticks([0.55, 0.6, 0.7, 0.75])
pl.grid(axis='y')
pl.ylabel("AUC")
pl.xticks(sp.arange(5), [
']0.0,1.0[',
'[0.1,0.9]',
'[0.2,0.8]',
'[0.3,0.7]',
'[0.4,0.6]',
],
fontsize=font_size,
rotation=90)
remove_border()
fig.text(-0.05,
1.02,
"c",
fontsize=15,
fontweight="bold",
va="top",
transform=fig.transAxes)
pl.title("SwissVarSelected")
pl.subplots_adjust(left=0.06,
bottom=0.11,
right=0.98,
top=0.91,
wspace=0.03,
hspace=0.5)
if release:
pl.savefig(os.path.abspath('Output/Supplementary/FigureS12.pdf'))
else:
pl.savefig(os.path.abspath('Output/Supplementary/FigureS12.pdf'))
pl.savefig(os.path.abspath('Output/Supplementary/FigureS12.tiff'),
dpi=300)
pl.savefig(os.path.abspath('Output/Supplementary/FigureS12.jpg'))
pl.close()
def plotFigure(score_data=None,release=True):
pl.ion()
font_size = 13
rc('font',**{'family':'sans-serif','sans-serif':['Arial']})
color_t = ['#F7977A','#FDC68A','#A2D39C','#6ECFF6','#8493CA','#BC8DBF','#F6989D','#FFF79A','#998675','#A4A4A4','#5AFF00','#29A3A3']
hatch = pl.Rectangle((0,0),1,1,fill=None,hatch="///")
circle = pl.Rectangle((0,0),1,1,fill=None,hatch="ooo")
#load data
val = utils.Validation()
mv_predictor = utils.ProteinMajorityVote()
lr_predictor = utils.LogisticRegression()
biased_map = {'mutationtaster':['humvar','exovar','varibench',
'varibench_selected','predictSNP_selected','swissvar_selected'],
'mutationassessor':['humvar','exovar','varibench'],
'polyphen2':['humvar','exovar','varibench'],
'sift':[''],
'fathmm_u':[''],
'fathmm_w':['humvar','exovar','varibench'],
'gerp++':[''],
'phylop':[''],
'CADD':['']
}
datasets = ['humvar','exovar','varibench_selected','predictSNP_selected','swissvar_selected']
n_datasets = len(datasets)
mt_biased = np.zeros(n_datasets)
mt = np.zeros(n_datasets)
mass_biased = np.zeros(n_datasets)
mass = np.zeros(n_datasets)
pp2_biased = np.zeros(n_datasets)
pp2 = np.zeros(n_datasets)
fathmmw = np.zeros(n_datasets)
fathmmw_biased = np.zeros(n_datasets)
fathmmw_type2_biased = np.zeros(n_datasets)
sift = np.zeros(n_datasets)
lrt = np.zeros(n_datasets)
fathmmu = np.zeros(n_datasets)
gerp = np.zeros(n_datasets)
phylop = np.zeros(n_datasets)
cadd = np.zeros(n_datasets)
mv = np.zeros(n_datasets)
mv_biased = np.zeros(n_datasets)
features = np.zeros(n_datasets)
features_biased = np.zeros(n_datasets)
features_type2_biased = np.zeros(n_datasets)
for i,dataset in enumerate(datasets):
score_data.selectDataset(dataset)
labels = score_data.getTrueLabels()
print "\tPerforming a Logistic Regression over the weighting features of FatHMM-W for dataset: " + dataset
lr_value = lr_predictor.run(true_labels=labels,features=score_data.getFatHMMFeatures(),folds=10)['auc']
print "\tComputing AUC values for dataset: " + dataset
if dataset in biased_map['mutationtaster']:
mt_biased[i] = val.getROCStats(labels,score_data.getScores('mutationtaster'))['auc']
else:
mt[i] = val.getROCStats(labels,score_data.getScores('mutationtaster'))['auc']
if dataset in biased_map['mutationassessor']:
mass_biased[i] = val.getROCStats(labels,score_data.getScores('mutationassessor'))['auc']
else:
mass[i] = val.getROCStats(labels,score_data.getScores('mutationassessor'))['auc']
if dataset in biased_map['polyphen2']:
pp2_biased[i] = val.getROCStats(labels,score_data.getScores('polyphen2'))['auc']
else:
pp2[i] = val.getROCStats(labels,score_data.getScores('polyphen2'))['auc']
if dataset in biased_map['fathmm_w']:
fathmmw_biased[i] = val.getROCStats(labels,score_data.getScores('fathmm_w'))['auc']
features_biased[i] = lr_value
else:
if dataset == "varibench_selected" or dataset=="predictSNP_selected" or dataset=="swissvar_selected":
fathmmw_type2_biased[i] = val.getROCStats(labels,score_data.getScores('fathmm_w'))['auc']
features_type2_biased[i] = lr_value
else:
fathmmw[i] = val.getROCStats(labels,score_data.getScores('fathmm_w'))['auc']
features[i] = lr_value
fathmmu[i] = val.getROCStats(labels,score_data.getScores('fathmm_u'))['auc']
sift[i] = val.getROCStats(labels,score_data.getScores('sift'))['auc']
lrt[i] = val.getROCStats(labels,score_data.getScores('lrt'))['auc']
gerp[i] = val.getROCStats(labels,score_data.getScores('gerp++'))['auc']
phylop[i] = val.getROCStats(labels,score_data.getScores('phylop'))['auc']
cadd[i] = val.getROCStats(labels,score_data.getScores('CADD'))['auc']
print "\tPerforming a Protein Majority Vote for dataset: " + dataset
mv_biased[i] = mv_predictor.getMV4Dataset(true_labels=labels,proteins=score_data.getUniprotIDs(),folds=10)['auc']
pl.figure(figsize=(15,5))
fig1 = pl.subplot(111)
width = 0.05
x=np.arange(n_datasets)
tool_names = np.array(['FatHMM-W','MutationTaster-2','PolyPhen-2','MutationAssessor','CADD','SIFT','LRT','FatHMM-U','Gerp++','phyloP','Features ln(Wn), ln(Wd)','Protein Majority Vote','Potentially Type 1 Biased','Potentially Type 2 Biased'])
spines_to_remove = ['top','right','bottom']
ax = fig1.get_axes()
for spine in spines_to_remove:
ax.spines[spine].set_visible(False)
ax.xaxis.set_ticks_position('none')
ax.yaxis.set_ticks_position('left')
ax.yaxis.grid(True)
t0 = fig1.bar(x-width/2.0-5*width-0.1,fathmmw,width=width,color=color_t[5])
t1 = fig1.bar(x-width/2.0-4*width-0.08,mt,width=width,color=color_t[0])
t2 = fig1.bar(x-width/2.0-3*width-0.06,pp2,width=width,color=color_t[1])
t3 = fig1.bar(x-width/2.0-2*width-0.04,mass,width=width,color=color_t[8])
t4 = fig1.bar(x-width/2.0-width-0.02,cadd,width=width,color=color_t[9])
t5 = fig1.bar(x-width/2.0,sift,width=width,color=color_t[2])
t6 = fig1.bar(x+width/2.0+0.02,lrt,width=width,color=color_t[3])
t7 = fig1.bar(x+width/2.0+width+0.04,fathmmu,width=width,color=color_t[4])
t8 = fig1.bar(x+width/2.0+2*width+0.06,gerp,width=width,color=color_t[6])
t9 = fig1.bar(x+width/2.0+3*width+0.08,phylop,width=width,color=color_t[7])
t10 = fig1.bar(x+width/2.0+4*width+0.1,features,width=width,color=color_t[11])
t11 = fig1.bar(x+width/2.0+5*width+0.12,mv,width=width,color=color_t[10])
#fig1.text(-0.05,1.02,"b",fontsize=15,fontweight="bold",va="top",transform=fig1.transAxes)
light_grey = np.array([float(248)/float(255)]*3)
light_grey = "#FFFFFF"
almost_black = '#262626'
legend = ax.legend([t0,t1,t2,t3,t4,t5,t6,t7,t8,t9,t10,t11,hatch,circle],tool_names,frameon=True, scatterpoints=1,
prop={'size':11},ncol=7,loc="upper center",fancybox=True,bbox_to_anchor=(0.5, 1.02))
legend.get_frame().set_alpha(0.5)
rect = legend.get_frame()
rect.set_facecolor(light_grey)
rect.set_linewidth(0.0)
# Change the legend label colors to almost black, too
texts = legend.texts
for t in texts:
t.set_color(almost_black)
fig1.bar(x-width/2.0-5*width-0.1,fathmmw_biased,width=width,color=color_t[5],hatch="/o/o/")
fig1.bar(x-width/2.0-5*width-0.1,fathmmw_type2_biased,width=width,color=color_t[5],hatch="ooo")
fig1.bar(x-width/2.0-4*width-0.08,mt_biased,width=width,color=color_t[0],hatch="///")
fig1.bar(x-width/2.0-3*width-0.06,pp2_biased,width=width,color=color_t[1],hatch="///")
fig1.bar(x-width/2.0-2*width-0.04,mass_biased,width=width,color=color_t[8],hatch="///")
fig1.bar(x+width/2.0+4*width+0.1,features_biased,width=width,color=color_t[11],hatch="/o/o/")
fig1.bar(x+width/2.0+4*width+0.1,features_type2_biased,width=width,color=color_t[11],hatch="ooo")
fig1.bar(x+width/2.0+5*width+0.12,mv_biased,width=width,color=color_t[10],hatch="ooo")
pl.xticks(x,['HumVar','ExoVar','VariBenchSelected','predictSNPSelected','SwissVarSelected'],fontsize=font_size)
fig1.set_ylabel("AUC")
fig1.set_ylim(0.5,1.06)
fig1.set_xlim(-0.5,n_datasets-0.5)
pl.yticks([0.45,0.5,0.55,0.6,0.65,0.7,0.75,0.8,0.85,0.9,0.95,1.0])
pl.subplots_adjust(left=0.06,bottom=0.07,right=0.99,top=0.99,wspace=0.05)
if release:
pl.savefig(os.path.abspath('Output/Figures/Figure1.tiff'),dpi=300)
else:
pl.savefig(os.path.abspath('Output/Figures/Figure1.pdf'))
pl.savefig(os.path.abspath('Output/Figures/Figure1.tiff'),dpi=300)
pl.savefig(os.path.abspath('Output/Figures/Figure1.jpg'))
pl.close()
def plotFigure(score_data=None, release=True):
pl.ion()
font_size = 13
mpl.rcParams['font.family'] = "sans-serif"
mpl.rcParams['font.sans-serif'] = "Arial"
mpl.rcParams['font.size'] = font_size
mpl.rcParams['font.weight'] = 'medium'
mpl.rcParams['figure.dpi'] = 150
mpl.rcParams['figure.facecolor'] = 'white'
mpl.rcParams['lines.linewidth'] = 2
mpl.rcParams['axes.facecolor'] = 'white'
mpl.rcParams['font.size'] = font_size
mpl.rcParams['patch.edgecolor'] = 'black'
color_t = [
'#F7977A', '#FDC68A', '#A2D39C', '#6ECFF6', '#8493CA', '#BC8DBF',
'#F6989D', '#FFF79A', '#998675', '#A4A4A4', '#5AFF00', '#29A3A3',
'#F53DD6', '#F2800D', '#3399FF'
]
hatch = pl.Rectangle((0, 0), 1, 1, fill=None, hatch="///")
circle = pl.Rectangle((0, 0), 1, 1, fill=None, hatch="ooo")
#load data
val = utils.Validation()
biased_map = {
'mutationtaster': [
'humvar', 'exovar', 'varibench', 'varibench_selected',
'swissvar_selected'
],
'mutationassessor': ['humvar', 'exovar', 'varibench'],
'polyphen2': ['humvar', 'exovar', 'varibench'],
'logit_condel_old': ['humvar', 'exovar', 'varibench'],
'logit_condel_new': ['humvar', 'exovar', 'varibench'],
'condel_old': ['humvar', 'exovar', 'varibench'],
'condel_new': ['humvar', 'exovar', 'varibench'],
'sift': [''],
'fathmm_u': [''],
'fathmm_w': ['humvar', 'exovar', 'varibench'],
'gerp++': [''],
'phylop': [''],
'CADD': ['']
}
datasets = [
'humvar', 'exovar', 'varibench_selected', 'predictSNP_selected',
'swissvar_selected'
]
n = len(datasets)
mass_biased = np.zeros(n)
mass = np.zeros(n)
pp2 = np.zeros(n)
pp2_biased = np.zeros(n)
sift = np.zeros(n)
logit_biased = np.zeros(n)
logit = np.zeros(n)
condel_biased = np.zeros(n)
condel = np.zeros(n)
fathmmw = np.zeros(n)
fathmmw_biased = np.zeros(n)
fathmmw_type2_biased = np.zeros(n)
logit_p_biased = np.zeros(n)
logit_p_type2_biased = np.zeros(n)
logit_p = np.zeros(n)
condel_p_biased = np.zeros(n)
condel_p_type2_biased = np.zeros(n)
condel_p = np.zeros(n)
for i, dataset in enumerate(datasets):
score_data.selectDataset(dataset)
labels = score_data.getTrueLabels()
print "\tComputing AUC values for dataset: " + dataset
if dataset in biased_map['mutationassessor']:
mass_biased[i] = val.getROCStats(
labels, score_data.getScores('mutationassessor'))['auc']
else:
mass[i] = val.getROCStats(
labels, score_data.getScores('mutationassessor'))['auc']
if dataset in biased_map['polyphen2']:
pp2_biased[i] = val.getROCStats(
labels, score_data.getScores('polyphen2'))['auc']
else:
pp2[i] = val.getROCStats(labels,
score_data.getScores('polyphen2'))['auc']
if dataset in biased_map['logit_condel_old']:
logit_biased[i] = val.getROCStats(
labels, score_data.getScores('logit_condel_old'))['auc']
else:
logit[i] = val.getROCStats(
labels, score_data.getScores('logit_condel_old'))['auc']
if dataset in biased_map['condel_old']:
condel_biased[i] = val.getROCStats(
labels, score_data.getScores('condel_old'))['auc']
else:
condel[i] = val.getROCStats(
labels, score_data.getScores('condel_old'))['auc']
if dataset in biased_map['fathmm_w']:
fathmmw_biased[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
else:
if dataset == "varibench_selected" or dataset == "predictSNP_selected" or dataset == "swissvar_selected":
fathmmw_type2_biased[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
else:
fathmmw[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
if dataset in biased_map['condel_new']:
condel_p_biased[i] = val.getROCStats(
labels, score_data.getScores('condel_new'))['auc']
else:
if dataset == "varibench_selected" or dataset == "predictSNP_selected" or dataset == "swissvar_selected":
condel_p_type2_biased[i] = val.getROCStats(
labels, score_data.getScores('condel_new'))['auc']
else:
condel_p[i] = val.getROCStats(
labels, score_data.getScores('condel_new'))['auc']
if dataset in biased_map['logit_condel_new']:
logit_p_biased[i] = val.getROCStats(
labels, score_data.getScores('logit_condel_new'))['auc']
else:
if dataset == "varibench_selected" or dataset == "predictSNP_selected" or dataset == "swissvar_selected":
logit_p_type2_biased[i] = val.getROCStats(
labels, score_data.getScores('logit_condel_new'))['auc']
else:
logit_p[i] = val.getROCStats(
labels, score_data.getScores('logit_condel_new'))['auc']
sift[i] = val.getROCStats(labels, score_data.getScores('sift'))['auc']
pl.figure(figsize=(10, 5))
fig1 = pl.subplot(111)
width = 0.07
x = np.arange(n)
tool_names = np.array([
'FatHMM-W', 'PolyPhen-2', 'MutationAssessor', 'SIFT', 'Condel',
'Logit', 'Condel+', 'Logit+', 'Type 1 Biased', 'Type 2 Biased'
])
spines_to_remove = ['top', 'right', 'bottom']
ax = fig1.get_axes()
for spine in spines_to_remove:
ax.spines[spine].set_visible(False)
ax.xaxis.set_ticks_position('none')
ax.yaxis.set_ticks_position('left')
ax.yaxis.grid(True)
t0 = fig1.bar(x - width / 2.0 - 3 * width - 0.06,
fathmmw,
width=width,
color=color_t[5])
t1 = fig1.bar(x - width / 2.0 - 2 * width - 0.04,
pp2,
width=width,
color=color_t[1])
t2 = fig1.bar(x - width / 2.0 - width - 0.02,
mass,
width=width,
color=color_t[8])
t3 = fig1.bar(x - width / 2.0, sift, width=width, color=color_t[2])
t4 = fig1.bar(x + width / 2.0 + 0.02,
condel,
width=width,
color=color_t[11])
t5 = fig1.bar(x + width / 2.0 + width + 0.04,
logit,
width=width,
color=color_t[12])
t6 = fig1.bar(x + width / 2.0 + 2.0 * width + 0.06,
condel_p,
width=width,
color=color_t[13])
t7 = fig1.bar(x + width / 2.0 + 3.0 * width + 0.08,
logit_p,
width=width,
color=color_t[14])
light_grey = np.array([float(248) / float(255)] * 3)
light_grey = "#FFFFFF"
almost_black = '#262626'
legend = ax.legend([t0, t1, t2, t3, t4, t5, t6, t7, hatch, circle],
tool_names,
frameon=True,
scatterpoints=1,
prop={'size': font_size},
ncol=5,
loc="upper center",
fancybox=True,
bbox_to_anchor=(0.5, 1.0))
rect = legend.get_frame()
rect.set_facecolor(light_grey)
rect.set_linewidth(0.0)
texts = legend.texts
for t in texts:
t.set_color(almost_black)
fig1.bar(x - width / 2.0 - 3 * width - 0.06,
fathmmw_biased,
width=width,
color=color_t[5],
hatch="/o/o/")
fig1.bar(x - width / 2.0 - 3 * width - 0.06,
fathmmw_type2_biased,
width=width,
color=color_t[5],
hatch="ooo")
fig1.bar(x - width / 2.0 - 2 * width - 0.04,
pp2_biased,
width=width,
color=color_t[1],
hatch="///")
fig1.bar(x - width / 2.0 - width - 0.02,
mass_biased,
width=width,
color=color_t[8],
hatch="///")
fig1.bar(x + width / 2.0 + 0.02,
condel_biased,
width=width,
color=color_t[11],
hatch="///")
fig1.bar(x + width / 2.0 + width + 0.04,
logit_biased,
width=width,
color=color_t[12],
hatch="///")
fig1.bar(x + width / 2.0 + 2.0 * width + 0.06,
condel_p_biased,
width=width,
color=color_t[13],
hatch="/o/o/")
fig1.bar(x + width / 2.0 + 3.0 * width + 0.08,
logit_p_type2_biased,
width=width,
color=color_t[14],
hatch="ooo")
fig1.bar(x + width / 2.0 + 2.0 * width + 0.06,
condel_p_type2_biased,
width=width,
color=color_t[13],
hatch="ooo")
fig1.bar(x + width / 2.0 + 3.0 * width + 0.08,
logit_p_biased,
width=width,
color=color_t[14],
hatch="/o/o/")
pl.xticks(x, [
'HumVar', 'ExoVar', 'VaribenchSelected', 'predictSNPSelected',
'SwissVarSelected'
],
fontsize=font_size)
pl.ylabel("AUC")
fig1.set_ylim(0.5, 1.03)
fig1.set_xlim(-0.5, n - 0.5)
fig1.text(-0.05,
1.02,
"a",
fontsize=15,
fontweight="bold",
va="top",
transform=fig1.transAxes)
pl.yticks([0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95, 1.0])
pl.subplots_adjust(left=0.07,
bottom=0.07,
right=0.99,
top=0.97,
wspace=0.05)
if release:
pl.savefig(os.path.abspath('Output/Figures/Figure4.tiff'), dpi=300)
else:
pl.savefig(os.path.abspath('Output/Figures/Figure4.pdf'))
pl.savefig(os.path.abspath('Output/Figures/Figure4.tiff'), dpi=300)
pl.savefig(os.path.abspath('Output/Figures/Figure4.jpg'))
pl.close()
def printTableS1(score_data=None):
val = utils.Validation()
f = open(os.path.abspath("Output/Supplementary/tableS1.csv"), 'w')
f.write(";;MT2;PP2;MASS;CADD;SIFT;LRT;FatHMM-U;FatHMM-W;Gerp++;phyloP\n")
datasets = [
'humvar', 'exovar', 'varibench_selected', 'predictSNP_selected',
'swissvar_selected'
]
for i, dataset in enumerate(datasets):
score_data.selectDataset(dataset)
labels = score_data.getTrueLabels()
string = dataset + ";TP;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('polyphen2'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationassessor'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_u'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_w'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['TP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['TP']
f.write(string + "\n")
string = ";FP;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('polyphen2'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationassessor'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_u'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_w'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['FP'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['FP']
f.write(string + "\n")
string = ";TN;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('polyphen2'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationassessor'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_u'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_w'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['TN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['TN']
f.write(string + "\n")
string = ";FN;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('polyphen2'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationassessor'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_u'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_w'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['FN'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['FN']
f.write(string + "\n")
string = ";AUC;"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('mutationtaster'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('polyphen2'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('mutationassessor'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('CADD'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('sift'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('lrt'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('fathmm_u'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('gerp++'))['auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('phylop'))['auc']
f.write(string + "\n")
string = ";AUC-PR;"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('mutationtaster'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('polyphen2'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('mutationassessor'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('CADD'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('sift'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('lrt'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('fathmm_u'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('fathmm_w'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('gerp++'))['pr_auc'] + ";"
string += "%.2f" % val.getROCStats(
labels, score_data.getScores('phylop'))['pr_auc']
f.write(string + "\n")
string = ";Accuracy;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('polyphen2'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels(
'mutationassessor'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_u'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_w'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['accuracy'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['accuracy']
f.write(string + "\n")
string = ";F-Score;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('polyphen2'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels(
'mutationassessor'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_u'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_w'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['f1_score'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['f1_score']
f.write(string + "\n")
string = ";MCC;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('polyphen2'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationassessor'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_u'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_w'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['mcc'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['mcc']
f.write(string + "\n")
string = ";Precision/Positive Predictive Value;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('polyphen2'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels(
'mutationassessor'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_u'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_w'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['precision'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['precision']
f.write(string + "\n")
string = ";Recall/Sensitivity;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('polyphen2'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationassessor'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_u'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_w'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['recall'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['recall']
f.write(string + "\n")
string = ";Specificity;"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels(
'mutationtaster'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('polyphen2'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels(
'mutationassessor'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_u'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('fathmm_w'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('gerp++'))['specificity'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['specificity']
f.write(string + "\n")
string = ";Negative Predictive Value;"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationtaster'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('polyphen2'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels,
score_data.getPredictedLabels('mutationassessor'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('CADD'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('sift'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('lrt'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_u'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('fathmm_w'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('gerp++'))['npv'] + ";"
string += "%.2f" % val.getPredictionStats(
labels, score_data.getPredictedLabels('phylop'))['npv']
f.write(string + "\n")
f.close()
def plotFigure(score_data=None, release=True):
pl.ion()
font_size = 10
mpl.rcParams['font.family'] = "sans-serif"
mpl.rcParams['font.sans-serif'] = "Arial"
mpl.rcParams['font.size'] = font_size
mpl.rcParams['figure.dpi'] = 150
mpl.rcParams['font.weight'] = 'medium'
mpl.rcParams['figure.facecolor'] = 'white'
mpl.rcParams['lines.linewidth'] = 0.8
mpl.rcParams['axes.facecolor'] = 'white'
mpl.rcParams['font.size'] = font_size
mpl.rcParams['patch.edgecolor'] = 'black'
val = utils.Validation()
categories = ['all', 'pure', 'cat1', 'cat2', 'cat3', 'cat4', 'cat5']
humvar = sp.zeros(7)
exovar = sp.zeros(7)
varibench_selected = sp.zeros(7)
predictSNP_selected = sp.zeros(7)
swissvar_selected = sp.zeros(7)
datasets = [
'humvar', 'exovar', 'varibench_selected', 'predictSNP_selected',
'swissvar_selected'
]
for dataset in datasets:
score_data.selectDataset(dataset)
score_data.loadCategories()
for i, cat in enumerate(categories):
if cat == "all":
labels = score_data.getTrueLabels()
if dataset == 'humvar':
humvar[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
elif dataset == 'exovar':
exovar[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
elif dataset == 'swissvar_selected':
swissvar_selected[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
elif dataset == 'predictSNP_selected':
predictSNP_selected[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
else:
varibench_selected[i] = val.getROCStats(
labels, score_data.getScores('fathmm_w'))['auc']
else:
[labels,
scores] = score_data.getData4Categorie(category=cat,
tool_name='fathmm_w')
if dataset == 'humvar':
humvar[i] = val.getROCStats(labels, scores)['auc']
elif dataset == 'exovar':
exovar[i] = val.getROCStats(labels, scores)['auc']
elif dataset == 'predictSNP_selected':
predictSNP_selected[i] = val.getROCStats(labels,
scores)['auc']
elif dataset == 'swissvar_selected':
swissvar_selected[i] = val.getROCStats(labels,
scores)['auc']
else:
varibench_selected[i] = val.getROCStats(labels,
scores)['auc']
pl.figure(figsize=(10, 7))
font = FontProperties()
font.set_weight('bold')
plotBar(humvar, 0, 321, "HumVar", flag=False)
ax = plotBar(exovar, 1, 322, "ExoVar", flag=True)
plotBar(varibench_selected, 3, 323, "VariBenchSelected", flag=False)
plotBar(predictSNP_selected, 4, 324, "predictSNPSelected", flag=True)
plotBar(swissvar_selected, 5, 325, "SwissVarSelected", flag=False)
rect = pl.Rectangle((0, 0), 1, 1, fill=None)
leg = ax.legend([rect, hatch, circle],
['FatHMM-W', 'Type 1 Biased', 'Type 2 Biased'],
loc='upper center',
bbox_to_anchor=(0.0, 1.15),
fancybox=True,
ncol=5,
prop={'size': 10},
numpoints=1)
leg.get_frame().set_alpha(0.2)
leg.get_frame().set_edgecolor("none")
pl.subplots_adjust(left=0.065,
bottom=0.11,
right=0.99,
top=0.94,
wspace=0.03,
hspace=0.5)
if release:
pl.savefig(os.path.abspath('Output/Supplementary/FigureS11.pdf'))
else:
pl.savefig(os.path.abspath('Output/Supplementary/FigureS11.pdf'))
pl.savefig(os.path.abspath('Output/Supplementary/FigureS11.tiff'),
dpi=300)
pl.savefig(os.path.abspath('Output/Supplementary/FigureS11.jpg'))
pl.close()
def printTableS2(score_data=None):
val = utils.Validation()
mv_predictor = utils.ProteinMajorityVote()
lr_predictor = utils.LogisticRegression()
f = open(os.path.abspath("Output/Supplementary/tableS2.csv"),'w')
f.write(";;FatHMM-W;Logistic Regression over the features ln(Wn) & ln(Wd);Protein Majority Vote (MV)\n")
datasets = ['humvar','exovar','varibench_selected','predictSNP_selected','swissvar_selected']
for i,dataset in enumerate(datasets):
score_data.selectDataset(dataset)
labels = score_data.getTrueLabels()
print "\tTraining Logistic Regression on Features ln(Wn) and ln(Wd)"
lr = lr_predictor.run(true_labels=labels,features=score_data.getFatHMMFeatures(),folds=10)
print "\tPerforming a Protein Majority Vote for dataset: " + dataset
mv = mv_predictor.getMV4Dataset(true_labels=labels,proteins=score_data.getUniprotIDs(),folds=10)
string = dataset + ";AUC;"
string += "%.2f"% val.getROCStats(labels,score_data.getScores('fathmm_w'))['auc'] + ";"
string += "%.2f (%.2f)"% (lr['auc'],lr['auc_std']) + ";"
string += "%.2f (%.2f)"% (mv['auc'],mv['auc_std'])
f.write(string + "\n")
string = ";AUC-PR;"
string += "%.2f"% val.getROCStats(labels,score_data.getScores('fathmm_w'))['pr_auc'] + ";"
string += "%.2f (%.2f)"% (lr['auc_pr'],lr['auc_pr_std']) + ";"
string += "%.2f (%.2f)"% (mv['auc_pr'],mv['auc_pr_std'])
f.write(string + "\n")
string = ";Accuracy;"
string += "%.2f"% val.getPredictionStats(labels,score_data.getPredictedLabels('fathmm_w'))['accuracy'] + ";"
string += "%.2f (%.2f)"% (lr['accuracy'],lr['accuracy_std']) + ";"
string += "%.2f (%.2f)"% (mv['accuracy'],mv['accuracy_std'])
f.write(string + "\n")
string = ";F1-Score;"
string += "%.2f"% val.getPredictionStats(labels,score_data.getPredictedLabels('fathmm_w'))['f1_score'] + ";"
string += "%.2f (%.2f)"% (lr['f1_score'],lr['f1_score_std']) + ";"
string += "%.2f (%.2f)"% (mv['f1_score'],mv['f1_score_std'])
f.write(string + "\n")
string = ";MCC;"
string += "%.2f"% val.getPredictionStats(labels,score_data.getPredictedLabels('fathmm_w'))['mcc'] + ";"
string += "%.2f (%.2f)"% (lr['mcc'],lr['mcc_std']) + ";"
string += "%.2f (%.2f)"% (mv['mcc'],mv['mcc_std'])
f.write(string + "\n")
string = ";Precision;"
string += "%.2f"% val.getPredictionStats(labels,score_data.getPredictedLabels('fathmm_w'))['precision'] + ";"
string += "%.2f (%.2f)"% (lr['precision'],lr['precision_std']) + ";"
string += "%.2f (%.2f)"% (mv['precision'],mv['precision_std'])
f.write(string + "\n")
string = ";Recall;"
string += "%.2f"% val.getPredictionStats(labels,score_data.getPredictedLabels('fathmm_w'))['recall'] + ";"
string += "%.2f (%.2f)"% (lr['recall'],lr['recall_std']) + ";"
string += "%.2f (%.2f)"% (mv['recall'],mv['recall_std'])
f.write(string + "\n")
string = ";Negative Predictive Value;"
string += "%.2f"% val.getPredictionStats(labels,score_data.getPredictedLabels('fathmm_w'))['npv'] + ";"
string += "%.2f (%.2f)"% (lr['npv'],lr['npv_std']) + ";"
string += "%.2f (%.2f)"% (mv['npv'],mv['npv_std'])
f.write(string + "\n")
string = ";Specificity;"
string += "%.2f"% val.getPredictionStats(labels,score_data.getPredictedLabels('fathmm_w'))['specificity'] + ";"
string += "%.2f (%.2f)"% (lr['specificity'],lr['specificity_std']) + ";"
string += "%.2f (%.2f)"% (mv['specificity'],mv['specificity_std'])
f.write(string + "\n")
f.close()
datas = utils.get_datas(PREPROCESSING_TRAIN_DATA_DIR)
datas_y = preprocessing_to_train.select_data_y(datas)
train_x, train_y, val_x, val_y = preprocessing_to_train.classification_data(
data_x=datas, data_y=datas_y)
with tf.Session() as sess:
inputs, labels, keep_prob = utils.build_input()
predictions, cost, optimizer, initial_state, final_state, cell = LSTM.LSTM_cell(
LSTM_SIZE, keep_prob, LSTM_LAYERS, BATCH_SIZE, train_x, labels,
LEARNING_RATE)
accuracy = utils.Validation(predictions, labels)
utils.draw_scalar(cost, 'loss')
utils.draw_scalar(accuracy, 'Batch accurcy')
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter('logs/', sess.graph)
sess.run(tf.global_variables_initializer())
with graph.as_default():
saver = tf.train.Saver()
iteration = 1
