diff --git a/modules/base/pymod/table.py b/modules/base/pymod/table.py
index e5de1d5bf64f503ef0ae11110a9781636f457d77..04fefecc05bd98600ee91a9bba143ba45cbe93d9 100644
--- a/modules/base/pymod/table.py
+++ b/modules/base/pymod/table.py
@@ -2721,6 +2721,63 @@ Statistics for column %(col)s
except ImportError:
LogError("Function needs numpy, but I could not import it.")
raise
+
+ def ComputeLogROCAUC(self, score_col, class_col, score_dir='-',
+ class_dir='-', class_cutoff=2.0):
+ '''
+ Computes the area under the curve of the log receiver operating
+ characteristics (logROC) where the x-axis is semilogarithmic
+ using the trapezoidal rule.
+
+ The logROC is computed with a lambda of 0.001 according to
+ Rapid Context-Dependent Ligand Desolvation in Molecular Docking
+ Mysinger M. and Shoichet B., Journal of Chemical Information and Modeling
+ 2010 50 (9), 1561-1573
+
+ For more information about parameters of the ROC, see
+ :meth:`ComputeROC`.
+
+ :warning: The function depends on *numpy*
+ '''
+ try:
+ import numpy as np
+
+ roc = self.ComputeROC(score_col, class_col, score_dir,
+ class_dir, class_cutoff)
+
+ if not roc:
+ return None
+
+ rocxt, rocyt = roc
+ rocx=[]
+ rocy=[]
+
+ # define lambda
+ l=0.001
+
+ # remove all duplicate x-values
+ rocxt = [x if x>0 else l for x in rocxt]
+ for i in range(len(rocxt)-1):
+ if rocxt[i]==rocxt[i+1]:
+ continue
+ rocx.append(rocxt[i])
+ rocy.append(rocyt[i])
+ rocx.append(1.0)
+ rocy.append(1.0)
+
+ # compute logauc
+ value = 0
+ for i in range(len(rocx)-1):
+ x = rocx[i]
+ if rocx[i]==rocx[i+1]:
+ continue
+ b = rocy[i+1]-rocx[i+1]*((rocy[i+1]-rocy[i])/(rocx[i+1]-rocx[i]))
+ value += ((rocy[i+1]-rocy[i])/math.log(10))+b*(math.log10(rocx[i+1])-math.log10(rocx[i]))
+ return value/math.log10(1.0/l)
+
+ except ImportError:
+ LogError("Function needs numpy, but I could not import it.")
+ raise
def PlotROC(self, score_col, class_col, score_dir='-',
class_dir='-', class_cutoff=2.0,
@@ -2772,6 +2829,63 @@ Statistics for column %(col)s
LogError("Function needs matplotlib, but I could not import it.")
raise
+ def PlotLogROC(self, score_col, class_col, score_dir='-',
+ class_dir='-', class_cutoff=2.0,
+ style='-', title=None, x_title=None, y_title=None,
+ clear=True, save=None):
+ '''
+ Plot an logROC curve where the x-axis is semilogarithmic using matplotlib
+
+ For more information about parameters of the ROC, see
+ :meth:`ComputeROC`, and for plotting see :meth:`Plot`.
+
+ :warning: The function depends on *matplotlib*
+ '''
+
+ try:
+ import matplotlib.pyplot as plt
+
+ roc = self.ComputeROC(score_col, class_col, score_dir,
+ class_dir, class_cutoff)
+
+ if not roc:
+ return None
+
+ rocx, rocy = roc
+
+ if not title:
+ title = 'logROC of %s'%score_col
+
+ if not x_title:
+ x_title = 'false positive rate'
+
+ if not y_title:
+ y_title = 'true positive rate'
+
+ if clear:
+ plt.clf()
+
+ rocx = [x if x>0 else 0.001 for x in rocx]
+
+
+ plt.plot(rocx, rocy, style)
+
+ plt.title(title, size='x-large', fontweight='bold')
+ plt.ylabel(y_title, size='x-large')
+ plt.xlabel(x_title, size='x-large')
+
+ plt.xscale('log', basex=10)
+ plt.xlim(0.001, 1.0)
+
+
+ if save:
+ plt.savefig(save)
+
+ return plt
+ except ImportError:
+ LogError("Function needs matplotlib, but I could not import it.")
+ raise
+
def ComputeMCC(self, score_col, class_col, score_dir='-',
class_dir='-', score_cutoff=2.0, class_cutoff=2.0):
'''
diff --git a/modules/base/tests/test_table.py b/modules/base/tests/test_table.py
index 818fcc16b82d7ff7683df5f31891e83f274e246a..e18e3092f3c149781b430d1a82196b6c7774bfcf 100644
--- a/modules/base/tests/test_table.py
+++ b/modules/base/tests/test_table.py
@@ -32,7 +32,7 @@ try:
except ImportError:
HAS_SCIPY_NDIMG=False
print "Could not find scipy.ndimage: ignoring some table class unit tests"
-
+
try:
import matplotlib
@@ -1286,6 +1286,28 @@ class TestTable(unittest.TestCase):
save=os.path.join("testfiles","roc-out.png"))
self.assertEquals(pl, None)
+ def testPlotLogROC(self):
+ if not HAS_MPL or not HAS_PIL:
+ return
+ tab = Table(['classific', 'score'], 'bf',
+ classific=[True, True, False, True, True, True, False, False, True, False, True, False, True, False, False, False, True, False, True, False],
+ score=[0.9, 0.8, 0.7, 0.6, 0.55, 0.54, 0.53, 0.52, 0.51, 0.505, 0.4, 0.39, 0.38, 0.37, 0.36, 0.35, 0.34, 0.33, 0.30, 0.1])
+ pl = tab.PlotLogROC(score_col='score', score_dir='+',
+ class_col='classific',
+ save=os.path.join("testfiles","logroc-out.png"))
+ img1 = Image.open(os.path.join("testfiles","logroc-out.png"))
+ #img2 = Image.open(os.path.join("testfiles","roc.png"))
+ #self.CompareImages(img1, img2)
+
+ # no true positives
+ tab = Table(['classific', 'score'], 'bf',
+ classific=[False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False],
+ score=[0.9, 0.8, 0.7, 0.6, 0.55, 0.54, 0.53, 0.52, 0.51, 0.505, 0.4, 0.39, 0.38, 0.37, 0.36, 0.35, 0.34, 0.33, 0.30, 0.1])
+ pl = tab.PlotLogROC(score_col='score', score_dir='+',
+ class_col='classific',
+ save=os.path.join("testfiles","logroc-out.png"))
+ self.assertEquals(pl, None)
+
def testPlotROCSameValues(self):
if not HAS_MPL or not HAS_PIL:
return
@@ -1334,6 +1356,40 @@ class TestTable(unittest.TestCase):
score=[0.9, 0.8, 0.7, 0.6, 0.55, 0.54, 0.53, 0.52, 0.51, 0.505, 0.4, 0.39, 0.38, 0.37, 0.36, 0.35, 0.34, 0.33, 0.30, 0.1])
auc = tab.ComputeROCAUC(score_col='score', score_dir='+', class_col='classific')
self.assertEquals(auc, None)
+
+ def testLogROCAUCforPerfectCurve(self):
+ if not HAS_NUMPY:
+ return
+ auc_ref = 1.0
+ tab = Table(['classific', 'score'], 'bf',
+ classific=[True, True, True, True, True, True, False, False, False, False, False, False],
+ score=[0.9, 0.8, 0.7, 0.6, 0.55, 0.54, 0.4, 0.39, 0.38, 0.37, 0.36, 0.35])
+
+ # test logAUC
+ auc = tab.ComputeLogROCAUC(score_col='score', score_dir='+', class_col='classific')
+ self.assertAlmostEquals(auc, auc_ref)
+
+ # test linear AUC
+ auc = tab.ComputeROCAUC(score_col='score', score_dir='+', class_col='classific')
+ self.assertAlmostEquals(auc, auc_ref)
+
+ def testCalcLogROCAUCRandomCurve(self):
+ if not HAS_NUMPY:
+ return
+ tab = Table(['classific', 'score'], 'bf',
+ classific=[True, False, True, False, True, False, True, False, True, False, True, False],
+ score=[0.9, 0.9, 0.7, 0.7, 0.55, 0.55, 0.4, 0.4, 0.3, 0.3, 0.3, 0.3])
+
+ # test logAUC
+ auc_ref = 0.1440197405305
+ auc = tab.ComputeLogROCAUC(score_col='score', score_dir='+', class_col='classific')
+ self.assertAlmostEquals(auc, auc_ref)
+
+ # test linear AUC
+ auc_ref = 0.5
+ auc = tab.ComputeROCAUC(score_col='score', score_dir='+', class_col='classific')
+ self.assertAlmostEquals(auc, auc_ref)
+
def testCalcROCAUCWithCutoff(self):
if not HAS_NUMPY: