def test1(self):
# simple test with a known answer
cont = TopNContainer(4)
for foo in range(10):
cont.Insert(foo, str(foo))
assert cont.GetPts() == list(range(6, 10))
assert cont.GetExtras() == [str(x) for x in range(6, 10)]
def test_keepAll(self):
# simple test with a known answer where we keep all
cont = TopNContainer(-1)
for i in range(10):
cont.Insert(9 - i, str(9 - i))
self.assertEqual(len(cont), i + 1)
assert cont.GetPts() == list(range(10))
assert cont.GetExtras() == [str(x) for x in range(10)]
def test2(self):
# larger scale random test
cont = TopNContainer(50)
for _ in range(1000):
cont.Insert(random.random())
vs = cont.GetPts()
last = vs.pop(0)
while vs:
assert vs[0] >= last
last = vs.pop(0)
def GetNeighbors(self, example):
""" Returns the k nearest neighbors of the example
"""
nbrs = TopNContainer(self._k)
for trex in self._trainingExamples:
dist = self._dfunc(trex, example, self._attrs)
if self._radius is None or dist < self._radius:
nbrs.Insert(-dist, trex)
nbrs.reverse()
return [x for x in nbrs]
def test3(self):
# random test with extras
cont = TopNContainer(10)
for _ in range(100):
v = random.random()
cont.Insert(v, v + 1)
vs = cont.GetExtras()
last = vs.pop(0)
while vs:
assert vs[0] >= last
last = vs.pop(0)
def GetNeighbors(self, example):
""" Returns the k nearest neighbors of the example
"""
nbrs = TopNContainer(self._k)
for trex in self._trainingExamples:
dist = self._dfunc(trex, example, self._attrs)
if self._radius is None or dist < self._radius:
nbrs.Insert(-dist, trex)
nbrs.reverse()
return [x for x in nbrs]
def test4(self):
# random test with extras and getitem
cont = TopNContainer(10)
for i in range(100):
v = random.random()
cont.Insert(v, v + 1)
lastV, lastE = cont[0]
for i in range(1, len(cont)):
v, e = cont[i]
assert v >= lastV
assert e >= lastE
lastV, lastE = v, e
def test5(self):
""" random test with extras and getitem, include reverse"""
cont = TopNContainer(10)
for i in range(100):
v = random.random()
cont.Insert(v,v+1)
cont.reverse()
lastV,lastE = cont[0]
for i in range(1,len(cont)):
v,e = cont[i]
assert v<=lastV
assert e<=lastE
lastV,lastE = v,e
def test5(self):
# random test with extras and getitem, include reverse
cont = TopNContainer(10)
for i in range(100):
v = random.random()
cont.Insert(v, v + 1)
cont.reverse()
lastV, lastE = cont[0]
for i in range(1, len(cont)):
v, e = cont[i]
assert v <= lastV
assert e <= lastE
lastV, lastE = v, e
def MakePicks(self, force=False):
if self._picks is not None and not force:
return
picks = TopNContainer(self.numToPick)
for fp in self.data:
origFp = fp
bestScore = -1.0
for probeFp in self.probes:
score = DataStructs.FingerprintSimilarity(origFp, probeFp, self.simMetric)
bestScore = max(score, bestScore)
picks.Insert(bestScore, fp)
self._picks = []
for score, pt in picks:
self._picks.append((pt, score))
self._picks.reverse()
def ScreenFingerprints(details, data, mol=None, probeFp=None):
""" Returns a list of results
"""
if probeFp is None:
try:
probeFp = apply(FingerprintMols.FingerprintMol, (mol, ),
details.__dict__)
except:
import traceback
FingerprintMols.error('Error: problems fingerprinting molecule.\n')
traceback.print_exc()
return []
if not probeFp:
return []
res = []
if not details.doThreshold and details.topN > 0:
topN = TopNContainer(details.topN)
else:
topN = []
res = []
count = 0
for pt in data:
fp1 = probeFp
if not details.noPickle:
if type(pt) in (types.TupleType, types.ListType):
id, fp = pt
else:
fp = pt
id = pt._fieldsFromDb[0]
score = DataStructs.FingerprintSimilarity(fp1, fp, details.metric)
else:
id, pkl = pt
score = details.metric(fp1, str(pkl))
if topN:
topN.Insert(score, id)
elif not details.doThreshold or \
(details.doThreshold and score>=details.screenThresh):
res.append((id, score))
count += 1
if hasattr(details, 'stopAfter') and count >= details.stopAfter:
break
for score, id in topN:
res.append((id, score))
return res
def GetNeighborLists(probes,
topN,
pool,
simMetric=DataStructs.DiceSimilarity,
simThresh=-1.,
silent=False,
**kwargs):
probeFps = [x[1] for x in probes]
validProbes = [x for x in range(len(probeFps)) if probeFps[x] is not None]
validFps = [probeFps[x] for x in validProbes]
from rdkit.DataStructs.TopNContainer import TopNContainer
if simThresh <= 0:
nbrLists = [TopNContainer(topN) for x in range(len(probeFps))]
else:
nbrLists = [TopNContainer(-1) for x in range(len(probeFps))]
nDone = 0
for nm, fp in pool:
nDone += 1
if not silent and not nDone % 1000:
logger.info(' searched %d rows' % nDone)
if (simMetric == DataStructs.DiceSimilarity):
scores = DataStructs.BulkDiceSimilarity(fp, validFps)
for i, score in enumerate(scores):
if score > simThresh:
nbrLists[validProbes[i]].Insert(score, nm)
elif (simMetric == DataStructs.TanimotoSimilarity):
scores = DataStructs.BulkTanimotoSimilarity(fp, validFps)
for i, score in enumerate(scores):
if score > simThresh:
nbrLists[validProbes[i]].Insert(score, nm)
elif (simMetric == DataStructs.TverskySimilarity):
av = float(kwargs.get('tverskyA', 0.5))
bv = float(kwargs.get('tverskyB', 0.5))
scores = DataStructs.BulkTverskySimilarity(fp, validFps, av, bv)
for i, score in enumerate(scores):
if score > simThresh:
nbrLists[validProbes[i]].Insert(score, nm)
else:
for i in range(len(probeFps)):
pfp = probeFps[i]
if pfp is not None:
score = simMetric(probeFps[i], fp)
if score > simThresh:
nbrLists[validProbes[i]].Insert(score, nm)
return nbrLists
def ScreenFingerprints(details, data, mol=None, probeFp=None):
""" Returns a list of results
"""
if probeFp is None:
try:
probeFp = FingerprintMols.FingerprintMol(mol, **details.__dict__)
except Exception:
import traceback
FingerprintMols.error('Error: problems fingerprinting molecule.\n')
traceback.print_exc()
return []
if not probeFp:
return []
if not details.doThreshold and details.topN > 0:
topN = TopNContainer(details.topN)
else:
topN = []
res = []
count = 0
for pt in data:
fp1 = probeFp
if not details.noPickle:
if isinstance(pt, (tuple, list)):
ID, fp = pt
else:
fp = pt
ID = pt._fieldsFromDb[0]
score = DataStructs.FingerprintSimilarity(fp1, fp, details.metric)
else:
ID, pkl = pt
score = details.metric(fp1, str(pkl))
if topN:
topN.Insert(score, ID)
elif not details.doThreshold or (details.doThreshold and score >= details.screenThresh):
res.append((ID, score))
count += 1
if hasattr(details, 'stopAfter') and count >= details.stopAfter:
break
for score, ID in topN:
res.append((ID, score))
return res
