def test0FromList(self):
bv1 = DataStructs.SparseBitVect(1000)
bv2 = DataStructs.SparseBitVect(1000)
obits = range(0, 1000, 3)
for bit in obits:
bv1.SetBit(bit)
bv2.SetBitsFromList(obits)
for i in range(1000):
assert bv1.GetBit(i) == bv2.GetBit(i)
self.failUnless(bv1 == bv2)
bv2.SetBit(1)
self.failUnless(bv1 != bv2)
bv2.UnSetBit(1)
self.failUnless(bv1 == bv2)
bv1 = DataStructs.ExplicitBitVect(1000)
bv2 = DataStructs.ExplicitBitVect(1000)
obits = range(0, 1000, 3)
for bit in obits:
bv1.SetBit(bit)
bv2.SetBitsFromList(obits)
for i in range(1000):
assert bv1.GetBit(i) == bv2.GetBit(i)
def test1FromList(self):
examples = []
bv = DS.SparseBitVect(5)
bv.SetBitsFromList([0, 2, 4])
examples.append([0, bv, 0])
bv = DS.SparseBitVect(5)
bv.SetBitsFromList([0, 2, 4])
examples.append([0, bv, 0])
bv = DS.SparseBitVect(5)
bv.SetBitsFromList([0, 3, 4])
examples.append([0, bv, 1])
bv = DS.SparseBitVect(5)
bv.SetBitsFromList([0, 2, 4])
examples.append([0, bv, 0])
bv = DS.SparseBitVect(5)
bv.SetBitsFromList([0, 2])
examples.append([0, bv, 1])
r = FS.selectCMIM(examples, 2)
self.failUnlessEqual(r, (2, 4))
r = FS.selectCMIM(examples, 1)
self.failUnlessEqual(r, (2, ))
r = FS.selectCMIM(examples, 3)
self.failUnlessEqual(r, (2, 4, -1))
def test2ranker(self):
nbits = 100
ninst = 100
dm = 50
nact = 10
nc = 2
RDRandom.seed(23)
rn = rdit.InfoBitRanker(nbits, nc, rdit.InfoType.ENTROPY)
rn.SetMaskBits([63, 70, 15, 25, 10])
fps = []
na = 0
ni = 0
for i in range(ninst):
v = DataStructs.SparseBitVect(nbits)
for j in range(dm):
v.SetBit(RDRandom.randrange(0, nbits))
if (RDRandom.randrange(0, ninst) < nact):
na += 1
rn.AccumulateVotes(v, 1)
fps.append((v, 1))
else:
ni += 1
rn.AccumulateVotes(v, 0)
fps.append((v, 0))
res = rn.GetTopN(5)
ids = [int(x[0]) for x in res]
ids.sort()
self.assertTrue(ids == [10, 15, 25, 63, 70])
with self.assertRaisesRegexp(Exception, ""):
res = rn.GetTopN(10)
def GetAtomPairFingerprintAsBitVect(mol):
""" Returns the Atom-pair fingerprint for a molecule as
a SparseBitVect. Note that this doesn't match the standard
definition of atom pairs, which uses counts of the
pairs, not just their presence.
**Arguments**:
- mol: a molecule
**Returns**: a SparseBitVect
>>> m = Chem.MolFromSmiles('CCC')
>>> v = [ pyScorePair(m.GetAtomWithIdx(0),m.GetAtomWithIdx(1),1),
... pyScorePair(m.GetAtomWithIdx(0),m.GetAtomWithIdx(2),2),
... ]
>>> v.sort()
>>> fp = GetAtomPairFingerprintAsBitVect(m)
>>> list(fp.GetOnBits())==v
True
"""
res = DataStructs.SparseBitVect(fpLen)
fp = rdMolDescriptors.GetAtomPairFingerprint(mol)
for val in fp.GetNonzeroElements().keys():
res.SetBit(val)
return res
def test1SparsePickle(self):
nbits = 10000
bv1 = DataStructs.SparseBitVect(nbits)
for i in range(1000):
x = random.randrange(0, nbits)
bv1.SetBit(x)
pkl = pickle.dumps(bv1, 1)
bv2 = pickle.loads(pkl)
for i in range(nbits):
assert bv1[i] == bv2[i]
def test5sbv(self):
n = 30
m = 2048
dm = 800
lst = []
for i in range(n):
v = DataStructs.SparseBitVect(m)
for j in range(dm):
v.SetBit(random.randrange(0, m))
lst.append(v)
dMat = rdmmc.GetTanimotoDistMat(lst)
sMat = rdmmc.GetTanimotoSimMat(lst)
for i in range(n * (n - 1) // 2):
assert feq(sMat[i] + dMat[i], 1.0)
def test1ranker(self):
nbits = 100
ninst = 100
dm = 50
nact = 10
nc = 2
rn = rdit.InfoBitRanker(nbits, nc, rdit.InfoType.ENTROPY)
fps = []
na = 0
ni = 0
for i in range(ninst):
v = DataStructs.SparseBitVect(nbits)
for j in range(dm):
v.SetBit(RDRandom.randrange(0, nbits))
if (RDRandom.randrange(0, ninst) < nact):
na += 1
rn.AccumulateVotes(v, 1)
fps.append((v, 1))
else:
ni += 1
rn.AccumulateVotes(v, 0)
fps.append((v, 0))
res = rn.GetTopN(50)
rn2 = rdit.InfoBitRanker(nbits, nc)
for fp in fps:
rn2.AccumulateVotes(fp[0], fp[1])
res2 = rn2.GetTopN(50)
self.assertTrue((res == res2).all())
rn3 = rdit.InfoBitRanker(nbits, nc, rdit.InfoType.BIASENTROPY)
#rn3.SetBiasList([0])
for fp in fps:
rn3.AccumulateVotes(fp[0], fp[1])
res3 = rn3.GetTopN(50)
for i in range(50):
fan = res3[i, 2] / na
fin = res3[i, 3] / ni
self.assertTrue(fan > fin)
