def testMurtaghUPGMA(self):
if Murtagh is None: return
nPts = 5
sz = 5
dataP = numpy.random.random((nPts, sz))
newClust = Murtagh.ClusterData(dataP, nPts, Murtagh.UPGMA)[0]
ds = []
for i in range(nPts):
for j in range(i):
d = dataP[i] - dataP[j]
ds.append(sum(d * d))
ds = numpy.array(ds)
newClust2 = Murtagh.ClusterData(ds, nPts, Murtagh.UPGMA,
isDistData=1)[0]
assert len(newClust) == len(newClust2), 'length mismatch2'
assert not newClust.Compare(newClust2,
ignoreExtras=0), 'equality failed3'
newClust2 = Murtagh.ClusterData(dataP,
nPts,
Murtagh.UPGMA,
isDistData=0)[0]
assert len(newClust) == len(newClust2), 'length mismatch2'
assert not newClust.Compare(newClust2,
ignoreExtras=0), 'equality failed3'
def ClusterPoints(data,
metric,
algorithmId,
haveLabels=False,
haveActs=True,
returnDistances=False):
message('Generating distance matrix.\n')
dMat = GetDistanceMatrix(data, metric)
message('Clustering\n')
clustTree = Murtagh.ClusterData(dMat, len(data), algorithmId,
isDistData=1)[0]
acts = []
if haveActs and len(data[0]) > 2:
# we've got activities... use them:
acts = [int(x[2]) for x in data]
if not haveLabels:
labels = ['Mol: %s' % str(x[0]) for x in data]
else:
labels = [x[0] for x in data]
clustTree._ptLabels = labels
if acts:
clustTree._ptValues = acts
for pt in clustTree.GetPoints():
idx = pt.GetIndex() - 1
pt.SetName(labels[idx])
if acts:
try:
pt.SetData(int(acts[idx]))
except Exception:
pass
if not returnDistances:
return clustTree
else:
return clustTree, dMat
def WardsClustering(dists, nfps):
print "-------------------------------------------------"
print "starting Wards clustering"
start_time = time.time()
c_tree = Murtagh.ClusterData(dists, nfps, Murtagh.WARDS, isDistData=True)
print "time taken: ", time.time() - start_time
return c_tree
def gen_coarseclusters(dists, nfps):
"""
A function to generate coarse grained clusters (i.e. Murtagh) from Tanimoto distance matrices
:param dists: Tanimoto distance matrix
:param nfps: number of fingerprints
:return: cs (clusters)
"""
from rdkit import DataStructs
from rdkit.ML.Cluster import Murtagh
# now cluster the data:
cs = Murtagh.ClusterData(dists, nfps, Murtagh.WARDS, isDistData=1)
return cs
def ClusterFps_Murtagh(self, dists, nfps, method, ncluster):
self.cdict = {}
cs = None
if method == 'Wards':
cs = Murtagh.ClusterData(dists,
len(self.fplist),
Murtagh.WARDS,
isDistData=1)
elif method == 'SLINK':
cs = Murtagh.ClusterData(dists,
len(self.fplist),
Murtagh.SLINK,
isDistData=1)
elif method == 'CLINK':
cs = Murtagh.ClusterData(dists,
len(self.fplist),
Murtagh.CLINK,
isDistData=1)
elif method == 'UPGMA':
cs = Murtagh.ClusterData(dists,
len(self.fplist),
Murtagh.UPGMA,
isDistData=1)
splitClusts = ClusterUtils.SplitIntoNClusters(cs[0], ncluster)
#centroids = [ClusterUtils.FindClusterCentroidFromDists(x,dists) for x in splitClusts]
for index, cluster in enumerate(splitClusts):
children = cluster.GetPoints()
pts = [x.GetData() for x in children]
self.clustdict[index + 1] = pts
for pt in pts:
self.cdict[pt] = [index + 1]
if pt == pts[0]:
self.cdict[pt].append("true")
else:
self.cdict[pt].append("flase")
[[10.0, 5.0], [20.0, 20.0], [30.0, 10.0], [30.0, 15.0], [5.0, 10.0]],
numpy.float)
print '2'
#clusters = Murtagh.ClusterData(d,len(d),Murtagh.WARDS)
#for i in range(len(clusters)):
# clusters[i].Print()
#print '3'
dists = []
for i in range(len(d)):
for j in range(i):
dist = sum((d[i] - d[j])**2)
dists.append(dist)
dists = numpy.array(dists)
print 'Wards:'
clusters = Murtagh.ClusterData(dists, len(d), Murtagh.WARDS, isDistData=1)
clusters[0].Print()
print 'SLINK:'
clusters = Murtagh.ClusterData(dists, len(d), Murtagh.SLINK, isDistData=1)
clusters[0].Print()
print 'CLINK:'
clusters = Murtagh.ClusterData(dists, len(d), Murtagh.CLINK, isDistData=1)
clusters[0].Print()
print 'UPGMA:'
clusters = Murtagh.ClusterData(dists, len(d), Murtagh.UPGMA, isDistData=1)
clusters[0].Print()