def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb"%id)
active_sites.append(io.read_active_site(filepath))
# update this assertion
clusters_out = cluster.cluster_hierarchically(active_sites)
print('the clusters out look like', clusters_out)
#Check length and structure
hierarchical_structure_check(clusters_out)
#Let's also check what this hierarchical test looks like for much longer, even-length list:
pdb_ids = [1806, 3458, 3733, 4629, 6040, 7674, 7780, 8208, 8304, 9776, 10701, 10814]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb"%id)
active_sites.append(io.read_active_site(filepath))
clusters_out = cluster.cluster_hierarchically(active_sites)
print('the clusters out look like', clusters_out)
#Check length and structure
hierarchical_structure_check(clusters_out)
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 52954, 34088]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
sim_matrix = cluster.similarity_matrix(active_sites)
# update this assertion
assert cluster.cluster_hierarchically([], {}) == []
assert cluster.cluster_hierarchically([active_sites[0]],
{}) == [[active_sites[0]]]
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701, 10701,10814,13052,14181,15813]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb"%id)
active_sites.append(io.read_active_site(filepath))
cluster.get_order_residues(active_sites)
# update this assertion
assert len(cluster.cluster_hierarchically(active_sites,2)[0]) == 2
assert len(cluster.cluster_hierarchically(active_sites,3)[0]) == 3
assert len(cluster.cluster_hierarchically(active_sites,4)[0]) == 4
def test_compareTwoMethods():
allPossible = []
for filename in os.listdir("data"):
allPossible.append(int(filename.split(".")[0]))
hierTotal = 0
partTotal = 0
iterations = 100
numPDBs = 15
for i in range(0, iterations):
indices = random.sample(range(0, len(allPossible)), numPDBs)
pdb_ids = []
for j in indices:
pdb_ids.append(allPossible[j])
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
hierScore = cluster.qualityMetric(
cluster.cluster_hierarchically(active_sites))
partScore = cluster.qualityMetric(
cluster.cluster_by_partitioning(active_sites))
hierTotal += hierScore
partTotal += partScore
print("hierScoreAverage: ", hierTotal / float(iterations))
print("partScoreAverage: ", partTotal / float(iterations))
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
# clusters more similar clusters together
clustering = cluster.cluster_hierarchically(active_sites, [2])
assert get_names(flatten(clustering[0])) in [['4629', '276'], ['10701']]
assert get_names(flatten(clustering[1])) in [['4629', '276'], ['10701']]
# len(clustered_list.unique()==k)
active_sites = read_active_sites("data")
assert len(cluster.cluster_hierarchically(active_sites, [2])) == 2
assert len(cluster.cluster_hierarchically(active_sites, [3])) == 3
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
# update this assertion
assert cluster.cluster_hierarchically(active_sites) == [[1, 0], 2]
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 276, 10701, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
assert cluster.cluster_hierarchically(
active_sites, 2) == [[10701, 10701], [276, 276]] or [[276, 276],
[10701, 10701]]
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
print()
assert len(cluster.cluster_hierarchically(active_sites).keys()) >= 2
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
# clusters produce k number of final labels
assert len(
cluster.cluster_hierarchically(active_sites)) == 3 #k = 3 in my code
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
# update this assertion
# checking the the three sites cluster as expected
assert np.array_equal(
cluster.cluster_hierarchically(active_sites, 2)[0], [0, 0, 1])
def test_hierarchical_clustering():
random.seed(40)
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
# update this assertion
label = cluster.cluster_hierarchically(active_sites)
assert all(label[1] == [0, 0, 1])
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
assignment = cluster.cluster_hierarchically(active_sites)
assert cluster.convert_indices_to_active_sites(assignment,
active_sites) == [["276"],
["4629"],
["10701"]]
pdb_ids = [276, 4629]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
# check empty active sites doesn't crash
assert cluster.cluster_hierarchically(None, K=10) is None
def test_hierarchical_clustering():
# tractable subset
pdb_ids = [276, 4629, 10701]
active_sites = []
for id in pdb_ids:
filepath = os.path.join("data", "%i.pdb" % id)
active_sites.append(io.read_active_site(filepath))
clusters = []
for clust in cluster.cluster_hierarchically(active_sites, 1):
elems = []
for e in clust:
elems.append([int(n.name) for n in e])
clusters.append(elems)
assert clusters == [[[276], [4629], [10701]], [[276], [10701, 4629]],
[[276, 10701, 4629]]]
number_clusters = []
p_sil_scores = []
h_sil_scores = []
r_sil_scores = []
for i in range(2,10):
# clustering all sites by partition
print('Finding %d clusters by partitioning'% i)
p_clusters, p_distances = cluster.cluster_by_partitioning(active_sites, i)
# clustering all sites hierarchically
print('Finding %d clusters hierarchically'% i)
h_clusters, h_distances = cluster.cluster_hierarchically(active_sites, i)
# generating random clusters labels
print('Generating %d random cluster labels'% i)
r_clusters = np.random.choice(range(0,i), 136)
# to evaluate the clusters I will get the silhouette score for each method of clustering
# as well as randomly generated cluster labels
p_sil_scores.append(silhouette_score(p_distances, p_clusters))
h_sil_scores.append(silhouette_score(h_distances, h_clusters))
r_sil_scores.append(silhouette_score(p_distances, r_clusters))
fig1 = plt.figure(dpi = 300)
plt.xlabel("Number of Clusters")
plt.ylabel('Silhouette Score')
plt.title('Evaluating Clustering Methods')
active_sites = io.read_active_sites(
"C:\\Users\Zoë\Documents\GitHub\hw2-skeleton\data")
#site1 = active_sites[5]
#site2 = active_sites[7]
#print('site1: ', site1.categories)
#print('site2: ', site2.categories)
#sim = cluster.compute_similarity(site1,site2)
# Run for one clustering by kmeans
Pclusters, PmaxDistance = cluster.cluster_by_partitioning(active_sites)
##for i in clusters:
## print(i.toStr())
io.write_clustering("clusterPk=10", Pclusters)
# Run for just one clustering by agglomerative clustering
Hclusters, distH = cluster.cluster_hierarchically(active_sites)
io.write_clustering("clusterHcutoff=0.3", Hclusters)
## Run for one clustering by agglomerative clustering
#Hclusters, HmaxDist, Hclusterings = cluster.cluster_hierarchically(active_sites)
#io.write_mult_clusterings("clusteringsH1", Hclusterings)
#%%
## Clusterings of multiple k values in kmeans
#kvals = [2,3,4,5,6,7,8,9,10,20,30,40,50,60,70,80,90,100,120,130,136]
#kvalstest = [2,10,100]
#
#numtrials = 10
#allscores = np.zeros((numtrials, len(kvals)))
#for i in range(numtrials):
# Pclusterings = cluster.cluster_by_partitioning(active_sites)