annotate chemfp_clustering/old/butina_clustering_old.py @ 1:43a9e7d9b24f draft

planemo upload for repository https://github.com/bgruening/galaxytools/tree/master/chemicaltoolbox/chemfp commit a44c0a13283e873a740eabcad04f021208290dfe-dirty

1 #!/usr/bin/env python
2 """
3     Modified version of code examples from the chemfp project.
4     http://code.google.com/p/chem-fingerprints/
5     Thanks to Andrew Dalke of Andrew Dalke Scientific!
6 """
7 
8 import chemfp
9 import sys
10 import os
11 import tempfile
12 
13 temp_file = tempfile.NamedTemporaryFile()
14 temp_link = "%s.%s" % (temp_file.name, 'fps')
15 temp_file.close()
16 os.system('ln -s %s %s' % (os.path.realpath(sys.argv[1]), temp_link) )
17 
18 
19 chemfp_fingerprint_file = temp_link
20 tanimoto_threshold = float(sys.argv[2])
21 outfile = sys.argv[3]
22 processors = int(sys.argv[4])
23 
24 
25 def get_hit_indicies(hits):
26     return [id for (id, score) in hits]
27 
28 out = open(outfile, 'w')
29 dataset = chemfp.load_fingerprints( chemfp_fingerprint_file )
30 
31 chemfp.set_num_threads( processors )
32 search = dataset.threshold_tanimoto_search_arena(dataset, threshold = tanimoto_threshold)
33 #search = chemfp.search.threshold_tanimoto_search_symmetric (dataset, threshold = tanimoto_threshold)
34 
35 # Reorder so the centroid with the most hits comes first.
36 # (That's why I do a reverse search.)
37 # Ignore the arbitrariness of breaking ties by fingerprint index
38 results = sorted( (  (len(hits), i, hits) for (i, hits) in enumerate(search.iter_indices_and_scores())  ),reverse=True)
39 
40 
41 # Determine the true/false singletons and the clusters
42 true_singletons = []
43 false_singletons = []
44 clusters = []
45 
46 seen = set()
47 
48 for (size, fp_idx, hits) in results:
49     if fp_idx in seen:
50         # Can't use a centroid which is already assigned
51         continue
52     seen.add(fp_idx)
53     print size, fp_idx, hits
54     if size == 1:
55         # The only fingerprint in the exclusion sphere is itself
56         true_singletons.append(fp_idx)
57         continue
58 
59     members = get_hit_indicies(hits)
60     # Figure out which ones haven't yet been assigned
61     unassigned = [target_idx for target_idx in members if target_idx not in seen]
62 
63     if not unassigned:
64         false_singletons.append(fp_idx)
65         continue
66 
67     # this is a new cluster
68     clusters.append( (fp_idx, unassigned) )
69     seen.update(unassigned)
70 
71 len_cluster = len(clusters)
72 #out.write( "#%s true singletons: %s\n" % ( len(true_singletons), " ".join(sorted(dataset.ids[idx] for idx in true_singletons)) ) )
73 #out.write( "#%s false singletons: %s\n" % ( len(false_singletons), " ".join(sorted(dataset.ids[idx] for idx in false_singletons)) ) )
74 
75 out.write( "#%s true singletons\n" % len(true_singletons) )
76 out.write( "#%s false singletons\n" % len(false_singletons) )
77 out.write( "#clusters: %s\n" % len_cluster )
78 
79 # Sort so the cluster with the most compounds comes first,
80 # then by alphabetically smallest id
81 def cluster_sort_key(cluster):
82     centroid_idx, members = cluster
83     return -len(members), dataset.ids[centroid_idx]
84 
85 clusters.sort(key=cluster_sort_key)
86 
87 
88 for centroid_idx, members in clusters:
89     centroid_name = dataset.ids[centroid_idx]
90     out.write("%s\t%s\t%s\n" % (centroid_name, len(members), " ".join(sorted(dataset.ids[idx] for idx in members))))
91     #ToDo: len(members) need to be some biggest top 90% or something ...
92 
93 for idx in sorted(true_singletons):
94     out.write("%s\t%s\n" % (dataset.ids[idx], 0))
95 
96 out.close()
97 os.remove( temp_link )