def main(argv): parser = argparse.ArgumentParser( description=__DESCRIPTION__, formatter_class=argparse.RawDescriptionHelpFormatter) parser.add_argument("target_trees", metavar='target_trees', type=str, nargs="*", help='a list of target tree files') parser.add_argument( "--targets_file", dest="targets_file", type=str, help="""path to a file containing target trees, one per line""") parser.add_argument("-o", dest="output", type=str, help="""Path to the tab delimited report file""") parser.add_argument("-r", dest="reftree", type=str, required=True, help="""Reference tree""") parser.add_argument( "--outgroup", dest="outgroup", nargs="+", help= """outgroup used to root reference and target trees before distance computation""" ) parser.add_argument("--expand_polytomies", dest="polytomies", action="store_true", help="""expand politomies if necessary""") parser.add_argument("--unrooted", dest="unrooted", action="store_true", help="""compare trees as unrooted""") parser.add_argument( "--min_support", dest="min_support", type=float, default=0.0, help= ("min support value for branches to be counted in the distance computation (RF, treeko and refTree/targeGene compatibility)" )) parser.add_argument( "--extract_species", dest="extract_species", action="store_true", help= """When used, reference tree is assumed to contain species names, while target trees as expected to be gene trees. Species name will be extracted from gene tree nodes and treeko will be used if duplication events are found.""" ) parser.add_argument("--spname_delimiter", dest="spname_delimiter", type=str, default="_", help=("species code delimiter in node names")) parser.add_argument( "--spname_field", dest="spname_field", type=int, default=-1, help= ("position of the species code extracted from node names. -1 = last field" )) parser.add_argument("--collateral", dest="collateral", action='store_true', help=("")) parser.add_argument("--ref_attr", dest="ref_attr", type=str, help=("attribute in ref tree used as leaf name")) parser.add_argument("--target_attr", dest="target_attr", type=str, help=("attribute in target tree used as leaf name")) args = parser.parse_args(argv) print __DESCRIPTION__ reftree = args.reftree if args.targets_file and args.target_trees: print >> sys.stderr, 'The use of targets_file and targets at the same time is not supported.' sys.exit(1) if args.targets_file: target_trees = tree_iterator(args.targets_file) else: target_trees = args.target_trees t = Tree(reftree) if args.ref_attr: for lf in t.iter_leaves(): lf._origname = lf.name if args.ref_attr not in lf.features: print lf lf.name = getattr(lf, args.ref_attr) if args.outgroup: if len(args.outgroup) > 1: out = t.get_common_ancestor(args.outgroup) else: out = t.search_nodes(name=args.outgroup[0])[0] t.set_outgroup(out) ref_names = set(t.get_leaf_names()) reftree_len = len(t) reftree_edges = (reftree_len * 2) - 2 ncollapsed_branches = len([ n for n in t.traverse() if n.children and n.support < args.min_support ]) #reftree_edges -= ncollapsed_branches #if ncollapsed_branches: # print '%d branches collapsed in reference tree' %ncollapsed_branches HEADER = ("target tree", 'dups', 'subtrees', 'used trees', 'treeko', "RF", "maxRF", 'normRF', "%reftree", "%genetree", "avgSize", "minSize", "common tips", "refSize", "targetSize") if args.output: OUT = open(args.output, "w") print >> OUT, '# ' + ctime() print >> OUT, '# ' + ' '.join(sys.argv) print >> OUT, '#' + '\t'.join(HEADER) else: print '# ' + ctime() print '# ' + ' '.join(sys.argv) COL_WIDTHS = [20, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7] print_table([HEADER], fix_col_width=COL_WIDTHS, wrap_style='wrap') prev_tree = None for counter, tfile in enumerate(target_trees): if args.targets_file: seedid, tfile = tfile else: seedid = None if args.extract_species: tt = PhyloTree(tfile, sp_naming_function=lambda name: name.split( args.spname_delimiter)[args.spname_field]) else: tt = Tree(tfile) if args.target_attr: for lf in tt.iter_leaves(): lf._origname = lf.name lf.name = getattr(lf, args.target_attr) if args.outgroup: if len(args.outgroup) > 1: out = tt.get_common_ancestor(args.outgroup) else: out = tt.search_nodes(name=args.outgroup[0])[0] tt.set_outgroup(out) if args.target_trees: fname = os.path.basename(tfile) else: fname = '%05d' % counter max_size, min_size, avg_size, common = -1, -1, -1, -1 total_rf, max_rf, norm_rf = -1, -1, -1 treeko_d = -1 ref_branches_in_target, target_branches_in_ref = -1, -1 target_tree_len = -1 used_subtrees = -1 if args.extract_species: orig_target_size = len(tt) ntrees, ndups, sp_trees = tt.get_speciation_trees( autodetect_duplications=True, newick_only=True) if ntrees < 1000: all_rf = [] ref_found = [] target_found = [] tree_sizes = [] all_max_rf = [] common_names = 0 for subtree_nw in sp_trees: if seedid and not args.collateral and (seedid not in subtree_nw): continue subtree = PhyloTree( subtree_nw, sp_naming_function=lambda name: name.split( args.spname_delimiter)[args.spname_field]) # only necessary if rf function is going to filter by support value. It slows downs the analysis, obviously if args.min_support: subtree_content = subtree.get_cached_content( store_attr='name') for n in subtree.traverse(): if n.children: n.support = tt.get_common_ancestor( subtree_content[n]).support rf, maxr, common, p1, p2, d1, d2 = t.robinson_foulds( subtree, expand_polytomies=args.polytomies, unrooted_trees=args.unrooted, attr_t2='species', min_support_t2=args.min_support) if maxr > 0 and p1 and p2: all_rf.append(rf) tree_sizes.append(len(common)) all_max_rf.append(maxr) common_names = max(common_names, len(common)) ref_found.append(float(len(p2 & p1)) / reftree_edges) p2bis = set([ p for p in (p2 - d2) if len(p[0]) > 1 and len(p[1]) > 1 ]) # valid edges in target not leaves if p2bis: incompatible_target_branches = float( len((p2 - d2) - p1)) target_found.append(1 - (incompatible_target_branches / (len(p2 - d2)))) # valid_target = p2-d2 # valid_ref = p1-d1 # ref_found.append(float(len(valid_target & valid_ref)) / reftree_edges) # p2bis = set([p for p in (p2-d2) if len(p[0])>1 and len(p[1])>1]) # if p2bis-d2: # incompatible_target_branches = float(len((p2-d2) - p1)) # target_found.append(1 - (incompatible_target_branches / (len(p2-d2)))) if all_rf: # Treeko speciation distance alld = [(all_rf[i] / float(all_max_rf[i])) for i in xrange(len(all_rf))] a = numpy.sum( [alld[i] * tree_sizes[i] for i in xrange(len(all_rf))]) b = float(numpy.sum(tree_sizes)) treeko_d = a / b total_rf = numpy.mean(all_rf) norm_rf = numpy.mean([(all_rf[i] / float(all_max_rf[i])) for i in xrange(len(all_rf))]) max_rf = numpy.max(all_max_rf) ref_branches_in_target = numpy.mean(ref_found) target_branches_in_ref = numpy.mean( target_found) if target_found else -1 target_tree_len = numpy.mean(tree_sizes) used_subtrees = len(all_rf) else: target_tree_len = len(tt) ndups, ntrees, used_subtrees = 0, 1, 1 treeko_d = -1 total_rf, max_rf, common, p1, p2, d1, d2 = tt.robinson_foulds( t, expand_polytomies=args.polytomies, unrooted_trees=args.unrooted) common_names = len(common) if max_rf: norm_rf = total_rf / float(max_rf) if p1 and p2: sizes = [len(p) for p in p2 ^ p1] if sizes: avg_size = sum(sizes) / float(len(sizes)) max_size, min_size = max(sizes), min(sizes) else: max_size, min_size, avg_size = 0, 0, 0 ref_branches_in_target = float(len(p2 & p1)) / reftree_edges #if p2-d2: # incompatible_target_branches = float(len((p2-d2) - p1)) # target_found.append(1 - (incompatible_target_branches / (len(p2-d2)))) else: ref_branches_in_target = 0.0 target_branches_in_ref = 0.0 max_size, min_size, avg_size = -1, -1, -1 if args.output: print >> OUT, '\t'.join( map(str, (fname, ndups, ntrees, used_subtrees, treeko_d, total_rf, max_rf, norm_rf, ref_branches_in_target, target_branches_in_ref, avg_size, min_size, common_names, reftree_len, target_tree_len))) else: print_table([ map(istr, (fname[-30:], ndups, ntrees, used_subtrees, treeko_d, total_rf, max_rf, norm_rf, '%0.4f' % ref_branches_in_target, '%0.4f' % target_branches_in_ref, avg_size, min_size, common_names, reftree_len, target_tree_len)) ], fix_col_width=COL_WIDTHS, wrap_style='cut') if args.output: OUT.close()
def main(argv): global args parser = argparse.ArgumentParser( description=__DESCRIPTION__, formatter_class=argparse.RawDescriptionHelpFormatter) parser.add_argument("-r", dest="reftree", type=str, required=True, help="""Reference tree""") parser.add_argument( "--source_trees", dest="source_trees", type=str, required=True, help= ("A list of *rooted* genetrees, one per line, in the format: TreeID/SeedID [TAB] newick " )) parser.add_argument("--plot_newick", dest="plot_newick", type=str, help=("")) parser.add_argument("--spname_delimiter", dest="spname_delimiter", type=str, default="_", help=("species code delimiter in node names")) parser.add_argument( "--spname_field", dest="spname_field", type=int, default=-1, help= ("position of the species code extracted from node names. -1 = last field" )) parser.add_argument( "--collateral", dest="use_collateral", action="store_true", help=("If enabled, collateral information will be used as" " equally qualified data. Otherwise, such data will" " be reported separatedly. Use this if your set of" " trees are not overlaping. ")) parser.add_argument( "--skip_dup_detection", dest="skip_dup_detection", action="store_true", help=('If used, duplications will be expected to be annotated' ' in the source gene trees with the evoltype="D" tag.' ' Otherwise they will be inferred on the fly using' ' the species overlap algorithm.')) parser.add_argument( "--spoverlap", dest="species_overlap", type=float, default=0.0, help=("Species overlap cutoff. A number between 0 and 1 " "representing the percentage of species that should be " "shared between two sister partitions to be considered a" " duplication. 0 = any overlap represents a duplication. ")) parser.add_argument( "--debug", dest="debug", action="store_true", help= ("generate an image of every input gene tree tree, so the result can be inspected" )) parser.add_argument( "--snapshot_step", dest="snapshot_step", type=int, default=1000, help=("How many trees should be processed between snapshots dumps?")) parser.add_argument( "--reftree_constraint", dest="reftree_constraint", type=str, help=("A python module from from which a function called " "*is_valid_treeid(treeid, refbranch)* should be importable. " "The function will be used to decide if the info of a given " "source tree is informative or not for each reftree branch. ")) parser.add_argument("-o", dest="output", type=str, required=True, help=("output tag name (extensions will be added)")) parser.add_argument("--cpu", dest="cpu", type=int, default=1, help=("enable parallel computation")) parser.add_argument( "--img_report", dest="img_report", action="store_true", help= ("If true, it generates a summary image results with all the computed data" )) parser.add_argument( "--report_supports", dest="report_supports", action="store_true", help= ("If used, supported ref tree branches are individually reported for each gene tree " )) args = parser.parse_args(argv) if args.plot_newick: t = Tree(args.plot_newick) ts = TreeStyle() ts.layout_fn = info_layout t.render("tree_analysis.png", tree_style=ts) sys.exit(0) SPNAME_FIELD, SPNAME_DELIMITER = args.spname_field, args.spname_delimiter USE_COLLATERAL = args.use_collateral DETECT_DUPLICATIONS = True if not args.skip_dup_detection else False REPORT_PER_TREE_SUPPORTS = True if args.report_supports else False SP_OVERLAP = args.species_overlap DEBUG = args.debug IMG_REPORT = args.img_report reftree = PhyloTree(args.reftree, sp_naming_function=None) for nid, n in enumerate(reftree.traverse()): n.add_features(nid=nid) REFTREE_SPECIES = set(reftree.get_leaf_names()) print __DESCRIPTION__ if REPORT_PER_TREE_SUPPORTS: REPORT_SUPPORT_FILE = open("%s.gentree_supports" % args.output, "w") print >> REPORT_SUPPORT_FILE, '#' + '\t'.join( map(str, [ "treeId", "spCoverage", "mean_support", "mean_coll_support", "tested_branches", 'tested_coll_branches' ])) TOTAL_TREES = int( commands.getoutput("wc -l %s" % args.source_trees).split()[0]) + 1 print >> sys.stderr, "Processing %d source trees" % TOTAL_TREES if args.reftree_constraint: import imp constraint = imp.load_source('constraint', args.reftree_constraint) IS_VALID_TREEID = constraint.is_valid_treeid else: IS_VALID_TREEID = None if args.cpu > 1: MONITOR_STEP = 0 #return (informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, # coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) # The output of the process_trees function are 9 dictionaries in which keys are refbranches target_dicts = [{} for x in range(9)] def merge_dict_results(target, source): def merge_dict(target, source): for k, v in source.iteritems(): if k not in target: target[k] = v elif isinstance(v, list): target[k].extend(v) elif isinstance(v, set): target[k].update(v) elif isinstance(v, int): target[k] += v else: raise ValueError("Impossible to merge str results") for index in xrange(len(target)): merge_dict(target[index], out[index]) from multiprocessing import Process, Queue from Queue import Empty as QueueEmpty outputs_queue = Queue() if TOTAL_TREES > args.cpu: trees_per_cpu = TOTAL_TREES / args.cpu trees_per_cpu += 1 if TOTAL_TREES % args.cpu else 0 else: trees_per_cpu = 1 args.cpu = TOTAL_TREES all_workers = set() for cpu_num in xrange(args.cpu): sline = (cpu_num * trees_per_cpu) eline = (cpu_num * trees_per_cpu) + trees_per_cpu data_iter = tree_iterator(args.source_trees, restrict_species=REFTREE_SPECIES, start_line=sline, end_line=eline) print >> sys.stderr, "Launching worker %d from %d to %d" % ( cpu_num, sline, eline) worker = Process(target=run_parallel, args=(cpu_num, outputs_queue, process_trees, data_iter, reftree, trees_per_cpu)) worker.name = "Worker_%d" % cpu_num all_workers.add(worker) worker.start() while all_workers: # clear done threads for w in list(all_workers): if not w.is_alive(): print >> sys.stderr, "%s thread is done!" % w.name all_workers.discard(w) # get and merge results while 1: try: out = outputs_queue.get(False) except QueueEmpty: break else: # This merge depends on process_trees return output!!!!! merge_dict_results(target_dicts, out) # Dump a snapshot dump_results(reftree, *target_dicts) time.sleep(0.1) if all_workers: time.sleep(1) # collected data (informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) = target_dicts else: MONITOR_STEP = args.snapshot_step data_iter = tree_iterator(args.source_trees, restrict_species=REFTREE_SPECIES) (informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) = process_trees(data_iter, reftree, TOTAL_TREES) if REPORT_PER_TREE_SUPPORTS: REPORT_SUPPORT_FILE.close() dump_results(reftree, informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) print >> sys.stderr, "Dumping full analysis..." # Full dump, including duplication details cPickle.dump(reftree, open("%s.pkl" % args.output, "w"))
def get_supported_branches(source_tree, reftree, refclades, seedid=None): """ Given a reference species tree and a rooted gene tree in which duplication events are already mapped, this function does the following: - Split gene tree into all possible species tree (Treeko method) - Find matches between each subtree branch and all branches in the reference tree. - Each branch in each species subtree is compared to all branches in the reftree. If left/right side of the subtree branch coincide with a the left/right side of a reference tree branch, this is considered a gene tree support point. Coincidences must comply with the following conditions: - All species in the left/right sides of the subtree branch exist in the left/right sides of the reference branch. - Species in the left/right sides of the reference branch are never mixed in the subtree branch. - Missing species are allowed in the subtree split, only if such species are not present in any other part of the original gene tree. """ # Run Treeko to get all possible species tree combinations. We assume dups are already mapped ntrees, ndups, sp_trees = source_tree.get_speciation_trees( autodetect_duplications=DETECT_DUPLICATIONS, newick_only=True) if ntrees > 100000: return {}, {} branches_found = [] branch2supports = defaultdict(list) branch2coll_supports = defaultdict(list) for nw in sp_trees: # Use all treeko trees or only those subtrees containing the seed? if seedid and seedid not in nw: container = branch2coll_supports else: container = branch2supports subtree = PhyloTree(nw, sp_naming_function=extract_species) subtreenode2content = subtree.get_cached_content(store_attr="species") #set([phy3(_c.name) for _c in subtreenode2content[subtree]]) all_sp_in_subtree = subtreenode2content[subtree] # Visit all nodes in the tree for n in subtree.traverse("preorder"): if not n.is_leaf(): c1 = subtreenode2content[n.children[0]] c2 = subtreenode2content[n.children[1]] #branches_found.append([all_sp_in_subtree, c1, c2]) for refnode, m1, m2 in refclades: all_expected_sp = m1 | m2 # We add one supporting point to every observed split that coincides # with a reference tree branch. This is, seqs in one side and seqs # on the other side of the observed split matches a ref_tree branch # without having extra seqs in any of the sides. However, we allow # for split matches where some seqs are lost in the observed split. #for all_sp_in_subtree, c1, c2 in branches_found: all_seen_sp = c1 | c2 notfound, found = 0, 0 false_missing = (all_expected_sp - all_seen_sp) & all_sp_in_subtree outside_species = (all_seen_sp - all_expected_sp) # Compare expected (m1,m2) splits with observed splits (c1,c2). a_straight = m1 & c1 b_straight = m2 & c2 a_cross = m1 & c2 b_cross = m2 & c1 # if matches are found for one of the first possible comparison if (a_straight and b_straight): # and the match contains all the observed species, species # from both sides are not mixed and missing species are real if not outside_species and not a_cross and not b_cross and not false_missing: found += 1 else: notfound += 1 # if matches are found for the second possible comparison (This # would never occur if found variable was increased in the # previous if) if (a_cross and b_cross): # and the match contains all the observed species, species # from both sides are not mixed and missing species are real if not outside_species and not a_straight and not b_straight and not false_missing: found += 1 else: notfound += 1 if notfound > 0: container[refnode].append(0) elif found > 0: container[refnode].append(1) if found == 2: raw_input( "Two possible matches? This should never occur!!") return branch2supports, branch2coll_supports
def main(argv): parser = argparse.ArgumentParser(description=__DESCRIPTION__, formatter_class=argparse.RawDescriptionHelpFormatter) parser.add_argument("target_trees", metavar='target_trees', type=str, nargs="*", help='a list of target tree files') parser.add_argument("--targets_file", dest="targets_file", type=str, help="""path to a file containing target trees, one per line""") parser.add_argument("-o", dest="output", type=str, help="""Path to the tab delimited report file""") parser.add_argument("-r", dest="reftree", type=str, required=True, help="""Reference tree""") parser.add_argument("--outgroup", dest="outgroup", nargs = "+", help="""outgroup used to root reference and target trees before distance computation""") parser.add_argument("--expand_polytomies", dest="polytomies", action = "store_true", help="""expand politomies if necessary""") parser.add_argument("--unrooted", dest="unrooted", action = "store_true", help="""compare trees as unrooted""") parser.add_argument("--min_support", dest="min_support", type=float, default=0.0, help=("min support value for branches to be counted in the distance computation (RF, treeko and refTree/targeGene compatibility)")) parser.add_argument("--extract_species", dest="extract_species", action = "store_true", help="""When used, reference tree is assumed to contain species names, while target trees as expected to be gene trees. Species name will be extracted from gene tree nodes and treeko will be used if duplication events are found.""") parser.add_argument("--spname_delimiter", dest="spname_delimiter", type=str, default="_", help=("species code delimiter in node names")) parser.add_argument("--spname_field", dest="spname_field", type=int, default=-1, help=("position of the species code extracted from node names. -1 = last field")) parser.add_argument("--collateral", dest="collateral", action='store_true', help=("")) parser.add_argument("--ref_attr", dest="ref_attr", type=str, help=("attribute in ref tree used as leaf name")) parser.add_argument("--target_attr", dest="target_attr", type=str, help=("attribute in target tree used as leaf name")) args = parser.parse_args(argv) print __DESCRIPTION__ reftree = args.reftree if args.targets_file and args.target_trees: print >>sys.stderr, 'The use of targets_file and targets at the same time is not supported.' sys.exit(1) if args.targets_file: target_trees = tree_iterator(args.targets_file) else: target_trees = args.target_trees t = Tree(reftree) if args.ref_attr: for lf in t.iter_leaves(): lf._origname = lf.name if args.ref_attr not in lf.features: print lf lf.name = getattr(lf, args.ref_attr) if args.outgroup: if len(args.outgroup) > 1: out = t.get_common_ancestor(args.outgroup) else: out = t.search_nodes(name=args.outgroup[0])[0] t.set_outgroup(out) ref_names = set(t.get_leaf_names()) reftree_len = len(t) reftree_edges = (reftree_len*2)-2 ncollapsed_branches = len([n for n in t.traverse() if n.children and n.support < args.min_support]) #reftree_edges -= ncollapsed_branches #if ncollapsed_branches: # print '%d branches collapsed in reference tree' %ncollapsed_branches HEADER = ("target tree", 'dups', 'subtrees', 'used trees', 'treeko', "RF", "maxRF", 'normRF', "%reftree", "%genetree", "avgSize", "minSize", "common tips", "refSize", "targetSize") if args.output: OUT = open(args.output, "w") print >>OUT, '# ' + ctime() print >>OUT, '# ' + ' '.join(sys.argv) print >>OUT, '#'+'\t'.join(HEADER) else: print '# ' + ctime() print '# ' + ' '.join(sys.argv) COL_WIDTHS = [20, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7] print_table([HEADER], fix_col_width=COL_WIDTHS, wrap_style='wrap') prev_tree = None for counter, tfile in enumerate(target_trees): if args.targets_file: seedid, tfile = tfile else: seedid = None if args.extract_species: tt = PhyloTree(tfile, sp_naming_function = lambda name: name.split(args.spname_delimiter)[args.spname_field]) else: tt = Tree(tfile) if args.target_attr: for lf in tt.iter_leaves(): lf._origname = lf.name lf.name = getattr(lf, args.target_attr) if args.outgroup: if len(args.outgroup) > 1: out = tt.get_common_ancestor(args.outgroup) else: out = tt.search_nodes(name=args.outgroup[0])[0] tt.set_outgroup(out) if args.target_trees: fname = os.path.basename(tfile) else: fname = '%05d' %counter max_size, min_size, avg_size, common = -1, -1, -1, -1 total_rf, max_rf, norm_rf = -1, -1, -1 treeko_d = -1 ref_branches_in_target, target_branches_in_ref = -1, -1 target_tree_len = -1 used_subtrees = -1 if args.extract_species: orig_target_size = len(tt) ntrees, ndups, sp_trees = tt.get_speciation_trees(autodetect_duplications=True, newick_only=True) if ntrees < 1000: all_rf = [] ref_found = [] target_found = [] tree_sizes = [] all_max_rf = [] common_names = 0 for subtree_nw in sp_trees: if seedid and not args.collateral and (seedid not in subtree_nw): continue subtree = PhyloTree(subtree_nw, sp_naming_function = lambda name: name.split(args.spname_delimiter)[args.spname_field]) # only necessary if rf function is going to filter by support value. It slows downs the analysis, obviously if args.min_support: subtree_content = subtree.get_cached_content(store_attr='name') for n in subtree.traverse(): if n.children: n.support = tt.get_common_ancestor(subtree_content[n]).support rf, maxr, common, p1, p2, d1, d2 = t.robinson_foulds(subtree, expand_polytomies=args.polytomies, unrooted_trees=args.unrooted, attr_t2='species', min_support_t2=args.min_support) if maxr > 0 and p1 and p2: all_rf.append(rf) tree_sizes.append(len(common)) all_max_rf.append(maxr) common_names = max(common_names, len(common)) ref_found.append(float(len(p2 & p1)) / reftree_edges) p2bis = set([p for p in (p2-d2) if len(p[0])>1 and len(p[1])>1]) # valid edges in target not leaves if p2bis: incompatible_target_branches = float(len((p2-d2) - p1)) target_found.append(1 - (incompatible_target_branches / (len(p2-d2)))) # valid_target = p2-d2 # valid_ref = p1-d1 # ref_found.append(float(len(valid_target & valid_ref)) / reftree_edges) # p2bis = set([p for p in (p2-d2) if len(p[0])>1 and len(p[1])>1]) # if p2bis-d2: # incompatible_target_branches = float(len((p2-d2) - p1)) # target_found.append(1 - (incompatible_target_branches / (len(p2-d2)))) if all_rf: # Treeko speciation distance alld = [(all_rf[i]/float(all_max_rf[i])) for i in xrange(len(all_rf))] a = numpy.sum([alld[i] * tree_sizes[i] for i in xrange(len(all_rf))]) b = float(numpy.sum(tree_sizes)) treeko_d = a/b total_rf = numpy.mean(all_rf) norm_rf = numpy.mean([(all_rf[i]/float(all_max_rf[i])) for i in xrange(len(all_rf))]) max_rf = numpy.max(all_max_rf) ref_branches_in_target = numpy.mean(ref_found) target_branches_in_ref = numpy.mean(target_found) if target_found else -1 target_tree_len = numpy.mean(tree_sizes) used_subtrees = len(all_rf) else: target_tree_len = len(tt) ndups, ntrees, used_subtrees = 0, 1, 1 treeko_d = -1 total_rf, max_rf, common, p1, p2, d1, d2 = tt.robinson_foulds(t, expand_polytomies=args.polytomies, unrooted_trees=args.unrooted) common_names = len(common) if max_rf: norm_rf = total_rf / float(max_rf) if p1 and p2: sizes = [len(p) for p in p2 ^ p1] if sizes: avg_size = sum(sizes) / float(len(sizes)) max_size, min_size = max(sizes), min(sizes) else: max_size, min_size, avg_size = 0, 0, 0 ref_branches_in_target = float(len(p2 & p1)) / reftree_edges #if p2-d2: # incompatible_target_branches = float(len((p2-d2) - p1)) # target_found.append(1 - (incompatible_target_branches / (len(p2-d2)))) else: ref_branches_in_target = 0.0 target_branches_in_ref = 0.0 max_size, min_size, avg_size = -1, -1, -1 if args.output: print >>OUT, '\t'.join(map(str, (fname, ndups, ntrees, used_subtrees, treeko_d, total_rf, max_rf, norm_rf, ref_branches_in_target, target_branches_in_ref, avg_size, min_size, common_names, reftree_len, target_tree_len))) else: print_table([map(istr, (fname[-30:], ndups, ntrees, used_subtrees, treeko_d, total_rf, max_rf, norm_rf, '%0.4f' %ref_branches_in_target, '%0.4f' %target_branches_in_ref, avg_size, min_size, common_names, reftree_len, target_tree_len))], fix_col_width = COL_WIDTHS, wrap_style='cut') if args.output: OUT.close()
def main(argv): global args parser = argparse.ArgumentParser(description=__DESCRIPTION__, formatter_class=argparse.RawDescriptionHelpFormatter) parser.add_argument("-r", dest="reftree", type=str, required=True, help="""Reference tree""") parser.add_argument("--source_trees", dest="source_trees", type=str, required = True, help=("A list of *rooted* genetrees, one per line, in the format: TreeID/SeedID [TAB] newick ")) parser.add_argument("--plot_newick", dest="plot_newick", type=str, help=("")) parser.add_argument("--spname_delimiter", dest="spname_delimiter", type=str, default="_", help=("species code delimiter in node names")) parser.add_argument("--spname_field", dest="spname_field", type=int, default=-1, help=("position of the species code extracted from node names. -1 = last field")) parser.add_argument("--collateral", dest="use_collateral", action="store_true", help=("If enabled, collateral information will be used as" " equally qualified data. Otherwise, such data will" " be reported separatedly. Use this if your set of" " trees are not overlaping. ")) parser.add_argument("--skip_dup_detection", dest="skip_dup_detection", action="store_true", help=('If used, duplications will be expected to be annotated' ' in the source gene trees with the evoltype="D" tag.' ' Otherwise they will be inferred on the fly using' ' the species overlap algorithm.')) parser.add_argument("--spoverlap", dest="species_overlap", type=float, default=0.0, help=("Species overlap cutoff. A number between 0 and 1 " "representing the percentage of species that should be " "shared between two sister partitions to be considered a" " duplication. 0 = any overlap represents a duplication. ")) parser.add_argument("--debug", dest="debug", action="store_true", help=("generate an image of every input gene tree tree, so the result can be inspected")) parser.add_argument("--snapshot_step", dest="snapshot_step", type=int, default=1000, help=("How many trees should be processed between snapshots dumps?")) parser.add_argument("--reftree_constraint", dest="reftree_constraint", type=str, help=("A python module from from which a function called " "*is_valid_treeid(treeid, refbranch)* should be importable. " "The function will be used to decide if the info of a given " "source tree is informative or not for each reftree branch. ")) parser.add_argument("-o", dest="output", type=str, required=True, help=("output tag name (extensions will be added)")) parser.add_argument("--cpu", dest="cpu", type=int, default=1, help=("enable parallel computation")) parser.add_argument("--img_report", dest="img_report", action="store_true", help=("If true, it generates a summary image results with all the computed data")) parser.add_argument("--report_supports", dest="report_supports", action="store_true", help=("If used, supported ref tree branches are individually reported for each gene tree ")) args = parser.parse_args(argv) if args.plot_newick: t = Tree(args.plot_newick) ts = TreeStyle() ts.layout_fn = info_layout t.render("tree_analysis.png", tree_style=ts) sys.exit(0) SPNAME_FIELD, SPNAME_DELIMITER = args.spname_field, args.spname_delimiter USE_COLLATERAL = args.use_collateral DETECT_DUPLICATIONS = True if not args.skip_dup_detection else False REPORT_PER_TREE_SUPPORTS = True if args.report_supports else False SP_OVERLAP = args.species_overlap DEBUG = args.debug IMG_REPORT = args.img_report reftree = PhyloTree(args.reftree, sp_naming_function=None) for nid, n in enumerate(reftree.traverse()): n.add_features(nid = nid) REFTREE_SPECIES = set(reftree.get_leaf_names()) print __DESCRIPTION__ if REPORT_PER_TREE_SUPPORTS: REPORT_SUPPORT_FILE = open("%s.gentree_supports" %args.output, "w") print >>REPORT_SUPPORT_FILE, '#'+'\t'.join(map(str, ["treeId", "spCoverage", "mean_support", "mean_coll_support", "tested_branches", 'tested_coll_branches'])) TOTAL_TREES = int(commands.getoutput("wc -l %s" %args.source_trees).split()[0]) + 1 print >>sys.stderr, "Processing %d source trees" %TOTAL_TREES if args.reftree_constraint: import imp constraint = imp.load_source('constraint', args.reftree_constraint) IS_VALID_TREEID = constraint.is_valid_treeid else: IS_VALID_TREEID = None if args.cpu > 1: MONITOR_STEP = 0 #return (informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, # coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) # The output of the process_trees function are 9 dictionaries in which keys are refbranches target_dicts = [{} for x in range(9)] def merge_dict_results(target, source): def merge_dict(target, source): for k, v in source.iteritems(): if k not in target: target[k] = v elif isinstance(v, list): target[k].extend(v) elif isinstance(v, set): target[k].update(v) elif isinstance(v, int): target[k] += v else: raise ValueError("Impossible to merge str results") for index in xrange(len(target)): merge_dict(target[index], out[index]) from multiprocessing import Process, Queue from Queue import Empty as QueueEmpty outputs_queue = Queue() if TOTAL_TREES > args.cpu: trees_per_cpu = TOTAL_TREES / args.cpu trees_per_cpu += 1 if TOTAL_TREES % args.cpu else 0 else: trees_per_cpu = 1 args.cpu = TOTAL_TREES all_workers = set() for cpu_num in xrange(args.cpu): sline = (cpu_num*trees_per_cpu) eline = (cpu_num*trees_per_cpu) + trees_per_cpu data_iter = tree_iterator(args.source_trees, restrict_species=REFTREE_SPECIES, start_line=sline, end_line=eline) print >>sys.stderr, "Launching worker %d from %d to %d" %(cpu_num, sline, eline) worker = Process(target=run_parallel, args=(cpu_num, outputs_queue, process_trees, data_iter, reftree, trees_per_cpu)) worker.name = "Worker_%d" %cpu_num all_workers.add(worker) worker.start() while all_workers: # clear done threads for w in list(all_workers): if not w.is_alive(): print >>sys.stderr, "%s thread is done!" %w.name all_workers.discard(w) # get and merge results while 1: try: out = outputs_queue.get(False) except QueueEmpty: break else: # This merge depends on process_trees return output!!!!! merge_dict_results(target_dicts, out) # Dump a snapshot dump_results(reftree, *target_dicts) time.sleep(0.1) if all_workers: time.sleep(1) # collected data (informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) = target_dicts else: MONITOR_STEP = args.snapshot_step data_iter = tree_iterator(args.source_trees, restrict_species=REFTREE_SPECIES) (informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) = process_trees(data_iter, reftree, TOTAL_TREES) if REPORT_PER_TREE_SUPPORTS: REPORT_SUPPORT_FILE.close() dump_results(reftree, informed_branches, dup_per_branch, losses_per_branch, losses_per_dup_branch, refbranch_supports, coll_dup_per_branch, coll_losses_per_branch, coll_losses_per_dup_branch, coll_refbranch_supports) print >>sys.stderr, "Dumping full analysis..." # Full dump, including duplication details cPickle.dump(reftree, open("%s.pkl"%args.output, "w"))
def get_supported_branches(source_tree, reftree, refclades, seedid=None): """ Given a reference species tree and a rooted gene tree in which duplication events are already mapped, this function does the following: - Split gene tree into all possible species tree (Treeko method) - Find matches between each subtree branch and all branches in the reference tree. - Each branch in each species subtree is compared to all branches in the reftree. If left/right side of the subtree branch coincide with a the left/right side of a reference tree branch, this is considered a gene tree support point. Coincidences must comply with the following conditions: - All species in the left/right sides of the subtree branch exist in the left/right sides of the reference branch. - Species in the left/right sides of the reference branch are never mixed in the subtree branch. - Missing species are allowed in the subtree split, only if such species are not present in any other part of the original gene tree. """ # Run Treeko to get all possible species tree combinations. We assume dups are already mapped ntrees, ndups, sp_trees = source_tree.get_speciation_trees(autodetect_duplications=DETECT_DUPLICATIONS, newick_only=True) if ntrees > 100000: return {}, {} branches_found = [] branch2supports = defaultdict(list) branch2coll_supports = defaultdict(list) for nw in sp_trees: # Use all treeko trees or only those subtrees containing the seed? if seedid and seedid not in nw: container = branch2coll_supports else: container = branch2supports subtree = PhyloTree(nw, sp_naming_function = extract_species) subtreenode2content = subtree.get_cached_content(store_attr="species") #set([phy3(_c.name) for _c in subtreenode2content[subtree]]) all_sp_in_subtree = subtreenode2content[subtree] # Visit all nodes in the tree for n in subtree.traverse("preorder"): if not n.is_leaf(): c1 = subtreenode2content[n.children[0]] c2 = subtreenode2content[n.children[1]] #branches_found.append([all_sp_in_subtree, c1, c2]) for refnode, m1, m2 in refclades: all_expected_sp = m1 | m2 # We add one supporting point to every observed split that coincides # with a reference tree branch. This is, seqs in one side and seqs # on the other side of the observed split matches a ref_tree branch # without having extra seqs in any of the sides. However, we allow # for split matches where some seqs are lost in the observed split. #for all_sp_in_subtree, c1, c2 in branches_found: all_seen_sp = c1|c2 notfound, found = 0, 0 false_missing = (all_expected_sp - all_seen_sp) & all_sp_in_subtree outside_species = (all_seen_sp - all_expected_sp) # Compare expected (m1,m2) splits with observed splits (c1,c2). a_straight = m1 & c1 b_straight = m2 & c2 a_cross = m1 & c2 b_cross = m2 & c1 # if matches are found for one of the first possible comparison if (a_straight and b_straight): # and the match contains all the observed species, species # from both sides are not mixed and missing species are real if not outside_species and not a_cross and not b_cross and not false_missing: found += 1 else: notfound += 1 # if matches are found for the second possible comparison (This # would never occur if found variable was increased in the # previous if) if (a_cross and b_cross): # and the match contains all the observed species, species # from both sides are not mixed and missing species are real if not outside_species and not a_straight and not b_straight and not false_missing: found += 1 else: notfound += 1 if notfound > 0: container[refnode].append(0) elif found > 0: container[refnode].append(1) if found == 2: raw_input("Two possible matches? This should never occur!!") return branch2supports, branch2coll_supports