def load_refjson(self, refjson_fname):
try:
self.refjson = RefJsonParser(refjson_fname)
except ValueError:
self.cfg.exit_user_error("ERROR: Invalid json file format!")
#validate input json format
(valid, err) = self.refjson.validate()
if not valid:
self.cfg.log.error(
"ERROR: Parsing reference JSON file failed:\n%s", err)
self.cfg.exit_user_error()
self.rate = self.refjson.get_rate()
self.node_height = self.refjson.get_node_height()
self.origin_taxonomy = self.refjson.get_origin_taxonomy()
self.tax_tree = self.refjson.get_tax_tree()
self.cfg.compress_patterns = self.refjson.get_pattern_compression()
self.bid_taxonomy_map = self.refjson.get_branch_tax_map()
if not self.bid_taxonomy_map:
# old file format (before 1.6), need to rebuild this map from scratch
th = TaxTreeHelper(self.cfg, self.origin_taxonomy)
th.set_mf_rooted_tree(self.tax_tree)
th.set_bf_unrooted_tree(self.refjson.get_reftree())
self.bid_taxonomy_map = th.get_bid_taxonomy_map()
self.write_bid_tax_map(self.bid_taxonomy_map, final=False)
reftree_str = self.refjson.get_raxml_readable_tree()
self.reftree = Tree(reftree_str)
self.reftree_size = len(self.reftree.get_leaves())
# IMPORTANT: set EPA heuristic rate based on tree size!
self.cfg.resolve_auto_settings(self.reftree_size)
# If we're loading the pre-optimized model, we MUST set the same rate het. mode as in the ref file
if self.cfg.epa_load_optmod:
self.cfg.raxml_model = self.refjson.get_ratehet_model()
self.classify_helper = TaxClassifyHelper(self.cfg,
self.bid_taxonomy_map,
self.rate, self.node_height)
self.taxtree_helper = TaxTreeHelper(self.cfg, self.origin_taxonomy,
self.tax_tree)
tax_code_name = self.refjson.get_taxcode()
self.tax_code = TaxCode(tax_code_name)
self.taxonomy = Taxonomy(prefix=EpacConfig.REF_SEQ_PREFIX,
tax_map=self.origin_taxonomy)
self.tax_common_ranks = self.taxonomy.get_common_ranks()
# print "Common ranks: ", self.tax_common_ranks
self.mislabels_cnt = [0] * TaxCode.UNI_TAX_LEVELS
self.rank_mislabels_cnt = [0] * TaxCode.UNI_TAX_LEVELS
def load_refjson(self, refjson_fname):
try:
self.refjson = RefJsonParser(refjson_fname)
except ValueError:
self.cfg.exit_user_error("ERROR: Invalid json file format!")
#validate input json format
(valid, err) = self.refjson.validate()
if not valid:
self.cfg.log.error("ERROR: Parsing reference JSON file failed:\n%s", err)
self.cfg.exit_user_error()
self.rate = self.refjson.get_rate()
self.node_height = self.refjson.get_node_height()
self.origin_taxonomy = self.refjson.get_origin_taxonomy()
self.tax_tree = self.refjson.get_tax_tree()
self.cfg.compress_patterns = self.refjson.get_pattern_compression()
self.bid_taxonomy_map = self.refjson.get_branch_tax_map()
if not self.bid_taxonomy_map:
# old file format (before 1.6), need to rebuild this map from scratch
th = TaxTreeHelper(self.cfg, self.origin_taxonomy)
th.set_mf_rooted_tree(self.tax_tree)
th.set_bf_unrooted_tree(self.refjson.get_reftree())
self.bid_taxonomy_map = th.get_bid_taxonomy_map()
self.write_bid_tax_map(self.bid_taxonomy_map, final=False)
reftree_str = self.refjson.get_raxml_readable_tree()
self.reftree = Tree(reftree_str)
self.reftree_size = len(self.reftree.get_leaves())
# IMPORTANT: set EPA heuristic rate based on tree size!
self.cfg.resolve_auto_settings(self.reftree_size)
# If we're loading the pre-optimized model, we MUST set the same rate het. mode as in the ref file
if self.cfg.epa_load_optmod:
self.cfg.raxml_model = self.refjson.get_ratehet_model()
self.classify_helper = TaxClassifyHelper(self.cfg, self.bid_taxonomy_map, self.rate, self.node_height)
self.taxtree_helper = TaxTreeHelper(self.cfg, self.origin_taxonomy, self.tax_tree)
tax_code_name = self.refjson.get_taxcode()
self.tax_code = TaxCode(tax_code_name)
self.taxonomy = Taxonomy(prefix=EpacConfig.REF_SEQ_PREFIX, tax_map=self.origin_taxonomy)
self.tax_common_ranks = self.taxonomy.get_common_ranks()
# print "Common ranks: ", self.tax_common_ranks
self.mislabels_cnt = [0] * TaxCode.UNI_TAX_LEVELS
self.rank_mislabels_cnt = [0] * TaxCode.UNI_TAX_LEVELS
class TaxTreeHelperTests(unittest.TestCase):
def setUp(self):
self.testfile_dir = os.path.join(
os.path.dirname(os.path.abspath(__file__)), "testfiles")
self.tax_fname = os.path.join(self.testfile_dir, "test_clean.tax")
self.taxonomy = Taxonomy(EpacConfig.REF_SEQ_PREFIX, self.tax_fname)
tax_map = self.taxonomy.get_map()
cfg = EpacConfig()
self.taxtree_helper = TaxTreeHelper(cfg, tax_map)
outgr_fname = os.path.join(self.testfile_dir, "outgroup.nw")
self.expected_outgr = Tree(outgr_fname)
def tearDown(self):
self.taxonomy = None
self.taxtree_helper = None
def test_outgroup(self):
mfu_tree_fname = os.path.join(self.testfile_dir, "taxtree.nw")
mfu_tree = Tree(mfu_tree_fname)
self.taxtree_helper.set_mf_rooted_tree(mfu_tree)
outgr = self.taxtree_helper.get_outgroup()
self.assertEqual(outgr.get_leaf_names(),
self.expected_outgr.get_leaf_names())
def test_branch_labeling(self):
bfu_tree_fname = os.path.join(self.testfile_dir, "resolved_tree.nw")
bfu_tree = Tree(bfu_tree_fname)
map_fname = os.path.join(self.testfile_dir, "bid_tax_map.txt")
self.expected_map = {}
with open(map_fname) as inf:
for line in inf:
bid, rank_id, rdiff, brlen = line.strip().split("\t")
self.expected_map[bid] = (rank_id, int(rdiff), float(brlen))
self.taxtree_helper.set_outgroup(self.expected_outgr)
self.taxtree_helper.set_bf_unrooted_tree(bfu_tree)
bid_tax_map = self.taxtree_helper.get_bid_taxonomy_map()
self.assertEqual(len(bid_tax_map), 2 * len(bfu_tree) - 3)
for bid in self.expected_map.iterkeys():
e_rec = self.expected_map[bid]
rec = bid_tax_map[bid]
self.assertEqual(e_rec[0], rec[0])
self.assertEqual(e_rec[1], rec[1])
self.assertAlmostEqual(e_rec[2], rec[2], 6)
def __init__(self, config, args):
self.cfg = config
self.jplace_fname = args.jplace_fname
self.ignore_refalign = args.ignore_refalign
self.tmp_refaln = config.tmp_fname("%NAME%.refaln")
#here is the final alignment file for running EPA
self.epa_alignment = config.tmp_fname("%NAME%.afa")
self.hmmprofile = config.tmp_fname("%NAME%.hmmprofile")
self.tmpquery = config.tmp_fname("%NAME%.tmpquery")
self.noalign = config.tmp_fname("%NAME%.noalign")
self.seqs = None
assign_fname = args.output_name + ".assignment.txt"
self.out_assign_fname = os.path.join(args.output_dir, assign_fname)
jplace_fname = args.output_name + ".jplace"
self.out_jplace_fname = os.path.join(args.output_dir, jplace_fname)
try:
self.refjson = RefJsonParser(config.refjson_fname)
except ValueError:
self.cfg.exit_user_error("Invalid json file format: %s" %
config.refjson_fname)
#validate input json format
self.refjson.validate()
self.reftree = self.refjson.get_reftree()
self.rate = self.refjson.get_rate()
self.node_height = self.refjson.get_node_height()
self.cfg.compress_patterns = self.refjson.get_pattern_compression()
self.bid_taxonomy_map = self.refjson.get_branch_tax_map()
if not self.bid_taxonomy_map:
# old file format (before 1.6), need to rebuild this map from scratch
th = TaxTreeHelper(self.cfg, self.refjson.get_origin_taxonomy())
th.set_mf_rooted_tree(self.refjson.get_tax_tree())
th.set_bf_unrooted_tree(self.refjson.get_reftree())
self.bid_taxonomy_map = th.get_bid_taxonomy_map()
self.cfg.log.info("Loaded reference tree with %d taxa\n" %
len(self.reftree.get_leaves()))
self.classify_helper = TaxClassifyHelper(self.cfg,
self.bid_taxonomy_map,
self.rate, self.node_height)
class TaxTreeHelperTests(unittest.TestCase):
def setUp(self):
self.testfile_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "testfiles")
self.tax_fname = os.path.join(self.testfile_dir, "test_clean.tax")
self.taxonomy = Taxonomy(EpacConfig.REF_SEQ_PREFIX, self.tax_fname)
tax_map = self.taxonomy.get_map()
cfg = EpacConfig()
self.taxtree_helper = TaxTreeHelper(cfg, tax_map)
outgr_fname = os.path.join(self.testfile_dir, "outgroup.nw")
self.expected_outgr = Tree(outgr_fname)
def tearDown(self):
self.taxonomy = None
self.taxtree_helper = None
def test_outgroup(self):
mfu_tree_fname = os.path.join(self.testfile_dir, "taxtree.nw")
mfu_tree = Tree(mfu_tree_fname)
self.taxtree_helper.set_mf_rooted_tree(mfu_tree)
outgr = self.taxtree_helper.get_outgroup()
self.assertEqual(outgr.get_leaf_names(), self.expected_outgr.get_leaf_names())
def test_branch_labeling(self):
bfu_tree_fname = os.path.join(self.testfile_dir, "resolved_tree.nw")
bfu_tree = Tree(bfu_tree_fname)
map_fname = os.path.join(self.testfile_dir, "bid_tax_map.txt")
self.expected_map = {}
with open(map_fname) as inf:
for line in inf:
bid, rank_id, rdiff, brlen = line.strip().split("\t")
self.expected_map[bid] = (rank_id, int(rdiff), float(brlen))
self.taxtree_helper.set_outgroup(self.expected_outgr)
self.taxtree_helper.set_bf_unrooted_tree(bfu_tree)
bid_tax_map = self.taxtree_helper.get_bid_taxonomy_map()
self.assertEqual(len(bid_tax_map), 2 * len(bfu_tree) - 3)
for bid in self.expected_map.iterkeys():
e_rec = self.expected_map[bid]
rec = bid_tax_map[bid]
self.assertEqual(e_rec[0], rec[0])
self.assertEqual(e_rec[1], rec[1])
self.assertAlmostEqual(e_rec[2], rec[2], 6)
def __init__(self, config, args):
self.cfg = config
self.jplace_fname = args.jplace_fname
self.ignore_refalign = args.ignore_refalign
self.tmp_refaln = config.tmp_fname("%NAME%.refaln")
#here is the final alignment file for running EPA
self.epa_alignment = config.tmp_fname("%NAME%.afa")
self.hmmprofile = config.tmp_fname("%NAME%.hmmprofile")
self.tmpquery = config.tmp_fname("%NAME%.tmpquery")
self.noalign = config.tmp_fname("%NAME%.noalign")
self.seqs = None
assign_fname = args.output_name + ".assignment.txt"
self.out_assign_fname = os.path.join(args.output_dir, assign_fname)
jplace_fname = args.output_name + ".jplace"
self.out_jplace_fname = os.path.join(args.output_dir, jplace_fname)
try:
self.refjson = RefJsonParser(config.refjson_fname)
except ValueError:
self.cfg.exit_user_error("Invalid json file format: %s" % config.refjson_fname)
#validate input json format
self.refjson.validate()
self.reftree = self.refjson.get_reftree()
self.rate = self.refjson.get_rate()
self.node_height = self.refjson.get_node_height()
self.cfg.compress_patterns = self.refjson.get_pattern_compression()
self.bid_taxonomy_map = self.refjson.get_branch_tax_map()
if not self.bid_taxonomy_map:
# old file format (before 1.6), need to rebuild this map from scratch
th = TaxTreeHelper(self.cfg, self.refjson.get_origin_taxonomy())
th.set_mf_rooted_tree(self.refjson.get_tax_tree())
th.set_bf_unrooted_tree(self.refjson.get_reftree())
self.bid_taxonomy_map = th.get_bid_taxonomy_map()
self.cfg.log.info("Loaded reference tree with %d taxa\n" % len(self.reftree.get_leaves()))
self.classify_helper = TaxClassifyHelper(self.cfg, self.bid_taxonomy_map, self.rate, self.node_height)
class RefTreeBuilder:
def __init__(self, config):
self.cfg = config
self.mfresolv_job_name = self.cfg.subst_name("mfresolv_%NAME%")
self.epalbl_job_name = self.cfg.subst_name("epalbl_%NAME%")
self.optmod_job_name = self.cfg.subst_name("optmod_%NAME%")
self.raxml_wrapper = RaxmlWrapper(config)
self.outgr_fname = self.cfg.tmp_fname("%NAME%_outgr.tre")
self.reftree_mfu_fname = self.cfg.tmp_fname("%NAME%_mfu.tre")
self.reftree_bfu_fname = self.cfg.tmp_fname("%NAME%_bfu.tre")
self.optmod_fname = self.cfg.tmp_fname("%NAME%.opt")
self.lblalign_fname = self.cfg.tmp_fname("%NAME%_lblq.fa")
self.reftree_lbl_fname = self.cfg.tmp_fname("%NAME%_lbl.tre")
self.reftree_tax_fname = self.cfg.tmp_fname("%NAME%_tax.tre")
self.brmap_fname = self.cfg.tmp_fname("%NAME%_map.txt")
def load_alignment(self):
in_file = self.cfg.align_fname
self.input_seqs = None
formats = ["fasta", "phylip_relaxed", "iphylip_relaxed", "phylip", "iphylip"]
for fmt in formats:
try:
self.input_seqs = SeqGroup(sequences=in_file, format = fmt)
break
except:
self.cfg.log.debug("Guessing input format: not " + fmt)
if self.input_seqs == None:
self.cfg.exit_user_error("Invalid input file format: %s\nThe supported input formats are fasta and phylip" % in_file)
def validate_taxonomy(self):
self.input_validator = InputValidator(self.cfg, self.taxonomy, self.input_seqs)
self.input_validator.validate()
def build_multif_tree(self):
c = self.cfg
tb = TaxTreeBuilder(c, self.taxonomy)
(t, ids) = tb.build(c.reftree_min_rank, c.reftree_max_seqs_per_leaf, c.reftree_clades_to_include, c.reftree_clades_to_ignore)
self.reftree_ids = frozenset(ids)
self.reftree_size = len(ids)
self.reftree_multif = t
# IMPORTANT: select GAMMA or CAT model based on tree size!
self.cfg.resolve_auto_settings(self.reftree_size)
if self.cfg.debug:
refseq_fname = self.cfg.tmp_fname("%NAME%_seq_ids.txt")
# list of sequence ids which comprise the reference tree
with open(refseq_fname, "w") as f:
for sid in ids:
f.write("%s\n" % sid)
# original tree with taxonomic ranks as internal node labels
reftax_fname = self.cfg.tmp_fname("%NAME%_mfu_tax.tre")
t.write(outfile=reftax_fname, format=8)
# t.show()
def export_ref_alignment(self):
"""This function transforms the input alignment in the following way:
1. Filter out sequences which are not part of the reference tree
2. Add sequence name prefix (r_)"""
self.refalign_fname = self.cfg.tmp_fname("%NAME%_matrix.afa")
with open(self.refalign_fname, "w") as fout:
for name, seq, comment, sid in self.input_seqs.iter_entries():
seq_name = EpacConfig.REF_SEQ_PREFIX + name
if seq_name in self.input_validator.corr_seqid:
seq_name = self.input_validator.corr_seqid[seq_name]
if seq_name in self.reftree_ids:
fout.write(">" + seq_name + "\n" + seq + "\n")
# we do not need the original alignment anymore, so free its memory
self.input_seqs = None
def export_ref_taxonomy(self):
self.taxonomy_map = {}
for sid, ranks in self.taxonomy.iteritems():
if sid in self.reftree_ids:
self.taxonomy_map[sid] = ranks
if self.cfg.debug:
tax_fname = self.cfg.tmp_fname("%NAME%_tax.txt")
with open(tax_fname, "w") as fout:
for sid, ranks in self.taxonomy_map.iteritems():
ranks_str = self.taxonomy.seq_lineage_str(sid)
fout.write(sid + "\t" + ranks_str + "\n")
def save_rooting(self):
rt = self.reftree_multif
tax_map = self.taxonomy.get_map()
self.taxtree_helper = TaxTreeHelper(self.cfg, tax_map)
self.taxtree_helper.set_mf_rooted_tree(rt)
outgr = self.taxtree_helper.get_outgroup()
outgr_size = len(outgr.get_leaves())
outgr.write(outfile=self.outgr_fname, format=9)
self.reftree_outgroup = outgr
self.cfg.log.debug("Outgroup for rooting was saved to: %s, outgroup size: %d", self.outgr_fname, outgr_size)
# remove unifurcation at the root
if len(rt.children) == 1:
rt = rt.children[0]
# now we can safely unroot the tree and remove internal node labels to make it suitable for raxml
rt.write(outfile=self.reftree_mfu_fname, format=9)
# RAxML call to convert multifurcating tree to the strictly bifurcating one
def resolve_multif(self):
self.cfg.log.debug("\nReducing the alignment: \n")
self.reduced_refalign_fname = self.raxml_wrapper.reduce_alignment(self.refalign_fname)
self.cfg.log.debug("\nConstrained ML inference: \n")
raxml_params = ["-s", self.reduced_refalign_fname, "-g", self.reftree_mfu_fname, "--no-seq-check", "-N", str(self.cfg.rep_num)]
if self.cfg.mfresolv_method == "fast":
raxml_params += ["-D"]
elif self.cfg.mfresolv_method == "ultrafast":
raxml_params += ["-f", "e"]
if self.cfg.restart and self.raxml_wrapper.result_exists(self.mfresolv_job_name):
self.invocation_raxml_multif = self.raxml_wrapper.get_invocation_str(self.mfresolv_job_name)
self.cfg.log.debug("\nUsing existing ML tree found in: %s\n", self.raxml_wrapper.result_fname(self.mfresolv_job_name))
else:
self.invocation_raxml_multif = self.raxml_wrapper.run(self.mfresolv_job_name, raxml_params)
# self.invocation_raxml_multif = self.raxml_wrapper.run_multiple(self.mfresolv_job_name, raxml_params, self.cfg.rep_num)
if self.cfg.mfresolv_method == "ultrafast":
self.raxml_wrapper.copy_result_tree(self.mfresolv_job_name, self.raxml_wrapper.besttree_fname(self.mfresolv_job_name))
if self.raxml_wrapper.besttree_exists(self.mfresolv_job_name):
if not self.cfg.reopt_model:
self.raxml_wrapper.copy_best_tree(self.mfresolv_job_name, self.reftree_bfu_fname)
self.raxml_wrapper.copy_optmod_params(self.mfresolv_job_name, self.optmod_fname)
self.invocation_raxml_optmod = ""
job_name = self.mfresolv_job_name
else:
bfu_fname = self.raxml_wrapper.besttree_fname(self.mfresolv_job_name)
job_name = self.optmod_job_name
# RAxML call to optimize model parameters and write them down to the binary model file
self.cfg.log.debug("\nOptimizing model parameters: \n")
raxml_params = ["-f", "e", "-s", self.reduced_refalign_fname, "-t", bfu_fname, "--no-seq-check"]
if self.cfg.raxml_model.startswith("GTRCAT") and not self.cfg.compress_patterns:
raxml_params += ["-H"]
if self.cfg.restart and self.raxml_wrapper.result_exists(self.optmod_job_name):
self.invocation_raxml_optmod = self.raxml_wrapper.get_invocation_str(self.optmod_job_name)
self.cfg.log.debug("\nUsing existing optimized tree and parameters found in: %s\n", self.raxml_wrapper.result_fname(self.optmod_job_name))
else:
self.invocation_raxml_optmod = self.raxml_wrapper.run(self.optmod_job_name, raxml_params)
if self.raxml_wrapper.result_exists(self.optmod_job_name):
self.raxml_wrapper.copy_result_tree(self.optmod_job_name, self.reftree_bfu_fname)
self.raxml_wrapper.copy_optmod_params(self.optmod_job_name, self.optmod_fname)
else:
errmsg = "RAxML run failed (model optimization), please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.optmod_job_name)
self.cfg.exit_fatal_error(errmsg)
if self.cfg.raxml_model.startswith("GTRCAT"):
mod_name = "CAT"
else:
mod_name = "GAMMA"
self.reftree_loglh = self.raxml_wrapper.get_tree_lh(job_name, mod_name)
self.cfg.log.debug("\n%s-based logLH of the reference tree: %f\n" % (mod_name, self.reftree_loglh))
else:
errmsg = "RAxML run failed (mutlifurcation resolution), please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.mfresolv_job_name)
self.cfg.exit_fatal_error(errmsg)
def load_reduced_refalign(self):
formats = ["fasta", "phylip_relaxed"]
for fmt in formats:
try:
self.reduced_refalign_seqs = SeqGroup(sequences=self.reduced_refalign_fname, format = fmt)
break
except:
pass
if self.reduced_refalign_seqs == None:
errmsg = "FATAL ERROR: Invalid input file format in %s! (load_reduced_refalign)" % self.reduced_refalign_fname
self.cfg.exit_fatal_error(errmsg)
# dummy EPA run to label the branches of the reference tree, which we need to build a mapping to tax ranks
def epa_branch_labeling(self):
# create alignment with dummy query seq
self.refalign_width = len(self.reduced_refalign_seqs.get_seqbyid(0))
self.reduced_refalign_seqs.write(format="fasta", outfile=self.lblalign_fname)
with open(self.lblalign_fname, "a") as fout:
fout.write(">" + "DUMMY131313" + "\n")
fout.write("A"*self.refalign_width + "\n")
# TODO always load model regardless of the config file settings?
epa_result = self.raxml_wrapper.run_epa(self.epalbl_job_name, self.lblalign_fname, self.reftree_bfu_fname, self.optmod_fname, mode="epa_mp")
self.reftree_lbl_str = epa_result.get_std_newick_tree()
self.raxml_version = epa_result.get_raxml_version()
self.invocation_raxml_epalbl = epa_result.get_raxml_invocation()
if not self.raxml_wrapper.epa_result_exists(self.epalbl_job_name):
errmsg = "RAxML EPA run failed, please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.epalbl_job_name)
self.cfg.exit_fatal_error(errmsg)
def epa_post_process(self):
lbl_tree = Tree(self.reftree_lbl_str)
self.taxtree_helper.set_bf_unrooted_tree(lbl_tree)
self.reftree_tax = self.taxtree_helper.get_tax_tree()
self.bid_ranks_map = self.taxtree_helper.get_bid_taxonomy_map()
if self.cfg.debug:
self.reftree_tax.write(outfile=self.reftree_tax_fname, format=3)
with open(self.reftree_lbl_fname, "w") as outf:
outf.write(self.reftree_lbl_str)
with open(self.brmap_fname, "w") as outf:
for bid, br_rec in self.bid_ranks_map.iteritems():
outf.write("%s\t%s\t%d\t%f\n" % (bid, br_rec[0], br_rec[1], br_rec[2]))
def calc_node_heights(self):
"""Calculate node heights on the reference tree (used to define branch-length cutoff during classification step)
Algorithm is as follows:
Tip node or node resolved to Species level: height = 1
Inner node resolved to Genus or above: height = min(left_height, right_height) + 1
"""
nh_map = {}
dummy_added = False
for node in self.reftree_tax.traverse("postorder"):
if not node.is_root():
if not hasattr(node, "B"):
# In a rooted tree, there is always one more node/branch than in unrooted one
# That's why one branch will be always not EPA-labelled after the rooting
if not dummy_added:
node.B = "DDD"
dummy_added = True
species_rank = Taxonomy.EMPTY_RANK
else:
errmsg = "FATAL ERROR: More than one tree branch without EPA label (calc_node_heights)"
self.cfg.exit_fatal_error(errmsg)
else:
species_rank = self.bid_ranks_map[node.B][-1]
bid = node.B
if node.is_leaf() or species_rank != Taxonomy.EMPTY_RANK:
nh_map[bid] = 1
else:
lchild = node.children[0]
rchild = node.children[1]
nh_map[bid] = min(nh_map[lchild.B], nh_map[rchild.B]) + 1
# remove heights for dummy nodes, since there won't be any placements on them
if dummy_added:
del nh_map["DDD"]
self.node_height_map = nh_map
def __get_all_rank_names(self, root):
rnames = set([])
for node in root.traverse("postorder"):
ranks = node.ranks
for rk in ranks:
rnames.add(rk)
return rnames
def mono_index(self):
"""This method will calculate monophyly index by looking at the left and right hand side of the tree"""
children = self.reftree_tax.children
if len(children) == 1:
while len(children) == 1:
children = children[0].children
if len(children) == 2:
left = children[0]
right =children[1]
lset = self.__get_all_rank_names(left)
rset = self.__get_all_rank_names(right)
iset = lset & rset
return iset
else:
print("Error: input tree not birfurcating")
return set([])
def build_hmm_profile(self, json_builder):
print "Building the HMMER profile...\n"
# this stupid workaround is needed because RAxML outputs the reduced
# alignment in relaxed PHYLIP format, which is not supported by HMMER
refalign_fasta = self.cfg.tmp_fname("%NAME%_ref_reduced.fa")
self.reduced_refalign_seqs.write(outfile=refalign_fasta)
hmm = hmmer(self.cfg, refalign_fasta)
fprofile = hmm.build_hmm_profile()
json_builder.set_hmm_profile(fprofile)
def write_json(self):
jw = RefJsonBuilder()
jw.set_branch_tax_map(self.bid_ranks_map)
jw.set_tree(self.reftree_lbl_str)
jw.set_outgroup(self.reftree_outgroup)
jw.set_ratehet_model(self.cfg.raxml_model)
jw.set_tax_tree(self.reftree_multif)
jw.set_pattern_compression(self.cfg.compress_patterns)
jw.set_taxcode(self.cfg.taxcode_name)
jw.set_merged_ranks_map(self.input_validator.merged_ranks)
corr_ranks_reverse = dict((reversed(item) for item in self.input_validator.corr_ranks.items()))
jw.set_corr_ranks_map(corr_ranks_reverse)
corr_seqid_reverse = dict((reversed(item) for item in self.input_validator.corr_seqid.items()))
jw.set_corr_seqid_map(corr_seqid_reverse)
mdata = { "ref_tree_size": self.reftree_size,
"ref_alignment_width": self.refalign_width,
"raxml_version": self.raxml_version,
"timestamp": str(datetime.datetime.now()),
"invocation_epac": self.invocation_epac,
"invocation_raxml_multif": self.invocation_raxml_multif,
"invocation_raxml_optmod": self.invocation_raxml_optmod,
"invocation_raxml_epalbl": self.invocation_raxml_epalbl,
"reftree_loglh": self.reftree_loglh
}
jw.set_metadata(mdata)
seqs = self.reduced_refalign_seqs.get_entries()
jw.set_sequences(seqs)
if not self.cfg.no_hmmer:
self.build_hmm_profile(jw)
orig_tax = self.taxonomy_map
jw.set_origin_taxonomy(orig_tax)
self.cfg.log.debug("Calculating the speciation rate...\n")
tp = tree_param(tree = self.reftree_lbl_str, origin_taxonomy = orig_tax)
jw.set_rate(tp.get_speciation_rate_fast())
jw.set_nodes_height(self.node_height_map)
jw.set_binary_model(self.optmod_fname)
self.cfg.log.debug("Writing down the reference file...\n")
jw.dump(self.cfg.refjson_fname)
# top-level function to build a reference tree
def build_ref_tree(self):
self.cfg.log.info("=> Loading taxonomy from file: %s ...\n" , self.cfg.taxonomy_fname)
self.taxonomy = Taxonomy(prefix=EpacConfig.REF_SEQ_PREFIX, tax_fname=self.cfg.taxonomy_fname)
self.cfg.log.info("==> Loading reference alignment from file: %s ...\n" , self.cfg.align_fname)
self.load_alignment()
self.cfg.log.info("===> Validating taxonomy and alignment ...\n")
self.validate_taxonomy()
self.cfg.log.info("====> Building a multifurcating tree from taxonomy with %d seqs ...\n" , self.taxonomy.seq_count())
self.build_multif_tree()
self.cfg.log.info("=====> Building the reference alignment ...\n")
self.export_ref_alignment()
self.export_ref_taxonomy()
self.cfg.log.info("======> Saving the outgroup for later re-rooting ...\n")
self.save_rooting()
self.cfg.log.info("=======> Resolving multifurcation: choosing the best topology from %d independent RAxML runs ...\n" % self.cfg.rep_num)
self.resolve_multif()
self.load_reduced_refalign()
self.cfg.log.info("========> Calling RAxML-EPA to obtain branch labels ...\n")
self.epa_branch_labeling()
self.cfg.log.info("=========> Post-processing the EPA tree (re-rooting, taxonomic labeling etc.) ...\n")
self.epa_post_process()
self.calc_node_heights()
self.cfg.log.debug("\n==========> Checking branch labels ...")
self.cfg.log.debug("shared rank names before training: %s", repr(self.taxonomy.get_common_ranks()))
self.cfg.log.debug("shared rank names after training: %s\n", repr(self.mono_index()))
self.cfg.log.info("==========> Saving the reference JSON file: %s\n" % self.cfg.refjson_fname)
self.write_json()
class RefTreeBuilder:
def __init__(self, config):
self.cfg = config
self.mfresolv_job_name = self.cfg.subst_name("mfresolv_%NAME%")
self.epalbl_job_name = self.cfg.subst_name("epalbl_%NAME%")
self.optmod_job_name = self.cfg.subst_name("optmod_%NAME%")
self.raxml_wrapper = RaxmlWrapper(config)
self.outgr_fname = self.cfg.tmp_fname("%NAME%_outgr.tre")
self.reftree_mfu_fname = self.cfg.tmp_fname("%NAME%_mfu.tre")
self.reftree_bfu_fname = self.cfg.tmp_fname("%NAME%_bfu.tre")
self.optmod_fname = self.cfg.tmp_fname("%NAME%.opt")
self.lblalign_fname = self.cfg.tmp_fname("%NAME%_lblq.fa")
self.reftree_lbl_fname = self.cfg.tmp_fname("%NAME%_lbl.tre")
self.reftree_tax_fname = self.cfg.tmp_fname("%NAME%_tax.tre")
self.brmap_fname = self.cfg.tmp_fname("%NAME%_map.txt")
def load_alignment(self):
in_file = self.cfg.align_fname
self.input_seqs = None
formats = [
"fasta", "phylip_relaxed", "iphylip_relaxed", "phylip", "iphylip"
]
for fmt in formats:
try:
self.input_seqs = SeqGroup(sequences=in_file, format=fmt)
break
except:
self.cfg.log.debug("Guessing input format: not " + fmt)
if self.input_seqs == None:
self.cfg.exit_user_error(
"Invalid input file format: %s\nThe supported input formats are fasta and phylip"
% in_file)
def validate_taxonomy(self):
self.input_validator = InputValidator(self.cfg, self.taxonomy,
self.input_seqs)
self.input_validator.validate()
def build_multif_tree(self):
c = self.cfg
tb = TaxTreeBuilder(c, self.taxonomy)
(t, ids) = tb.build(c.reftree_min_rank, c.reftree_max_seqs_per_leaf,
c.reftree_clades_to_include,
c.reftree_clades_to_ignore)
self.reftree_ids = frozenset(ids)
self.reftree_size = len(ids)
self.reftree_multif = t
# IMPORTANT: select GAMMA or CAT model based on tree size!
self.cfg.resolve_auto_settings(self.reftree_size)
if self.cfg.debug:
refseq_fname = self.cfg.tmp_fname("%NAME%_seq_ids.txt")
# list of sequence ids which comprise the reference tree
with open(refseq_fname, "w") as f:
for sid in ids:
f.write("%s\n" % sid)
# original tree with taxonomic ranks as internal node labels
reftax_fname = self.cfg.tmp_fname("%NAME%_mfu_tax.tre")
t.write(outfile=reftax_fname, format=8)
# t.show()
def export_ref_alignment(self):
"""This function transforms the input alignment in the following way:
1. Filter out sequences which are not part of the reference tree
2. Add sequence name prefix (r_)"""
self.refalign_fname = self.cfg.tmp_fname("%NAME%_matrix.afa")
with open(self.refalign_fname, "w") as fout:
for name, seq, comment, sid in self.input_seqs.iter_entries():
seq_name = EpacConfig.REF_SEQ_PREFIX + name
if seq_name in self.input_validator.corr_seqid:
seq_name = self.input_validator.corr_seqid[seq_name]
if seq_name in self.reftree_ids:
fout.write(">" + seq_name + "\n" + seq + "\n")
# we do not need the original alignment anymore, so free its memory
self.input_seqs = None
def export_ref_taxonomy(self):
self.taxonomy_map = {}
for sid, ranks in self.taxonomy.iteritems():
if sid in self.reftree_ids:
self.taxonomy_map[sid] = ranks
if self.cfg.debug:
tax_fname = self.cfg.tmp_fname("%NAME%_tax.txt")
with open(tax_fname, "w") as fout:
for sid, ranks in self.taxonomy_map.iteritems():
ranks_str = self.taxonomy.seq_lineage_str(sid)
fout.write(sid + "\t" + ranks_str + "\n")
def save_rooting(self):
rt = self.reftree_multif
tax_map = self.taxonomy.get_map()
self.taxtree_helper = TaxTreeHelper(self.cfg, tax_map)
self.taxtree_helper.set_mf_rooted_tree(rt)
outgr = self.taxtree_helper.get_outgroup()
outgr_size = len(outgr.get_leaves())
outgr.write(outfile=self.outgr_fname, format=9)
self.reftree_outgroup = outgr
self.cfg.log.debug(
"Outgroup for rooting was saved to: %s, outgroup size: %d",
self.outgr_fname, outgr_size)
# remove unifurcation at the root
if len(rt.children) == 1:
rt = rt.children[0]
# now we can safely unroot the tree and remove internal node labels to make it suitable for raxml
rt.write(outfile=self.reftree_mfu_fname, format=9)
# RAxML call to convert multifurcating tree to the strictly bifurcating one
def resolve_multif(self):
self.cfg.log.debug("\nReducing the alignment: \n")
self.reduced_refalign_fname = self.raxml_wrapper.reduce_alignment(
self.refalign_fname)
self.cfg.log.debug("\nConstrained ML inference: \n")
raxml_params = [
"-s", self.reduced_refalign_fname, "-g", self.reftree_mfu_fname,
"--no-seq-check", "-N",
str(self.cfg.rep_num)
]
if self.cfg.mfresolv_method == "fast":
raxml_params += ["-D"]
elif self.cfg.mfresolv_method == "ultrafast":
raxml_params += ["-f", "e"]
if self.cfg.restart and self.raxml_wrapper.result_exists(
self.mfresolv_job_name):
self.invocation_raxml_multif = self.raxml_wrapper.get_invocation_str(
self.mfresolv_job_name)
self.cfg.log.debug(
"\nUsing existing ML tree found in: %s\n",
self.raxml_wrapper.result_fname(self.mfresolv_job_name))
else:
self.invocation_raxml_multif = self.raxml_wrapper.run(
self.mfresolv_job_name, raxml_params)
# self.invocation_raxml_multif = self.raxml_wrapper.run_multiple(self.mfresolv_job_name, raxml_params, self.cfg.rep_num)
if self.cfg.mfresolv_method == "ultrafast":
self.raxml_wrapper.copy_result_tree(
self.mfresolv_job_name,
self.raxml_wrapper.besttree_fname(self.mfresolv_job_name))
if self.raxml_wrapper.besttree_exists(self.mfresolv_job_name):
if not self.cfg.reopt_model:
self.raxml_wrapper.copy_best_tree(self.mfresolv_job_name,
self.reftree_bfu_fname)
self.raxml_wrapper.copy_optmod_params(self.mfresolv_job_name,
self.optmod_fname)
self.invocation_raxml_optmod = ""
job_name = self.mfresolv_job_name
else:
bfu_fname = self.raxml_wrapper.besttree_fname(
self.mfresolv_job_name)
job_name = self.optmod_job_name
# RAxML call to optimize model parameters and write them down to the binary model file
self.cfg.log.debug("\nOptimizing model parameters: \n")
raxml_params = [
"-f", "e", "-s", self.reduced_refalign_fname, "-t",
bfu_fname, "--no-seq-check"
]
if self.cfg.raxml_model.startswith(
"GTRCAT") and not self.cfg.compress_patterns:
raxml_params += ["-H"]
if self.cfg.restart and self.raxml_wrapper.result_exists(
self.optmod_job_name):
self.invocation_raxml_optmod = self.raxml_wrapper.get_invocation_str(
self.optmod_job_name)
self.cfg.log.debug(
"\nUsing existing optimized tree and parameters found in: %s\n",
self.raxml_wrapper.result_fname(self.optmod_job_name))
else:
self.invocation_raxml_optmod = self.raxml_wrapper.run(
self.optmod_job_name, raxml_params)
if self.raxml_wrapper.result_exists(self.optmod_job_name):
self.raxml_wrapper.copy_result_tree(
self.optmod_job_name, self.reftree_bfu_fname)
self.raxml_wrapper.copy_optmod_params(
self.optmod_job_name, self.optmod_fname)
else:
errmsg = "RAxML run failed (model optimization), please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.optmod_job_name)
self.cfg.exit_fatal_error(errmsg)
if self.cfg.raxml_model.startswith("GTRCAT"):
mod_name = "CAT"
else:
mod_name = "GAMMA"
self.reftree_loglh = self.raxml_wrapper.get_tree_lh(
job_name, mod_name)
self.cfg.log.debug("\n%s-based logLH of the reference tree: %f\n" %
(mod_name, self.reftree_loglh))
else:
errmsg = "RAxML run failed (mutlifurcation resolution), please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.mfresolv_job_name)
self.cfg.exit_fatal_error(errmsg)
def load_reduced_refalign(self):
formats = ["fasta", "phylip_relaxed"]
for fmt in formats:
try:
self.reduced_refalign_seqs = SeqGroup(
sequences=self.reduced_refalign_fname, format=fmt)
break
except:
pass
if self.reduced_refalign_seqs == None:
errmsg = "FATAL ERROR: Invalid input file format in %s! (load_reduced_refalign)" % self.reduced_refalign_fname
self.cfg.exit_fatal_error(errmsg)
# dummy EPA run to label the branches of the reference tree, which we need to build a mapping to tax ranks
def epa_branch_labeling(self):
# create alignment with dummy query seq
self.refalign_width = len(self.reduced_refalign_seqs.get_seqbyid(0))
self.reduced_refalign_seqs.write(format="fasta",
outfile=self.lblalign_fname)
with open(self.lblalign_fname, "a") as fout:
fout.write(">" + "DUMMY131313" + "\n")
fout.write("A" * self.refalign_width + "\n")
# TODO always load model regardless of the config file settings?
epa_result = self.raxml_wrapper.run_epa(self.epalbl_job_name,
self.lblalign_fname,
self.reftree_bfu_fname,
self.optmod_fname,
mode="epa_mp")
self.reftree_lbl_str = epa_result.get_std_newick_tree()
self.raxml_version = epa_result.get_raxml_version()
self.invocation_raxml_epalbl = epa_result.get_raxml_invocation()
if not self.raxml_wrapper.epa_result_exists(self.epalbl_job_name):
errmsg = "RAxML EPA run failed, please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.epalbl_job_name)
self.cfg.exit_fatal_error(errmsg)
def epa_post_process(self):
lbl_tree = Tree(self.reftree_lbl_str)
self.taxtree_helper.set_bf_unrooted_tree(lbl_tree)
self.reftree_tax = self.taxtree_helper.get_tax_tree()
self.bid_ranks_map = self.taxtree_helper.get_bid_taxonomy_map()
if self.cfg.debug:
self.reftree_tax.write(outfile=self.reftree_tax_fname, format=3)
with open(self.reftree_lbl_fname, "w") as outf:
outf.write(self.reftree_lbl_str)
with open(self.brmap_fname, "w") as outf:
for bid, br_rec in self.bid_ranks_map.iteritems():
outf.write("%s\t%s\t%d\t%f\n" %
(bid, br_rec[0], br_rec[1], br_rec[2]))
def calc_node_heights(self):
"""Calculate node heights on the reference tree (used to define branch-length cutoff during classification step)
Algorithm is as follows:
Tip node or node resolved to Species level: height = 1
Inner node resolved to Genus or above: height = min(left_height, right_height) + 1
"""
nh_map = {}
dummy_added = False
for node in self.reftree_tax.traverse("postorder"):
if not node.is_root():
if not hasattr(node, "B"):
# In a rooted tree, there is always one more node/branch than in unrooted one
# That's why one branch will be always not EPA-labelled after the rooting
if not dummy_added:
node.B = "DDD"
dummy_added = True
species_rank = Taxonomy.EMPTY_RANK
else:
errmsg = "FATAL ERROR: More than one tree branch without EPA label (calc_node_heights)"
self.cfg.exit_fatal_error(errmsg)
else:
species_rank = self.bid_ranks_map[node.B][-1]
bid = node.B
if node.is_leaf() or species_rank != Taxonomy.EMPTY_RANK:
nh_map[bid] = 1
else:
lchild = node.children[0]
rchild = node.children[1]
nh_map[bid] = min(nh_map[lchild.B], nh_map[rchild.B]) + 1
# remove heights for dummy nodes, since there won't be any placements on them
if dummy_added:
del nh_map["DDD"]
self.node_height_map = nh_map
def __get_all_rank_names(self, root):
rnames = set([])
for node in root.traverse("postorder"):
ranks = node.ranks
for rk in ranks:
rnames.add(rk)
return rnames
def mono_index(self):
"""This method will calculate monophyly index by looking at the left and right hand side of the tree"""
children = self.reftree_tax.children
if len(children) == 1:
while len(children) == 1:
children = children[0].children
if len(children) == 2:
left = children[0]
right = children[1]
lset = self.__get_all_rank_names(left)
rset = self.__get_all_rank_names(right)
iset = lset & rset
return iset
else:
print("Error: input tree not birfurcating")
return set([])
def build_hmm_profile(self, json_builder):
print "Building the HMMER profile...\n"
# this stupid workaround is needed because RAxML outputs the reduced
# alignment in relaxed PHYLIP format, which is not supported by HMMER
refalign_fasta = self.cfg.tmp_fname("%NAME%_ref_reduced.fa")
self.reduced_refalign_seqs.write(outfile=refalign_fasta)
hmm = hmmer(self.cfg, refalign_fasta)
fprofile = hmm.build_hmm_profile()
json_builder.set_hmm_profile(fprofile)
def write_json(self):
jw = RefJsonBuilder()
jw.set_branch_tax_map(self.bid_ranks_map)
jw.set_tree(self.reftree_lbl_str)
jw.set_outgroup(self.reftree_outgroup)
jw.set_ratehet_model(self.cfg.raxml_model)
jw.set_tax_tree(self.reftree_multif)
jw.set_pattern_compression(self.cfg.compress_patterns)
jw.set_taxcode(self.cfg.taxcode_name)
jw.set_merged_ranks_map(self.input_validator.merged_ranks)
corr_ranks_reverse = dict(
(reversed(item)
for item in self.input_validator.corr_ranks.items()))
jw.set_corr_ranks_map(corr_ranks_reverse)
corr_seqid_reverse = dict(
(reversed(item)
for item in self.input_validator.corr_seqid.items()))
jw.set_corr_seqid_map(corr_seqid_reverse)
mdata = {
"ref_tree_size": self.reftree_size,
"ref_alignment_width": self.refalign_width,
"raxml_version": self.raxml_version,
"timestamp": str(datetime.datetime.now()),
"invocation_epac": self.invocation_epac,
"invocation_raxml_multif": self.invocation_raxml_multif,
"invocation_raxml_optmod": self.invocation_raxml_optmod,
"invocation_raxml_epalbl": self.invocation_raxml_epalbl,
"reftree_loglh": self.reftree_loglh
}
jw.set_metadata(mdata)
seqs = self.reduced_refalign_seqs.get_entries()
jw.set_sequences(seqs)
if not self.cfg.no_hmmer:
self.build_hmm_profile(jw)
orig_tax = self.taxonomy_map
jw.set_origin_taxonomy(orig_tax)
self.cfg.log.debug("Calculating the speciation rate...\n")
tp = tree_param(tree=self.reftree_lbl_str, origin_taxonomy=orig_tax)
jw.set_rate(tp.get_speciation_rate_fast())
jw.set_nodes_height(self.node_height_map)
jw.set_binary_model(self.optmod_fname)
self.cfg.log.debug("Writing down the reference file...\n")
jw.dump(self.cfg.refjson_fname)
# top-level function to build a reference tree
def build_ref_tree(self):
self.cfg.log.info("=> Loading taxonomy from file: %s ...\n",
self.cfg.taxonomy_fname)
self.taxonomy = Taxonomy(prefix=EpacConfig.REF_SEQ_PREFIX,
tax_fname=self.cfg.taxonomy_fname)
self.cfg.log.info(
"==> Loading reference alignment from file: %s ...\n",
self.cfg.align_fname)
self.load_alignment()
self.cfg.log.info("===> Validating taxonomy and alignment ...\n")
self.validate_taxonomy()
self.cfg.log.info(
"====> Building a multifurcating tree from taxonomy with %d seqs ...\n",
self.taxonomy.seq_count())
self.build_multif_tree()
self.cfg.log.info("=====> Building the reference alignment ...\n")
self.export_ref_alignment()
self.export_ref_taxonomy()
self.cfg.log.info(
"======> Saving the outgroup for later re-rooting ...\n")
self.save_rooting()
self.cfg.log.info(
"=======> Resolving multifurcation: choosing the best topology from %d independent RAxML runs ...\n"
% self.cfg.rep_num)
self.resolve_multif()
self.load_reduced_refalign()
self.cfg.log.info(
"========> Calling RAxML-EPA to obtain branch labels ...\n")
self.epa_branch_labeling()
self.cfg.log.info(
"=========> Post-processing the EPA tree (re-rooting, taxonomic labeling etc.) ...\n"
)
self.epa_post_process()
self.calc_node_heights()
self.cfg.log.debug("\n==========> Checking branch labels ...")
self.cfg.log.debug("shared rank names before training: %s",
repr(self.taxonomy.get_common_ranks()))
self.cfg.log.debug("shared rank names after training: %s\n",
repr(self.mono_index()))
self.cfg.log.info("==========> Saving the reference JSON file: %s\n" %
self.cfg.refjson_fname)
self.write_json()
class RefTreeBuilder:
def __init__(self, config):
self.cfg = config
self.mfresolv_job_name = self.cfg.subst_name("mfresolv_%NAME%")
self.epalbl_job_name = self.cfg.subst_name("epalbl_%NAME%")
self.optmod_job_name = self.cfg.subst_name("optmod_%NAME%")
self.raxml_wrapper = RaxmlWrapper(config)
self.outgr_fname = self.cfg.tmp_fname("%NAME%_outgr.tre")
self.reftree_mfu_fname = self.cfg.tmp_fname("%NAME%_mfu.tre")
self.reftree_bfu_fname = self.cfg.tmp_fname("%NAME%_bfu.tre")
self.optmod_fname = self.cfg.tmp_fname("%NAME%.opt")
self.lblalign_fname = self.cfg.tmp_fname("%NAME%_lblq.fa")
self.reftree_lbl_fname = self.cfg.tmp_fname("%NAME%_lbl.tre")
self.reftree_tax_fname = self.cfg.tmp_fname("%NAME%_tax.tre")
self.brmap_fname = self.cfg.tmp_fname("%NAME%_map.txt")
def validate_taxonomy(self):
# make sure we don't taxonomy "irregularities" (more than 7 ranks or missing ranks in the middle)
action = self.cfg.wrong_rank_count
if action != "ignore":
autofix = action == "autofix"
errs = self.taxonomy.check_for_disbalance(autofix)
if len(errs) > 0:
if action == "autofix":
print "WARNING: %d sequences with invalid annotation (missing/redundant ranks) found and were fixed as follows:\n" % len(errs)
for err in errs:
print "Original: %s\t%s" % (err[0], err[1])
print "Fixed as: %s\t%s\n" % (err[0], err[2])
elif action == "skip":
print "WARNING: Following %d sequences with invalid annotation (missing/redundant ranks) were skipped:\n" % len(errs)
for err in errs:
self.taxonomy.remove_seq(err[0])
print "%s\t%s" % err
else: # abort
print "ERROR: %d sequences with invalid annotation (missing/redundant ranks) found:\n" % len(errs)
for err in errs:
print "%s\t%s" % err
print "\nPlease fix them manually (add/remove ranks) and run the pipeline again (or use -wrong-rank-count autofix option)"
print "NOTE: Only standard 7-level taxonomies are supported at the moment. Although missing trailing ranks (e.g. species) are allowed,"
print "missing intermediate ranks (e.g. family) or sublevels (e.g. suborder) are not!\n"
sys.exit()
# check for duplicate rank names
action = self.cfg.dup_rank_names
if action != "ignore":
autofix = action == "autofix"
dups = self.taxonomy.check_for_duplicates(autofix)
if len(dups) > 0:
if action == "autofix":
print "WARNING: %d sequences with duplicate rank names found and were renamed as follows:\n" % len(dups)
for dup in dups:
print "Original: %s\t%s" % (dup[0], dup[1])
print "Duplicate: %s\t%s" % (dup[2], dup[3])
print "Renamed to: %s\t%s\n" % (dup[2], dup[4])
elif action == "skip":
print "WARNING: Following %d sequences with duplicate rank names were skipped:\n" % len(dups)
for dup in dups:
self.taxonomy.remove_seq(dup[2])
print "%s\t%s\n" % (dup[2], dup[3])
else: # abort
print "ERROR: %d sequences with duplicate rank names found:\n" % len(dups)
for dup in dups:
print "%s\t%s\n%s\t%s\n" % dup
print "Please fix (rename) them and run the pipeline again (or use -dup-rank-names autofix option)"
sys.exit()
# check for invalid characters in rank names
self.taxonomy.normalize_rank_names()
self.taxonomy.close_taxonomy_gaps()
def build_multif_tree(self):
c = self.cfg
tb = TaxTreeBuilder(c, self.taxonomy)
(t, ids) = tb.build(c.reftree_min_rank, c.reftree_max_seqs_per_leaf, c.reftree_clades_to_include, c.reftree_clades_to_ignore)
self.reftree_ids = frozenset(ids)
self.reftree_size = len(ids)
self.reftree_multif = t
# IMPORTANT: select GAMMA or CAT model based on tree size!
self.cfg.resolve_auto_settings(self.reftree_size)
if self.cfg.debug:
refseq_fname = self.cfg.tmp_fname("%NAME%_seq_ids.txt")
# list of sequence ids which comprise the reference tree
with open(refseq_fname, "w") as f:
for sid in ids:
f.write("%s\n" % sid)
# original tree with taxonomic ranks as internal node labels
reftax_fname = self.cfg.tmp_fname("%NAME%_mfu_tax.tre")
t.write(outfile=reftax_fname, format=8)
# t.show()
def export_ref_alignment(self):
"""This function transforms the input alignment in the following way:
1. Filter out sequences which are not part of the reference tree
2. Add sequence name prefix (r_)"""
in_file = self.cfg.align_fname
ref_seqs = None
formats = ["fasta", "phylip", "iphylip", "phylip_relaxed", "iphylip_relaxed"]
for fmt in formats:
try:
ref_seqs = SeqGroup(sequences=in_file, format = fmt)
break
except:
if self.cfg.debug:
print("Guessing input format: not " + fmt)
if ref_seqs == None:
print("Invalid input file format: %s" % in_file)
print("The supported input formats are fasta and phylip")
sys.exit()
self.refalign_fname = self.cfg.tmp_fname("%NAME%_matrix.afa")
with open(self.refalign_fname, "w") as fout:
for name, seq, comment, sid in ref_seqs.iter_entries():
seq_name = EpacConfig.REF_SEQ_PREFIX + name
if seq_name in self.reftree_ids:
fout.write(">" + seq_name + "\n" + seq + "\n")
def export_ref_taxonomy(self):
self.taxonomy_map = {}
for sid, ranks in self.taxonomy.iteritems():
if sid in self.reftree_ids:
self.taxonomy_map[sid] = ranks
if self.cfg.debug:
tax_fname = self.cfg.tmp_fname("%NAME%_tax.txt")
with open(tax_fname, "w") as fout:
for sid, ranks in self.taxonomy_map.iteritems():
ranks_str = self.taxonomy.lineage_str(sid, True)
fout.write(sid + "\t" + ranks_str + "\n")
def save_rooting(self):
rt = self.reftree_multif
tax_map = self.taxonomy.get_map()
self.taxtree_helper = TaxTreeHelper(tax_map, self.cfg)
self.taxtree_helper.set_mf_rooted_tree(rt)
outgr = self.taxtree_helper.get_outgroup()
outgr_size = len(outgr.get_leaves())
outgr.write(outfile=self.outgr_fname, format=9)
self.reftree_outgroup = outgr
if self.cfg.verbose:
print "Outgroup for rooting was saved to: %s, outgroup size: %d" % (self.outgr_fname, outgr_size)
# remove unifurcation at the root
if len(rt.children) == 1:
rt = rt.children[0]
# now we can safely unroot the tree and remove internal node labels to make it suitable for raxml
rt.write(outfile=self.reftree_mfu_fname, format=9)
# RAxML call to convert multifurcating tree to the strictly bifurcating one
def resolve_multif(self):
print "\nReducing the alignment: \n"
self.reduced_refalign_fname = self.raxml_wrapper.reduce_alignment(self.refalign_fname)
print "\nResolving multifurcation: \n"
raxml_params = ["-s", self.reduced_refalign_fname, "-g", self.reftree_mfu_fname, "-F", "--no-seq-check"]
if self.cfg.mfresolv_method == "fast":
raxml_params += ["-D"]
elif self.cfg.mfresolv_method == "ultrafast":
raxml_params += ["-f", "e"]
self.invocation_raxml_multif = self.raxml_wrapper.run(self.mfresolv_job_name, raxml_params)
if self.raxml_wrapper.result_exists(self.mfresolv_job_name):
# self.raxml_wrapper.copy_result_tree(self.mfresolv_job_name, self.reftree_bfu_fname)
# self.raxml_wrapper.copy_optmod_params(self.mfresolv_job_name, self.optmod_fname)
bfu_fname = self.raxml_wrapper.result_fname(self.mfresolv_job_name)
# RAxML call to optimize model parameters and write them down to the binary model file
print "\nOptimizing model parameters: \n"
raxml_params = ["-f", "e", "-s", self.reduced_refalign_fname, "-t", bfu_fname, "--no-seq-check"]
if self.cfg.raxml_model == "GTRCAT" and not self.cfg.compress_patterns:
raxml_params += ["-H"]
self.invocation_raxml_optmod = self.raxml_wrapper.run(self.optmod_job_name, raxml_params)
if self.raxml_wrapper.result_exists(self.optmod_job_name):
self.raxml_wrapper.copy_result_tree(self.optmod_job_name, self.reftree_bfu_fname)
self.raxml_wrapper.copy_optmod_params(self.optmod_job_name, self.optmod_fname)
if not self.cfg.debug:
self.raxml_wrapper.cleanup(self.optmod_job_name)
else:
print "RAxML run failed (model optimization), please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.optmod_job_name)
sys.exit()
if not self.cfg.debug:
self.raxml_wrapper.cleanup(self.mfresolv_job_name)
else:
print "RAxML run failed (mutlifurcation resolution), please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.mfresolv_job_name)
sys.exit()
def load_reduced_refalign(self):
formats = ["fasta", "phylip_relaxed"]
for fmt in formats:
try:
self.reduced_refalign_seqs = SeqGroup(sequences=self.reduced_refalign_fname, format = fmt)
break
except:
pass
if self.reduced_refalign_seqs == None:
print("FATAL ERROR: Invalid input file format in %s! (load_reduced_refalign)" % self.reduced_refalign_fname)
sys.exit()
# dummy EPA run to label the branches of the reference tree, which we need to build a mapping to tax ranks
def epa_branch_labeling(self):
# create alignment with dummy query seq
self.refalign_width = len(self.reduced_refalign_seqs.get_seqbyid(0))
self.reduced_refalign_seqs.write(format="fasta", outfile=self.lblalign_fname)
with open(self.lblalign_fname, "a") as fout:
fout.write(">" + "DUMMY131313" + "\n")
fout.write("A"*self.refalign_width + "\n")
epa_result = self.raxml_wrapper.run_epa(self.epalbl_job_name, self.lblalign_fname, self.reftree_bfu_fname, self.optmod_fname)
self.reftree_lbl_str = epa_result.get_std_newick_tree()
self.raxml_version = epa_result.get_raxml_version()
self.invocation_raxml_epalbl = epa_result.get_raxml_invocation()
if self.raxml_wrapper.epa_result_exists(self.epalbl_job_name):
if not self.cfg.debug:
self.raxml_wrapper.cleanup(self.epalbl_job_name, True)
else:
print "RAxML EPA run failed, please examine the log for details: %s" \
% self.raxml_wrapper.make_raxml_fname("output", self.epalbl_job_name)
sys.exit()
def epa_post_process(self):
lbl_tree = Tree(self.reftree_lbl_str)
self.taxtree_helper.set_bf_unrooted_tree(lbl_tree)
self.reftree_tax = self.taxtree_helper.get_tax_tree()
self.bid_ranks_map = self.taxtree_helper.get_bid_taxonomy_map()
if self.cfg.debug:
self.reftree_tax.write(outfile=self.reftree_lbl_fname, format=5)
self.reftree_tax.write(outfile=self.reftree_tax_fname, format=3)
def build_branch_rank_map(self):
self.bid_ranks_map = {}
for node in self.reftree_tax.traverse("postorder"):
if not node.is_root() and hasattr(node, "B"):
parent = node.up
self.bid_ranks_map[node.B] = parent.ranks
# print "%s => %s" % (node.B, parent.ranks)
elif self.cfg.verbose:
print "INFO: EPA branch label missing, mapping to taxon skipped (%s)" % node.name
def write_branch_rank_map(self):
with open(self.brmap_fname, "w") as fbrmap:
for node in self.reftree_tax.traverse("postorder"):
if not node.is_root() and hasattr(node, "B"):
fbrmap.write(node.B + "\t" + ";".join(self.bid_ranks_map[node.B]) + "\n")
def calc_node_heights(self):
"""Calculate node heights on the reference tree (used to define branch-length cutoff during classification step)
Algorithm is as follows:
Tip node or node resolved to Species level: height = 1
Inner node resolved to Genus or above: height = min(left_height, right_height) + 1
"""
nh_map = {}
dummy_added = False
for node in self.reftree_tax.traverse("postorder"):
if not node.is_root():
if not hasattr(node, "B"):
# In a rooted tree, there is always one more node/branch than in unrooted one
# That's why one branch will be always not EPA-labelled after the rooting
if not dummy_added:
node.B = "DDD"
dummy_added = True
species_rank = Taxonomy.EMPTY_RANK
else:
print "FATAL ERROR: More than one tree branch without EPA label (calc_node_heights)"
sys.exit()
else:
species_rank = self.bid_ranks_map[node.B][6]
bid = node.B
if node.is_leaf() or species_rank != Taxonomy.EMPTY_RANK:
nh_map[bid] = 1
else:
lchild = node.children[0]
rchild = node.children[1]
nh_map[bid] = min(nh_map[lchild.B], nh_map[rchild.B]) + 1
# remove heights for dummy nodes, since there won't be any placements on them
if dummy_added:
del nh_map["DDD"]
self.node_height_map = nh_map
def __get_all_rank_names(self, root):
rnames = set([])
for node in root.traverse("postorder"):
ranks = node.ranks
for rk in ranks:
rnames.add(rk)
return rnames
def mono_index(self):
"""This method will calculate monophyly index by looking at the left and right hand side of the tree"""
children = self.reftree_tax.children
if len(children) == 1:
while len(children) == 1:
children = children[0].children
if len(children) == 2:
left = children[0]
right =children[1]
lset = self.__get_all_rank_names(left)
rset = self.__get_all_rank_names(right)
iset = lset & rset
return iset
else:
print("Error: input tree not birfurcating")
return set([])
def build_hmm_profile(self, json_builder):
print "Building the HMMER profile...\n"
# this stupid workaround is needed because RAxML outputs the reduced
# alignment in relaxed PHYLIP format, which is not supported by HMMER
refalign_fasta = self.cfg.tmp_fname("%NAME%_ref_reduced.fa")
self.reduced_refalign_seqs.write(outfile=refalign_fasta)
hmm = hmmer(self.cfg, refalign_fasta)
fprofile = hmm.build_hmm_profile()
json_builder.set_hmm_profile(fprofile)
if not self.cfg.debug:
FileUtils.remove_if_exists(refalign_fasta)
FileUtils.remove_if_exists(fprofile)
def write_json(self):
jw = RefJsonBuilder()
jw.set_taxonomy(self.bid_ranks_map)
jw.set_tree(self.reftree_lbl_str)
jw.set_outgroup(self.reftree_outgroup)
jw.set_ratehet_model(self.cfg.raxml_model)
jw.set_tax_tree(self.reftree_multif)
jw.set_pattern_compression(self.cfg.compress_patterns)
mdata = { "ref_tree_size": self.reftree_size,
"ref_alignment_width": self.refalign_width,
"raxml_version": self.raxml_version,
"timestamp": str(datetime.datetime.now()),
"invocation_epac": self.invocation_epac,
"invocation_raxml_multif": self.invocation_raxml_multif,
"invocation_raxml_optmod": self.invocation_raxml_optmod,
"invocation_raxml_epalbl": self.invocation_raxml_epalbl
}
jw.set_metadata(mdata)
seqs = self.reduced_refalign_seqs.get_entries()
jw.set_sequences(seqs)
if not self.cfg.no_hmmer:
self.build_hmm_profile(jw)
orig_tax = self.taxonomy_map
jw.set_origin_taxonomy(orig_tax)
print "Calculating the speciation rate...\n"
tp = tree_param(tree = self.reftree_lbl_str, origin_taxonomy = orig_tax)
jw.set_rate(tp.get_speciation_rate_fast())
jw.set_nodes_height(self.node_height_map)
jw.set_binary_model(self.optmod_fname)
print "Writing down the reference file...\n"
jw.dump(self.cfg.refjson_fname)
def cleanup(self):
FileUtils.remove_if_exists(self.outgr_fname)
FileUtils.remove_if_exists(self.reftree_mfu_fname)
FileUtils.remove_if_exists(self.reftree_bfu_fname)
FileUtils.remove_if_exists(self.optmod_fname)
FileUtils.remove_if_exists(self.lblalign_fname)
FileUtils.remove_if_exists(self.outgr_fname)
FileUtils.remove_if_exists(self.reduced_refalign_fname)
FileUtils.remove_if_exists(self.refalign_fname)
# top-level function to build a reference tree
def build_ref_tree(self):
start_time = time.time()
print "\n> Loading taxonomy from file: %s ...\n" % (self.cfg.taxonomy_fname)
self.taxonomy = GGTaxonomyFile(self.cfg.taxonomy_fname, EpacConfig.REF_SEQ_PREFIX)
print "\n=> Building a multifurcating tree from taxonomy with %d seqs ...\n" % self.taxonomy.seq_count()
self.validate_taxonomy()
self.build_multif_tree()
print "\n==> Building the reference alignment ...\n"
self.export_ref_alignment()
self.export_ref_taxonomy()
print "\n===> Saving the outgroup for later re-rooting ...\n"
self.save_rooting()
print "\n====> RAxML call: resolve multifurcation ...\n"
self.resolve_multif()
self.load_reduced_refalign()
print "\n=====> RAxML-EPA call: labeling the branches ...\n"
self.epa_branch_labeling()
print "\n======> Post-processing the EPA tree (re-rooting, taxonomic labeling etc.) ...\n"
self.epa_post_process()
self.calc_node_heights()
if self.cfg.verbose:
print "\n=======> Checking branch labels ...\n"
print "shared rank names before training: " + repr(self.taxonomy.get_common_ranks())
print "shared rank names after training: " + repr(self.mono_index())
print "\n=======> Saving the reference JSON file ...\n"
self.write_json()
elapsed_time = time.time() - start_time
print "\n*********** Done! (%.0f s) **********\n" % elapsed_time
def run_final_epa_test(self):
self.reftree_outgroup = self.refjson.get_outgroup()
tmp_reftree = self.reftree.copy(method="newick")
name2refnode = {}
for leaf in tmp_reftree.iter_leaves():
name2refnode[leaf.name] = leaf
tmp_taxtree = self.tax_tree.copy(method="newick")
name2taxnode = {}
for leaf in tmp_taxtree.iter_leaves():
name2taxnode[leaf.name] = leaf
for mis_rec in self.mislabels:
rname = mis_rec['name']
# rname = EpacConfig.REF_SEQ_PREFIX + name
if rname in name2refnode:
name2refnode[rname].delete()
else:
print "Node not found in the reference tree: %s" % rname
if rname in name2taxnode:
name2taxnode[rname].delete()
else:
print "Node not found in the taxonomic tree: %s" % rname
# remove unifurcation at the root
if len(tmp_reftree.children) == 1:
tmp_reftree = tmp_reftree.children[0]
self.mislabels = []
th = TaxTreeHelper(self.cfg, self.origin_taxonomy)
th.set_mf_rooted_tree(tmp_taxtree)
epa_result = self.run_epa_once(tmp_reftree)
reftree_epalbl_str = epa_result.get_std_newick_tree()
placements = epa_result.get_placement()
# update branchid-taxonomy mapping to account for possible changes in branch numbering
reftree_tax = Tree(reftree_epalbl_str)
th.set_bf_unrooted_tree(reftree_tax)
bid_tax_map = th.get_bid_taxonomy_map()
self.write_bid_tax_map(bid_tax_map, final=True)
cl = TaxClassifyHelper(self.cfg, bid_tax_map, self.rate, self.node_height)
# newtax_fname = self.cfg.subst_name("newtax_%NAME%.tre")
# th.get_tax_tree().write(outfile=newtax_fname, format=3)
final_ass = {}
for place in placements:
seq_name = place["n"][0]
# get original taxonomic label
orig_ranks = self.taxtree_helper.get_seq_ranks_from_tree(seq_name)
# EXPERIMENTAL FEATURE - disabled for now!
# It could happen that certain ranks were present in the "original" reference tree, but
# are completely missing in the pruned tree (e.g., all seqs of a species were considered "suspicious"
# after the leave-one-out test and thus pruned)
# In this case, EPA has no chance to infer full original taxonomic annotation (=species) since the corresponding clade
# is now missing. To account for this fact, we amend the original taxonomic annotation and set ranks missing from
# pruned tree to "Undefined".
# orig_ranks = th.strip_missing_ranks(orig_ranks)
# print orig_ranks
# get EPA tax label
ranks, lws = cl.classify_seq(place["p"])
final_ass[seq_name] = (ranks, lws)
#print seq_name, ": ", orig_ranks, "--->", ranks
# check if they match
mis_rec = self.check_seq_tax_labels(seq_name, orig_ranks, ranks, lws)
self.write_assignments(final_ass, final=True)
def run_final_epa_test(self):
self.reftree_outgroup = self.refjson.get_outgroup()
pruned_reftree = self.prune_mislabels_from_tree(self.reftree, "reference")
pruned_taxtree = self.prune_mislabels_from_tree(self.reftree, "taxonomic")
# remove unifurcation at the root
if len(pruned_reftree.children) == 1:
pruned_reftree = pruned_reftree.children[0]
self.mislabels = []
th = TaxTreeHelper(self.cfg, self.origin_taxonomy)
th.set_mf_rooted_tree(pruned_taxtree)
reftree_epalbl_str = None
if self.cfg.final_jplace_fname:
if os.path.isdir(self.cfg.final_jplace_fname):
jplace_fmask = os.path.join(self.cfg.final_jplace_fname, '*.jplace')
else:
jplace_fmask = self.cfg.final_jplace_fname
jplace_fname_list = glob.glob(jplace_fmask)
placements = []
for jplace_fname in jplace_fname_list:
jp = EpaJsonParser(jplace_fname)
placements += jp.get_placement()
if not reftree_epalbl_str:
reftree_epalbl_str = jp.get_std_newick_tree()
config.log.debug("Loaded %d final epa placements from %s\n", len(placements), jplace_fmask)
else:
epa_result = self.run_epa_once(pruned_reftree)
reftree_epalbl_str = epa_result.get_std_newick_tree()
placements = epa_result.get_placement()
# update branchid-taxonomy mapping to account for possible changes in branch numbering
reftree_tax = Tree(reftree_epalbl_str)
th.set_bf_unrooted_tree(reftree_tax)
bid_tax_map = th.get_bid_taxonomy_map()
self.write_bid_tax_map(bid_tax_map, final=True)
cl = TaxClassifyHelper(self.cfg, bid_tax_map, self.rate, self.node_height)
# newtax_fname = self.cfg.subst_name("newtax_%NAME%.tre")
# th.get_tax_tree().write(outfile=newtax_fname, format=3)
final_ass = {}
for place in placements:
seq_name = place["n"][0]
# get original taxonomic label
orig_ranks = self.taxtree_helper.get_seq_ranks_from_tree(seq_name)
# EXPERIMENTAL FEATURE - disabled for now!
# It could happen that certain ranks were present in the "original" reference tree, but
# are completely missing in the pruned tree (e.g., all seqs of a species were considered "suspicious"
# after the leave-one-out test and thus pruned)
# In this case, EPA has no chance to infer full original taxonomic annotation (=species) since the corresponding clade
# is now missing. To account for this fact, we amend the original taxonomic annotation and set ranks missing from
# pruned tree to "Undefined".
# orig_ranks = th.strip_missing_ranks(orig_ranks)
# print orig_ranks
# get EPA tax label
ranks, lws = cl.classify_seq(place["p"])
final_ass[seq_name] = (ranks, lws)
#print seq_name, ": ", orig_ranks, "--->", ranks
# check if they match
mis_rec = self.check_seq_tax_labels(seq_name, orig_ranks, ranks, lws)
self.write_assignments(final_ass, final=True)
def run_final_epa_test(self):
self.reftree_outgroup = self.refjson.get_outgroup()
pruned_reftree = self.prune_mislabels_from_tree(
self.reftree, "reference")
pruned_taxtree = self.prune_mislabels_from_tree(
self.reftree, "taxonomic")
# remove unifurcation at the root
if len(pruned_reftree.children) == 1:
pruned_reftree = pruned_reftree.children[0]
self.mislabels = []
th = TaxTreeHelper(self.cfg, self.origin_taxonomy)
th.set_mf_rooted_tree(pruned_taxtree)
reftree_epalbl_str = None
if self.cfg.final_jplace_fname:
if os.path.isdir(self.cfg.final_jplace_fname):
jplace_fmask = os.path.join(self.cfg.final_jplace_fname,
'*.jplace')
else:
jplace_fmask = self.cfg.final_jplace_fname
jplace_fname_list = glob.glob(jplace_fmask)
placements = []
for jplace_fname in jplace_fname_list:
jp = EpaJsonParser(jplace_fname)
placements += jp.get_placement()
if not reftree_epalbl_str:
reftree_epalbl_str = jp.get_std_newick_tree()
config.log.debug("Loaded %d final epa placements from %s\n",
len(placements), jplace_fmask)
else:
epa_result = self.run_epa_once(pruned_reftree)
reftree_epalbl_str = epa_result.get_std_newick_tree()
placements = epa_result.get_placement()
# update branchid-taxonomy mapping to account for possible changes in branch numbering
reftree_tax = Tree(reftree_epalbl_str)
th.set_bf_unrooted_tree(reftree_tax)
bid_tax_map = th.get_bid_taxonomy_map()
self.write_bid_tax_map(bid_tax_map, final=True)
cl = TaxClassifyHelper(self.cfg, bid_tax_map, self.rate,
self.node_height)
# newtax_fname = self.cfg.subst_name("newtax_%NAME%.tre")
# th.get_tax_tree().write(outfile=newtax_fname, format=3)
final_ass = {}
for place in placements:
seq_name = place["n"][0]
# get original taxonomic label
orig_ranks = self.taxtree_helper.get_seq_ranks_from_tree(seq_name)
# EXPERIMENTAL FEATURE - disabled for now!
# It could happen that certain ranks were present in the "original" reference tree, but
# are completely missing in the pruned tree (e.g., all seqs of a species were considered "suspicious"
# after the leave-one-out test and thus pruned)
# In this case, EPA has no chance to infer full original taxonomic annotation (=species) since the corresponding clade
# is now missing. To account for this fact, we amend the original taxonomic annotation and set ranks missing from
# pruned tree to "Undefined".
# orig_ranks = th.strip_missing_ranks(orig_ranks)
# print orig_ranks
# get EPA tax label
ranks, lws = cl.classify_seq(place["p"])
final_ass[seq_name] = (ranks, lws)
#print seq_name, ": ", orig_ranks, "--->", ranks
# check if they match
mis_rec = self.check_seq_tax_labels(seq_name, orig_ranks, ranks,
lws)
self.write_assignments(final_ass, final=True)
def run_final_epa_test(self):
self.reftree_outgroup = self.refjson.get_outgroup()
tmp_reftree = self.reftree.copy(method="newick")
name2refnode = {}
for leaf in tmp_reftree.iter_leaves():
name2refnode[leaf.name] = leaf
tmp_taxtree = self.tax_tree.copy(method="newick")
name2taxnode = {}
for leaf in tmp_taxtree.iter_leaves():
name2taxnode[leaf.name] = leaf
for mis_rec in self.mislabels:
rname = mis_rec['name']
# rname = EpacConfig.REF_SEQ_PREFIX + name
if rname in name2refnode:
name2refnode[rname].delete()
else:
print "Node not found in the reference tree: %s" % rname
if rname in name2taxnode:
name2taxnode[rname].delete()
else:
print "Node not found in the taxonomic tree: %s" % rname
# remove unifurcation at the root
if len(tmp_reftree.children) == 1:
tmp_reftree = tmp_reftree.children[0]
self.mislabels = []
th = TaxTreeHelper(self.cfg, self.origin_taxonomy)
th.set_mf_rooted_tree(tmp_taxtree)
epa_result = self.run_epa_once(tmp_reftree)
reftree_epalbl_str = epa_result.get_std_newick_tree()
placements = epa_result.get_placement()
# update branchid-taxonomy mapping to account for possible changes in branch numbering
reftree_tax = Tree(reftree_epalbl_str)
th.set_bf_unrooted_tree(reftree_tax)
bid_tax_map = th.get_bid_taxonomy_map()
self.write_bid_tax_map(bid_tax_map, final=True)
cl = TaxClassifyHelper(self.cfg, bid_tax_map, self.rate,
self.node_height)
# newtax_fname = self.cfg.subst_name("newtax_%NAME%.tre")
# th.get_tax_tree().write(outfile=newtax_fname, format=3)
final_ass = {}
for place in placements:
seq_name = place["n"][0]
# get original taxonomic label
orig_ranks = self.taxtree_helper.get_seq_ranks_from_tree(seq_name)
# EXPERIMENTAL FEATURE - disabled for now!
# It could happen that certain ranks were present in the "original" reference tree, but
# are completely missing in the pruned tree (e.g., all seqs of a species were considered "suspicious"
# after the leave-one-out test and thus pruned)
# In this case, EPA has no chance to infer full original taxonomic annotation (=species) since the corresponding clade
# is now missing. To account for this fact, we amend the original taxonomic annotation and set ranks missing from
# pruned tree to "Undefined".
# orig_ranks = th.strip_missing_ranks(orig_ranks)
# print orig_ranks
# get EPA tax label
ranks, lws = cl.classify_seq(place["p"])
final_ass[seq_name] = (ranks, lws)
#print seq_name, ": ", orig_ranks, "--->", ranks
# check if they match
mis_rec = self.check_seq_tax_labels(seq_name, orig_ranks, ranks,
lws)
self.write_assignments(final_ass, final=True)
