def scatterplot_phylodist_to_score(blast_dirname,filename_rex,phylo_dirname,NONMEMBER_DIST=10.):
"""
Similar to scatterplot_length_to_maxscore, except that we're plotting
the branch distance to maximum observed score. And we do it PER family
blast file since branch distances make sense only within a family....
Also:
We assume that the blast files not only match filename_rex, but also
ends in '.WUblast'
We assume that under <phylo_dirname>, there will only be 1 .outtree file
that fits the filename_rex-extracted family name
We also assume that this blast output is a "real" vs "padded_real + negative",
so id1 is always not padded, and id2 is potentially padded
"""
for filename in os.listdir(blast_dirname):
if os.path.exists(blast_dirname+'/'+filename+'.phylodist_to_score'):
continue
maxscore_of_dist = defaultdict(lambda: float("-inf"))
m = filename_rex.match(filename)
if m is None or not filename.endswith('WUblast'):
continue
family = m.group(1)
phylo_filename = fnmatch.filter(os.listdir(phylo_dirname),"*{0}*.outtree".format(family))[0]
p = NewickReader(phylo_dirname+'/'+phylo_filename)
print >> sys.stderr, filename
with open(blast_dirname+'/'+filename,'r') as f:
for line in f:
d = parseWUBLASTline(line)
id1,id2 = d['id1'],d['id2']
if id1[:id1.find('_')]==id2[:id2.find('_')] and id1.startswith(family):
try:
dist = p.distance(parse_id(id1,True)[1], parse_id(id2,False)[1])
maxscore_of_dist[dist] = max(maxscore_of_dist[dist], d['sprime'])
except:
print >> sys.stderr, "failed phylo dist on ",id1,id2
else: # a non-hit
maxscore_of_dist[NONMEMBER_DIST] = max(maxscore_of_dist[NONMEMBER_DIST], d['sprime'])
dists = maxscore_of_dist.keys()
dists.sort()
with open(blast_dirname+'/'+filename+'.phylodist_to_score','w') as f:
for k in dists:
f.write("{0}\t{1}\n".format(k, maxscore_of_dist[k]))
def evaluate_blast_graph_by_phylo(graph_or_filename,phylo_filename,ignore_prefix=['shuffled_','random_']):
"""
Reads the blast-processed XGraph (or the filename of the pickled XGraph)
The REAL ncRNA seq IDs should be in format <family>_<DB id>
The FAKE seq IDs should be in format <ignore_prefix>xxxxx.....
Returns (result,sens_by_family,spec_by_family)
where result is an obsolete junk (so just make it None) for now
phylo_sum_by_family is dict with family --> list of sums of neighbor phylo distances
for all nodes in this family
spec_by_family is dict with family --> list of spec(ratio) for all nodes in this family
"""
if type(graph_or_filename) is XGraph:
X = graph_or_filename
else:
X = read_gpickle(graph_or_filename)
p = NewickReader(phylo_filename)
phylo_sum_by_family = defaultdict(lambda: [])
spec_by_family = defaultdict(lambda: [])
total_size_by_family = defaultdict(lambda: 0)
for n in X.nodes_iter():
i = n.rfind('_')
family,db_id = n[:i],n[(i+1):]
if any(map(lambda x: family.startswith(x), ignore_prefix)): continue
total_size_by_family[family] += 1
if X.degree(n) == 0:
spec_by_family[family].append(0)
phylo_sum_by_family[family].append(0)
continue
phylo_sum,tp = 0,0
for m in X.neighbors_iter(n):
if any(map(lambda x: family.startswith(x), ignore_prefix)): continue
m1 = parse_id(m,True)
if m1[0] == family:
try:
phylo_sum += p.distance(db_id,m1[1])
except:
pass
tp += 1
phylo_sum_by_family[family].append(phylo_sum)
spec_by_family[family].append(tp*1./X.degree(n))
return (None,phylo_sum_by_family,spec_by_family)