def retrieve_tophit_names(hitsfile,
filetype,
mincols=False,
evalue=False,
title_split_char='',
useScore=False):
"""
Returns the hit(s) with the lowest evalue, optionally filtering by minimum alignment span and maximum evalue.
If evalue is not provided (or if useScore==True), takes hit(s) with highest score
"""
hitnames = {}
#first, check that evalues are present, if required
if not useScore:
#first, find a result with a hit
for result in reptools.hitsParser(hitsfile,
filetype,
mincols=mincols,
evalue=evalue):
if len(result.hits) > 0:
if len(result.hits[0].hsps) > 0:
if result.hits[0].hsps[0].evalue is None:
useScore = True
break
for result in reptools.hitsParser(hitsfile,
filetype,
mincols=mincols,
evalue=evalue):
hitnames[result.id.split(title_split_char)[0]] = set()
#loop through the hits, recording the hit names
minevalue = False
maxscore = 0
for hit in result.hits:
for hsp in hit.hsps:
if useScore: #if we're looking at bitscore
if hsp.bitscore > maxscore:
hitnames[result.id.split(title_split_char)[0]] = set(
[hit.id])
maxscore = hsp.bitscore
elif hsp.bitscore == maxscore:
hitnames[result.id.split(title_split_char)[0]].add(
hit.id)
else:
pass
else: #if we're looking at evalue
if hsp.evalue < minevalue or minevalue is False:
hitnames[result.id.split(title_split_char)[0]] = set(
[hit.id])
minevalue = hsp.evalue
elif hsp.evalue == minevalue:
hitnames[result.id.split(title_split_char)[0]].add(
hit.id)
else:
pass
return (hitnames)
def parseHitfile(fn,aligner,title_split=' ',clusterID_position=0,mincols=False,id=False,evalue=False):
results = {}
for result in reptools.hitsParser(fn,aligner):
results[result.id.split(title_split)[clusterID_position]] = reptools.filterHits(
result.hits,
mincols=mincols,
id=id,
evalue=evalue
)
results = {k:results[k] for k in results if len(results[k])>0} #remove queries left empty by filter
return(results)
def get_top_hits(
hitsfile,filetype='blast',
title_split=None,clusterID_position=0,criteria=['bitscore','high'],
mincols=False,id=False,evalue=False
):
"""
Filetype should be "stellar","usearch", "ublast", "local", global", "blast", "swipe", or "dict"
Returns a dictionary keyed by query id, containing the best hit or hits for each query
"""
besthits = {}
if criteria[1].lower().startswith('high'):
selectionoperator = max
elif criteria[1].lower().startswith('low'):
selectionoperator = min
else:
raise ValueError(
'"criteria" must be a list of len=2, with the first item giving the name of the property to select on, and '
'the second being either "low" or "high"'
)
if filetype == 'dict':
for result in hitsfile:
besthits[result.split(title_split)[clusterID_position]] = [
#this nested list comp. selects hits if their max (or min) hsp value is equal to the
#max/min value of any hit
hit for hit in hitsfile[result]
if
selectionoperator([getattr(hsp,criteria[0]) for hsp in hit.hsps])
==
selectionoperator([
selectionoperator([getattr(hsp,criteria[0]) for hsp in hit2.hsps])
for hit2 in hitsfile[result]
])
]
else:
for result in reptools.hitsParser(hitsfile,filetype=filetype):
title = result.id.split(title_split)[clusterID_position]
besthits[title] = [
#this nested list comp. selects hits if their max (or min) hsp value is equal to the
#max/min value of any hit
hit for hit in result.hits
if
selectionoperator([getattr(hsp,criteria[0]) for hsp in hit.hsps])
==
selectionoperator([
selectionoperator([getattr(hsp,criteria[0]) for hsp in hit2.hsps])
for hit2 in result.hits
])
]
besthits = {k:reptools.filterHits(besthits[k],mincols=mincols,id=id,evalue=evalue) for k in besthits}
besthits = {k:besthits[k] for k in besthits if len(besthits[k])>0} #remove queries left empty by filter
return(besthits)
def find_nearest_to_overlap_hits(
hitsfile,gene,genedictionary,dict_entry,
filetype='blast',evalue=False,title_split=None,clusterID_position=0,
criteria=('bitscore','high'),
top_hits_only=False,returnhitstrings=False,
mincols=False, id=False
):
"""
Filetype should be "stellar","usearch", "ublast", "local", global", "blast", or "swipe"
"""
if returnhitstrings and not top_hits_only:
raise ValueError('returnhitstrings=True must be used with top_hits_only=True')
if criteria[1].lower().startswith('high'):
#selectionoperator = operator.gt
selectionoperator = max
elif criteria[1].lower().startswith('low'):
#selectionoperator = operator.lt
selectionoperator = min
else:
raise ValueError(
'"criteria" must be a list of len=2, with the first item giving the name of the property to select on, and '
'the second being either "low" or "high"'
)
besthits = {}
hitstrings = {}
for result in reptools.hitsParser(hitsfile,filetype=filetype,evalue=evalue):
closesthsp={}
#loop through the hits, and from each, select the hsp closest to the junction
if len(result.hits) == 0: continue
hits = result.hits
if top_hits_only: #first, select only the top hits, if instructed
hits = [
h for h in hits if
selectionoperator([getattr(hsp,criteria[0]) for hsp in h.hsps])
==
selectionoperator([
selectionoperator([getattr(hsp,criteria[0]) for hsp in h2.hsps])
for h2 in hits
])
]
hits = reptools.filterHits(hits, mincols=mincols, id=id) #filter hits by mincols and id
if len(hits) == 0 : continue
if returnhitstrings:
hitstrings[result.id.split(title_split)[clusterID_position]] = '+'.join([h.id for h in hits])
for hit in hits:
pos_H = genedictionary[gene][dict_entry][hit.id]
distancefromjunction=[]
for hsp in hit.hsps:
#does hsp surround hit?
if (hsp.hit_start<pos_H and hsp.hit_end>pos_H) or (hsp.hit_start>pos_H and hsp.hit_end<pos_H):
distancefromjunction.append(0)
else: #hsp doesn't surround hit
distancefromjunction.append(min(abs(hsp.hit_start-pos_H),abs(hsp.hit_end-pos_H)))
#select the hsps with the minimal distance from the junction
closest_dist = min(distancefromjunction)
candidatehsps = [(hit.hsps[n],d) for n,d in enumerate(distancefromjunction) if d==closest_dist]
scores = [getattr(hsptuple[0],criteria[0]) for hsptuple in candidatehsps]
closesthsp[hit.id] = candidatehsps[scores.index(selectionoperator(scores))]
#now select the best overall hit: ties are broken
#need to pick closest hit by distance, then the best of the closest by selection criteria
dists = [closesthsp[hit][1] for hit in closesthsp]
closesthits = [closesthsp[hit][0] for hit in closesthsp if closesthsp[hit][1]==min(dists)]
scores = [getattr(hsp,criteria[0]) for hsp in closesthits]
besthits[result.id.split(title_split)[clusterID_position]]=closesthits[scores.index(selectionoperator(scores))]
if returnhitstrings:
return(tuple([besthits,hitstrings]))
else:
return(besthits)