def introns(self, name):
if not h.has_key(name):
return []
retv = []
r = h[name]
for i in r.Introns():
if i.stop - i.start > 60:
j = i.tail(60)
v = [Tools.translate_coordinate(j, d) for d in DataS3.query(j)]
retv.append({
"chr": j.chr,
"id": j.id,
"start": j.start,
"end": j.end,
"strand": j.strand,
"v": v
})
return retv
def addCoverageAndSeqScore():
'''
two step,
1. phase change threshold determin
2. find the nearest possible splicing sites
'''
#logging.debug(length)
#logging.debug(g)
diff=[array[0]] # assume the pos -1 is 0.0
for j in range(length-1):
diff.append(math.log((array[j+1]+1)/(array[j]+1.0)))
def mean_sd(indexes):
n=0
s=0.0
sd=0.0
a=[]
for k in indexes:
n+=1
max_diff=0.0
for k0 in nearbyIter(k,args.merge_bp,length):
if k0>=length: continue
if abs(max_diff) < abs(diff[k0]):
max_diff=diff[k0]
s+=max_diff
a.append(max_diff)
if n>0:
mean=s/n
else:
return 0.0,0.0
for k in a:
sd+=(k-mean)*(k-mean)
if n > 1:
sd=sd/(n-1)
if n > 1:
sd=math.sqrt(sd/(n-1))
else:
sd=0.0
return mean,sd
(meanDonor,sdDonor)=mean_sd(hDonor.keys())
(meanAcceptor,sdAcceptor)=mean_sd(hAcceptor.keys())
#print meanDonor,sdDonor
#print meanAcceptor,sdAcceptor
'''
adding possible splicing sites based on diff value
'''
def correctToNearAcceptor(j):
signal=0
for k in nearbyIter(j,args.merge_bp,length):
if hAcceptor.has_key(k):
return k
for k in nearbyIter(j,args.merge_bp,length):
try:
if ag[k]==1:
return k+2
except:
pass
return -1
def correctToNearDonor(j):
signal=0
for k in nearbyIter(j,args.merge_bp,length):
if hDonor.has_key(k):
return k
for k in nearbyIter(j,args.merge_bp,length):
try:
if gt[k]==1:
return k
except:
pass
return -1
for j in range(length):
#if diff[j] > sdAcceptor + meanAcceptor:
if diff[j] > meanAcceptor + sdAcceptor and diff[j] > 0.0:
k=correctToNearAcceptor(j)
if k!=-1 and not hAcceptor.has_key(k):
hAcceptor[k]=1
g.append(TuringCode(k,cb.BLOCKON))
# add to revise small fragrments ?
'''
elif k==-1 and diff[j] > meanAcceptor + 3 * sdAcceptor and not hAcceptor.has_key(j):
hAcceptor[j]=1
l.append((j,cb.BLOCKON))
'''
#if diff[j] < meanDonor - sdDonor:
if diff[j] < meanDonor - sdDonor and diff[j] < 0.0:
k=correctToNearDonor(j)
if k!=-1 and not hDonor.has_key(k):
hDonor[k]=1
g.append(TuringCode(k,cb.BLOCKOFF))
'''
elif k==-1 and diff[j] < meanDonor - 3 * sdDonor and not hDonor.has_key(j):
hDonor[j]=1
l.append((j,cb.BLOCKOFF))
'''
if isinstance(x,BED12): #strange could not use i?
ti=Tools.translate_coordinates(x,x)
#print "debug ti:",tia
ti_codes=[]
for j in ti.blockStarts:
k=correctToNearAcceptor(j)
if k==-1:
ti_codes.append(TuringCode(j,cb.BLOCKON))
else:
ti_codes.append(TuringCode(k,cb.BLOCKON))
if k!=-1 and not hAcceptor.has_key(k):
hAcceptor[k]=1
g.append(TuringCode(k,cb.BLOCKON))
if k==-1 and not hAcceptor.has_key(j):
hAcceptor[j]=1
g.append(TuringCode(j,cb.BLOCKON))
last_stop=0
for j,m in itertools.izip(ti.blockStarts,ti.blockSizes):
blockStop=j+m
#print "debug blockStop",blockStop
k=correctToNearDonor(blockStop)
if k==-1:
ti_codes.append(TuringCode(blockStop,cb.BLOCKOFF))
last_stop=j
else:
ti_codes.append(TuringCode(k,cb.BLOCKOFF))
last_stop=k
if k!=-1 and not hDonor.has_key(k):
hDonor[k]=1
g.append(TuringCode(k,cb.BLOCKOFF))
elif k==-1 and not hDonor.has_key(blockStop):
hDonor[blockStop]=1
g.append(TuringCode(blockStop,cb.BLOCKOFF))
ti_codes.append(TuringCode(correctToNearAcceptor(ti.blockStarts[0]),cb.ON))
ti_codes.append(TuringCode(last_stop,cb.OFF))
ti_codes.sort()
#print "debug ti_codes",ti_codes
#print ti_codes
#TODO GENERate start path.
# l.sort()
graph=TuringGraph(g)
g_len=len(graph)
initial_bits=graph.translate_path_into_bits(TuringGraph(ti_codes),args.merge_bp)
#print "debug ", initial_bits
for i in range(0,2*g_len,2):
if initial_bits[i]==True and initial_bits[i+1]==False:
initial_bits[i+1]=True
initial_bits[-1]=True
initial_bits[-2]=True
return initial_bits,graph
# l.sort()
return None,TuringGraph(g)
def query(i,dbi_bam,genome): # i is query iteem
h={}
hc={}
#retv="" #RETURN VALUE
#retv+="BEGIN\nQR\t"+str(i)+"\n"
x=i
ret_dict={}
ret_dict["QR"]=i
'''
PART A: scoring and adding splicing sites
'''
print >>sys.stderr,"QR",str(i)
length=i.stop-i.start
g=list()
g.append(TuringCode(0,cb.ON))
g.append(TuringCode(0,cb.BLOCKON))
g.append(TuringCode(length,cb.OFF))
g.append(TuringCode(length,cb.BLOCKOFF))
#logging.debug(g)
array=[0.0 for j in xrange(length)]
donorSitesScore=[0.0 for j in xrange(length+1)]
acceptorSitesScore=[0.0 for j in xrange(length+1)]
seq=genome.get_seq(i).upper()
for j in dbi_bam.query(i,method="bam1",strand=args.strand):
read=Tools.translate_coordinates(i,j)
#TODO : get read anc call peaks
if read.strand=="+":
for k in read.Exons():
for k0 in xrange(k.start,k.stop):
if k0>=length:
break
if k0>=0:
array[k0]+=1
for intron in read.Introns():
if len(intron)< MIN_INTRON_LENGTH: continue
if intron.start > 0 and intron.start <length:
donorSitesScore[intron.start]+=1
if intron.stop > 0 and intron.stop <length:
acceptorSitesScore[intron.stop]+=1
gt=[0 for j in range(length)]
ag=[0 for j in range(length)]
for j in xrange(length-1):
if seq[j]=="G" and seq[j+1]=="T":
gt[j]=1
elif seq[j]=="A" and seq[j+1]=="G":
ag[j]=1
'''
Score SplicingScores
'''
#DONE CLean Scores [ Minus the near by scores]
sharpScore(donorSitesScore,args.merge_bp)
sharpScore(acceptorSitesScore,args.merge_bp)
hDonor={}
hAcceptor={}
hAcceptor[0]=1
hDonor[length]=1
for j in xrange(length):
if donorSitesScore[j] > MIN_SPLICING_SITES_SCORE:
g.append(TuringCode(j,cb.BLOCKOFF))
hDonor[j]=1
elif acceptorSitesScore[j] > MIN_SPLICING_SITES_SCORE:
g.append(TuringCode(j,cb.BLOCKON))
hAcceptor[j]=1
'''
closure to score sites by coverage
'''
def addCoverageAndSeqScore():
'''
two step,
1. phase change threshold determin
2. find the nearest possible splicing sites
'''
#logging.debug(length)
#logging.debug(g)
diff=[array[0]] # assume the pos -1 is 0.0
for j in range(length-1):
diff.append(math.log((array[j+1]+1)/(array[j]+1.0)))
def mean_sd(indexes):
n=0
s=0.0
sd=0.0
a=[]
for k in indexes:
n+=1
max_diff=0.0
for k0 in nearbyIter(k,args.merge_bp,length):
if k0>=length: continue
if abs(max_diff) < abs(diff[k0]):
max_diff=diff[k0]
s+=max_diff
a.append(max_diff)
if n>0:
mean=s/n
else:
return 0.0,0.0
for k in a:
sd+=(k-mean)*(k-mean)
if n > 1:
sd=sd/(n-1)
if n > 1:
sd=math.sqrt(sd/(n-1))
else:
sd=0.0
return mean,sd
(meanDonor,sdDonor)=mean_sd(hDonor.keys())
(meanAcceptor,sdAcceptor)=mean_sd(hAcceptor.keys())
#print meanDonor,sdDonor
#print meanAcceptor,sdAcceptor
'''
adding possible splicing sites based on diff value
'''
def correctToNearAcceptor(j):
signal=0
for k in nearbyIter(j,args.merge_bp,length):
if hAcceptor.has_key(k):
return k
for k in nearbyIter(j,args.merge_bp,length):
try:
if ag[k]==1:
return k+2
except:
pass
return -1
def correctToNearDonor(j):
signal=0
for k in nearbyIter(j,args.merge_bp,length):
if hDonor.has_key(k):
return k
for k in nearbyIter(j,args.merge_bp,length):
try:
if gt[k]==1:
return k
except:
pass
return -1
for j in range(length):
#if diff[j] > sdAcceptor + meanAcceptor:
if diff[j] > meanAcceptor + sdAcceptor and diff[j] > 0.0:
k=correctToNearAcceptor(j)
if k!=-1 and not hAcceptor.has_key(k):
hAcceptor[k]=1
g.append(TuringCode(k,cb.BLOCKON))
# add to revise small fragrments ?
'''
elif k==-1 and diff[j] > meanAcceptor + 3 * sdAcceptor and not hAcceptor.has_key(j):
hAcceptor[j]=1
l.append((j,cb.BLOCKON))
'''
#if diff[j] < meanDonor - sdDonor:
if diff[j] < meanDonor - sdDonor and diff[j] < 0.0:
k=correctToNearDonor(j)
if k!=-1 and not hDonor.has_key(k):
hDonor[k]=1
g.append(TuringCode(k,cb.BLOCKOFF))
'''
elif k==-1 and diff[j] < meanDonor - 3 * sdDonor and not hDonor.has_key(j):
hDonor[j]=1
l.append((j,cb.BLOCKOFF))
'''
if isinstance(x,BED12): #strange could not use i?
ti=Tools.translate_coordinates(x,x)
#print "debug ti:",tia
ti_codes=[]
for j in ti.blockStarts:
k=correctToNearAcceptor(j)
if k==-1:
ti_codes.append(TuringCode(j,cb.BLOCKON))
else:
ti_codes.append(TuringCode(k,cb.BLOCKON))
if k!=-1 and not hAcceptor.has_key(k):
hAcceptor[k]=1
g.append(TuringCode(k,cb.BLOCKON))
if k==-1 and not hAcceptor.has_key(j):
hAcceptor[j]=1
g.append(TuringCode(j,cb.BLOCKON))
last_stop=0
for j,m in itertools.izip(ti.blockStarts,ti.blockSizes):
blockStop=j+m
#print "debug blockStop",blockStop
k=correctToNearDonor(blockStop)
if k==-1:
ti_codes.append(TuringCode(blockStop,cb.BLOCKOFF))
last_stop=j
else:
ti_codes.append(TuringCode(k,cb.BLOCKOFF))
last_stop=k
if k!=-1 and not hDonor.has_key(k):
hDonor[k]=1
g.append(TuringCode(k,cb.BLOCKOFF))
elif k==-1 and not hDonor.has_key(blockStop):
hDonor[blockStop]=1
g.append(TuringCode(blockStop,cb.BLOCKOFF))
ti_codes.append(TuringCode(correctToNearAcceptor(ti.blockStarts[0]),cb.ON))
ti_codes.append(TuringCode(last_stop,cb.OFF))
ti_codes.sort()
#print "debug ti_codes",ti_codes
#print ti_codes
#TODO GENERate start path.
# l.sort()
graph=TuringGraph(g)
g_len=len(graph)
initial_bits=graph.translate_path_into_bits(TuringGraph(ti_codes),args.merge_bp)
#print "debug ", initial_bits
for i in range(0,2*g_len,2):
if initial_bits[i]==True and initial_bits[i+1]==False:
initial_bits[i+1]=True
initial_bits[-1]=True
initial_bits[-2]=True
return initial_bits,graph
# l.sort()
return None,TuringGraph(g)
initial_bits,g=addCoverageAndSeqScore()
'''
PART B : Report wig
retv+="WIG"+"\n"
retv+= "index\tnt\tcoverage\tdonorSitesScore\tacceptorSitesScore\tGT\tAG\n"
for j in xrange(len(i)):
retv+= str(j)+"\t"+str(seq[j])+"\t"+str(array[j])+"\t"+str(donorSitesScore[j])+"\t"+str(acceptorSitesScore[j])+"\t"+str(gt[j])+"\t"+str(ag[j])+"\n"
'''
#logging.debug(initial_bits)
if args.report_seq:
ret_dict["WIG_TABLE"]=(seq,array,donorSitesScore,acceptorSitesScore,gt,ag)
'''
end of adding wig
'''
'''
PART C: GREEDY ADDING ISOFORMS
'''
# bitarray_path=bitarray(2*len(g))
# bitarray_path.setall(True)
#TODO change bitarray?
#logging.debug(g)
ret_dict["FIGURE"]=g.graph_str(600)
h={}
hc={}
j0=0;
total_frag=0
#for j in g: print "debug g:",j
#print "INIT",bitarray_to_rep(initial_bits)
g_len=len(g)
if initial_bits is None:
initial_bits=bitarray(2*g_len)
initial_bits.setall(True)
for j in dbi_bam.query(i,method="bam2fast",strand=args.strand):
p=[]
#logging.debug(j)
#print "debug",j
if j[0].strand!=i.strand: continue
for k in j:
p.append(TuringFactory(Tools.translate_coordinates(i,k))) #TODO check this
# p.append(TuringFactory(Tools.translate_coordinates(i,j))) #TODO check this
#print "debug glen:",g_len
a=g.translate_paths_into_bits(p,args.merge_bp)
#print "debug bits: ",a
#print "debug bed :",translate_bits_into_bed(g,a)
if isSelfIncompatible(a): continue
if h.has_key(a.tobytes()):
h[a.tobytes()]+=1
if hc[a.tobytes()]!=a:
print >>sys.stderr,"WARNING"
else:
h[a.tobytes()]=1
hc[a.tobytes()]=a
j0+=1
#retv+="PATTERN_NUMBER\t"+str(len(h.keys()))+"\n"
ret_dict["PATTERN_NUMBER"]=len(h.keys())
# print >>sys.stderr,"debug pattern number",ret_dict["PATTERN_NUMBER"]
#retv+="FRG_NUMBER\t"+str(j0)+"\n"
ret_dict["FRG_NUMBER"]=j0
total_frag=j0
if j0==0: return ret_dict
sorted_keys=h.keys()
if args.sort_model==0:
sorted_keys.sort(lambda x,y:bitarray_to_intron_number(hc[x])*h[x]-bitarray_to_intron_number(hc[y])*h[y] or h[x]-h[y], reverse=True)
elif args.sort_model==1:
sorted_keys.sort(lambda x,y:bitarray_to_intron_number(hc[x])-bitarray_to_intron_number(hc[y]) or h[x]-h[y], reverse=True)
else:
sorted_keys.sort(lambda x,y:h[x]-h[y] or bitarray_to_intron_number(hc[x])-bitarray_to_intron_number(hc[y]), reverse=True)
clique=[]
cliques=[]
bits_no_prior=bitarray(g_len*2)
bits_no_prior.setall(True)
cliques_pattern=[]
ret_dict["PATTERNS"]=list()
keys_len=len(sorted_keys)
joined_cliques=[False for i0 in range(keys_len)]
unjoined_pattern=range(0,keys_len)
for j,key in enumerate(sorted_keys):
ret_dict["PATTERNS"].append((bitarray_to_rep(hc[key]),h[key],bitarray_to_intron_number(hc[key])))
if joined_cliques[j] : continue
joined_clique=False
joined_cliques[j]=True
bits_no_prior=bitarray(g_len*2)
bits_no_prior.setall(True)
h_clique={}
if isCompatible(initial_bits,hc[sorted_keys[j]]):
max_index=0
clique=[]
clique.append(j)
cliques.append(clique)
cliques_pattern.append(hc[sorted_keys[j]])
h_clique[j]=1
update=True
while update:
update=False
logging.debug("update:%s",update)
for k0,k in enumerate(unjoined_pattern):
logging.debug("del %i",k)
if k==j:
del unjoined_pattern[k0]
elif isOverlapCompatible(cliques_pattern[-1],hc[sorted_keys[k]]):
update=True
cliques[-1].append(k)
cliques_pattern[-1]=bitarray_and(cliques_pattern[-1],hc[sorted_keys[k]])
joined_cliques[k]=True
del unjoined_pattern[k0]
h_clique[k]=1
break
if update==False:
if k==j:
continue
for k in range(keys_len):
if h_clique.has_key(k): continue
if isOverlapCompatible(cliques_pattern[-1],hc[sorted_keys[k]]):
update=True
cliques[-1].append(k)
cliques_pattern[-1]=bitarray_and(cliques_pattern[-1],hc[sorted_keys[k]])
h_clique[k]=1
break
ret_dict["CLIQUES"]=list()
cumulative_score=0
#update all cliques pattern.
j0=0
#buffer_keys=[] # those key are ignored in previous uniqs
max_uniq_fpk=0.0
max_uniq=0
for j,x in enumerate(cliques):
score=0
cliques_pattern[j][-1]=True
cliques_pattern[j][-2]=True
uniq_score=0
for k,y in enumerate(x):
uniq_score+=h[sorted_keys[y]]
'''
for k,y in enumerate(buffer_keys):
if isCompatible(cliques_pattern[buffer_keys[k]],
cliques[j].append(k)
'''
if float(uniq_score)/total_frag < args.min_uniq :
#print >>sys.stderr,"ignore due to small uniq frags:",float(uniq_score)/total_frag
continue
else:
logging.debug(cliques_pattern)
bed=g.translate_bits_into_bed(cliques_pattern[j])
cdna_length=bed.cdna_length()
if cdna_length==0:
#print >>sys.stderr,"ignore due to cdna_length",bed
continue
uniq_fpk=float(uniq_score)/cdna_length*1000.0
if uniq_fpk > max_uniq_fpk:
max_uniq_fpk=uniq_fpk
if uniq_fpk < max_uniq_fpk * MIN_FPK_RATIO:
#print >>sys.stderr,"ignore due to fpk",bed,"\tuniq_fpk:",uniq_fpk,"\t current max:",max_uniq_fpk
continue
rgb=200-uniq_score*200/total_frag
cumulative_score+=uniq_score
j0+=1
#retv+="NO."+str(j0)+" CLIQUE"+"\n"
#retv+="CLIQUE\t"+str(c)+"\nUNIQ\t"+str(cliques[j])+"\nPATTERN\t"+bitarray_to_rep(cliques_pattern[j])+"\n"
ret_dict["CLIQUES"].append(dict())
#ret_dict["CLIQUES"][-1]["CLIQUE"]=(c,cliques[j],bitarray_to_rep(cliques_pattern[j]))
ret_dict["CLIQUES"][-1]["ALL_GROUP"]=cliques[j]
ret_dict["CLIQUES"][-1]["UNIQ_GROUP"]=cliques[j]
ret_dict["CLIQUES"][-1]["REP"]=bitarray_to_rep(cliques_pattern[j])
#retv+="CLIQUE\t",c,"\nUNIQ\t",cliques[j],"\nPATTERN\t",cliques_pattern[j]
'''
pattern
need to revise
'''
pattern=cliques_pattern[j]
pattern[-1]=True
pattern[-2]=True
beds=[bed for bed in g.translate_bits_into_beds(cliques_pattern[j])]
ret_dict["CLIQUES"][-1]["BEDS"]=[]
for j1,bed in enumerate(beds):
#logging.debug(i)
score=score*1000.0/cdna_length
chr=i.id
id=i.id+"_"+"NO."+str(j0)+"_"+str(j1)
itemRgb=str(rgb)+","+str(rgb)+","+str(rgb)
bed=bed._replace(chr=chr,id=id,itemRgb=itemRgb)
ret_dict["CLIQUES"][-1]["BEDS"].append(str(Tools.translate_coordinates(i,bed,True)))
ret_dict["CLIQUES"][-1]["UNIQ_SCORE"]=uniq_score
ret_dict["CLIQUES"][-1]["UNIQ_FPK"]=uniq_fpk
ret_dict["CLIQUES"][-1]["SCORE"]=score
#retv+="UNIQ\t"+str(uniq_score)+"\tBED\t"+str(Tools.translate_coordinates(Bed(i),bed,True))+"\n"
#print >>sys.stderr,"UNIQ_FPK/CURRENT_MAX_UNIQ_FPK",uniq_fpk/max_uniq_fpk
#retv+="\n"
ret_dict["INTERPRET_FRG"]=float(cumulative_score)/total_frag
return ret_dict
def BamToBed12(handle,uniq=False,**kwargs):
'''
handle is an bam iterator
need references hash if handle is not filename.
score="NH" ,will report tag NH as score
'''
if type(handle)==type("string"):
handle=pysam.Samfile(handle,"rb");
for i in handle:
if i.tid<0: continue
if uniq:
nh=_get_tag_score(i,"NH")
if nh:
if nh > 1:
continue
else:
raise "no NH tag in your bam file"
strand="+"
if i.is_reverse:
strand="-"
score=i.mapq
'''
test
'''
if kwargs.has_key("references"):
if isinstance(kwargs["references"],str):
chr=kwargs["references"]
else:
chr=kwargs["references"][i.tid];
else:
try:
chr=handle.references[i.tid];
except:
chr="chr"
if kwargs.has_key("strand"):
if kwargs["strand"]=="read1" or kwargs["strand"]=="firstMate":
read1=True
else:
read1=False
else:
read1=True
start=i.pos
end=i.aend
name=i.qname
cds_start=start
cds_end=start
itemRgb="0,0,0"
if not uniq:
'''
try to put NH score in itemRgb
'''
try:
nh=_get_tag_score(i,"NH")
if nh:
itemRgb=str(nh)+",0,0"
except:
pass
'''
debug
import sys
if i.cigar is None:
print >>sys.stderr,"why cigar is Nonetype?"
print >>sys.stderr,i
exit(0)
end of debug
'''
if i.cigar==None: continue # IGNORE THIS READS?
(block_starts,block_sizes)=Tools.cigar_to_coordinates(i.cigar);
if i.is_read1 and not read1:
strand=Tools.reverse_strand(strand)
elif i.is_read2 and read1:
strand=Tools.reverse_strand(strand)
bed=Bed12(chr,start,end,name,score,strand,cds_start,cds_end,itemRgb,len(block_sizes),block_sizes,block_starts)
yield bed
