def RemoveDupSeq(infile, g_outpath, method, isUseMD5):#{{{
if g_outpath == "":
outpath = myfunc.my_dirname(infile)
else:
outpath = g_outpath
rootname = os.path.basename(os.path.splitext(infile)[0])
outfile = "%s%s%s"%(outpath, os.sep, rootname)
fpout = myfunc.myopen(outfile, None, "w", False)
if fpout == None:
return 1
hdl = myfunc.ReadFastaByBlock(infile)
if hdl.failure:
return -1
myset = set([])
recordList = hdl.readseq()
while recordList != None:
for rd in recordList:
if method == "id":
key = rd.seqid
elif method == "seq":
if isUseMD5:
key = md5.new(rd.seq).digest()
else:
key = rd.seq
if not key in myset:
myset.add(key)
fpout.write(">%s\n%s\n"%(rd.description, rd.seq))
recordList = hdl.readseq()
hdl.close()
myfunc.myclose(fpout)
return 0
def RemoveDupSeq(infile, g_outpath, method, isUseMD5): #{{{
if g_outpath == "":
outpath = myfunc.my_dirname(infile)
else:
outpath = g_outpath
rootname = os.path.basename(os.path.splitext(infile)[0])
outfile = "%s%s%s" % (outpath, os.sep, rootname)
fpout = myfunc.myopen(outfile, None, "w", False)
if fpout == None:
return 1
hdl = myfunc.ReadFastaByBlock(infile)
if hdl.failure:
return -1
myset = set([])
recordList = hdl.readseq()
while recordList != None:
for rd in recordList:
if method == "id":
key = rd.seqid
elif method == "seq":
if isUseMD5:
key = md5.new(rd.seq).digest()
else:
key = rd.seq
if not key in myset:
myset.add(key)
fpout.write(">%s\n%s\n" % (rd.description, rd.seq))
recordList = hdl.readseq()
hdl.close()
myfunc.myclose(fpout)
return 0
def main(g_params): #{{{
argv = sys.argv
numArgv = len(argv)
if numArgv < 2:
PrintHelp()
return 1
SPE_PAIR_LIST = [(2, 1), (2, 4), (2, 6), (2, 8), (3, 6), (3, 7), (4, 6),
(4, 8), (4, 10), (5, 7), (5, 10), (6, 8), (6, 10),
(6, 12), (7, 14), (8, 10), (8, 12), (10, 12), (10, 13),
(11, 13), (12, 14)]
outfile = ""
infile = ""
pfamDefFile = "%s/data/pfam/pfam26.0/Pfam-A.clans.tsv" % (DATADIR3)
signalpFile = "%s/wk/MPTopo/pfamAna_refpro/pred_signalp/refpro20120604-celluar.selmaxlength-m1.nr100.signalp_list" % (
DATADIR3)
#seqid2clanidMapFile = "%s/wk/MPTopo/pfamAna_refpro/pfammap_from_uniprot/refpro20120604-celluar.selmaxlength-m1.nr100.filter.fragmented.seqid2clanid"%(DATADIR3)
#seqid2pfamidMapFile = "%s/wk/MPTopo/pfamAna_refpro/pfammap_from_uniprot/refpro20120604-celluar.selmaxlength-m1.nr100.filter.fragmented.seqid2pfamid"%(DATADIR3)
seqid2clanidMapFile = ""
seqid2pfamidMapFile = ""
tm_pfamidListFile = ""
tm_clanidListFile = ""
pfamid2seqidMapFile = ""
clanid2seqidMapFile = ""
dbname_predTM = ""
pairlistwithpfamidFile = ""
pfamtype = ""
pairListFile = ""
#classList_TableNumTMHeatMap = ["ALL", "RMSP"]
classList_TableNumTMHeatMap = ["ALL"]
i = 1
isNonOptionArg = False
while i < numArgv:
if isNonOptionArg == True:
infile = argv[i]
isNonOptionArg = False
i += 1
elif argv[i] == "--":
isNonOptionArg = True
i += 1
elif argv[i][0] == "-":
if argv[i] in ["-h", "--help"]:
PrintHelp()
return 1
elif argv[i] in ["-o", "--o", "-outfile"]:
(outfile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-outpath", "--outpath"]:
(g_params['outpath'], i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-l", "--l"]:
(fileListFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-pfamdef", "--pfamdef"]:
(pfamDefFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-signalp", "--signalp"]:
(signalpFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-mp", "--mp"]:
g_params[
'pairwise_comparison_method'], i = myfunc.my_getopt_int(
argv, i)
elif argv[i] in ["-mindiffpair", "--mindiffpair"]:
g_params['mindiffpair'], i = myfunc.my_getopt_int(argv, i)
elif argv[i] in ["-pfamtype", "--pfamtype"]:
pfamtype, i = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-clanidlist", "--clanidlist"]:
(tm_clanidListFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-pfamidlist", "--pfamidlist"]:
(tm_pfamidListFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-seqid2clanid", "--seqid2clanid"]:
(seqid2clanidMapFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-seqid2pfamid", "--seqid2pfamid"]:
(seqid2pfamidMapFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-pfamid2seqid", "--pfamid2seqid"]:
(pfamid2seqidMapFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-clanid2seqid", "--clanid2seqid"]:
(clanid2seqidMapFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-pairlistwithpfamid", "--pairlistwithpfamid"]:
(pairlistwithpfamidFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-predTMdbname", "--predTMdbname"]:
(dbname_predTM, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-pairlist", "--pairlist"]:
(pairListFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-winsize", "--winsize"]:
(g_params['winsize'], i) = myfunc.my_getopt_int(argv, i)
elif argv[i] in ["-outname", "--outname"]:
(g_params['outname'], i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-q", "--q"]:
g_params['isQuiet'] = True
i += 1
elif argv[i] in ["-prokar", "--prokar"]:
g_params['isOnlyAnaProkar'] = True
i += 1
elif argv[i] in ["-eukar", "--eukar"]:
g_params['isOnlyAnaEukar'] = True
i += 1
else:
print >> sys.stderr, "Error! Wrong argument:", argv[i]
return 1
else:
infile = argv[i]
i += 1
if myfunc.checkfile(
infile, "%s (line %d): infile" %
(__file__, inspect.currentframe().f_lineno)) != 0:
return 1
dirpath = myfunc.my_dirname(infile)
# try to obtain Pfam family tag
tag = ""
if pfamtype != "":
if pfamtype.upper().find("FAM") != -1:
tag = ".Family"
elif pfamtype.upper().find("DOM") != -1:
tag = ".Domain"
elif pfamtype.upper().find("REP") != -1:
tag = ".Repeat"
elif pfamtype.upper().find("MOT") != -1:
tag = ".Motif"
else:
tag = ""
else:
if infile.find(".Family.") != -1:
tag = ".Family"
elif infile.find(".Domain.") != -1:
tag = ".Domain"
elif infile.find(".Repeat.") != -1:
tag = ".Repeat"
elif infile.find(".Motif.") != -1:
tag = ".Motif"
else:
tag = ""
if seqid2clanidMapFile == "":
seqid2clanidMapFile = "%s/wk/MPTopo/pfamAna_refpro/pfammap_from_uniprot/Pfam-A-full.seqfrompfamfasta.percentTMpro_scampi.perTM75_nseq20.nr100.filter.fragmented.seqid2clanid" % (
DATADIR3)
if myfunc.checkfile(
seqid2clanidMapFile, "%s (line %d): seqid2clanidMapFile" %
(__file__, inspect.currentframe().f_lineno)):
return 1
if seqid2pfamidMapFile == "":
seqid2pfamidMapFile = "%s/wk/MPTopo/pfamAna_refpro/pfammap_from_uniprot/Pfam-A-full.seqfrompfamfasta.percentTMpro_scampi.perTM75_nseq20%s.nr100.filter.fragmented.seqid2pfamid" % (
DATADIR3, tag)
if myfunc.checkfile(
seqid2pfamidMapFile, "%s (line %d): seqid2pfamidMapFile" %
(__file__, inspect.currentframe().f_lineno)):
return 1
if pfamid2seqidMapFile == "":
pfamid2seqidMapFile = "%s/wk/MPTopo/pfamAna_refpro/pfammap_from_uniprot/Pfam-A-full.seqfrompfamfasta.percentTMpro_scampi.perTM75_nseq20.nr100.filter.fragmented.pfamid2seqid" % (
DATADIR3)
if myfunc.checkfile(
pfamid2seqidMapFile, "%s (line %d): pfamid2seqidMapFile" %
(__file__, inspect.currentframe().f_lineno)):
return 1
if clanid2seqidMapFile == "":
clanid2seqidMapFile = "%s/wk/MPTopo/pfamAna_refpro/pfammap_from_uniprot/Pfam-A-full.seqfrompfamfasta.percentTMpro_scampi.perTM75_nseq20%s.nr100.filter.fragmented.clanid2seqid" % (
DATADIR3, tag)
if myfunc.checkfile(
clanid2seqidMapFile, "%s (line %d): clanid2seqidMapFile" %
(__file__, inspect.currentframe().f_lineno)):
return 1
if tm_pfamidListFile == "":
tm_pfamidListFile = "%s/data/pfam/pfam26.0/Pfam-A-full.seqfrompfamfasta.percentTMpro_scampi.perTM75_nseq20%s.pfamidlist" % (
DATADIR3, tag)
if myfunc.checkfile(
tm_pfamidListFile, "%s (line %d): tm_pfamidListFile" %
(__file__, inspect.currentframe().f_lineno)):
return 1
if tm_clanidListFile == "":
tm_clanidListFile = "%s/data/pfam/pfam26.0/Pfam-A-full.seqfrompfamfasta.percentTMpro_scampi.perTM75_nseq20.clanidlist" % (
DATADIR3)
if myfunc.checkfile(
tm_clanidListFile, "%s (line %d): tm_clanidListFile" %
(__file__, inspect.currentframe().f_lineno)):
return 1
if dbname_predTM == "":
dbname_predTM = "%s/wk/MPTopo/pfamAna_refpro/pred_topcons_single_method4/refpro20120604-celluar.selmaxlength-m1.topcons-single_topcons_single.m1.agree-44.RMSP" % (
DATADIR3)
if myfunc.checkfile(
"%s0.db" % (dbname_predTM), "%s (line %d): dbname_predTM" %
(__file__, inspect.currentframe().f_lineno)):
return 1
if g_params['isOnlyAnaProkar']:
prokarseqidfile = "%s/data/uniprot/reference_proteome/refpro20120604-celluar.selmaxlength-m1.nr100.filter.fragmented.Prokaryota.seqidlist" % (
DATADIR3)
g_params['prokarSeqIDSet'] = set(myfunc.ReadIDList(prokarseqidfile))
if len(g_params['prokarSeqIDSet']) < 1:
return 1
if g_params['isOnlyAnaEukar']:
eukarseqidfile = "%s/data/uniprot/reference_proteome/refpro20120604-celluar.selmaxlength-m1.nr100.filter.fragmented.Eukaryota.seqidlist" % (
DATADIR3)
g_params['eukarSeqIDSet'] = set(myfunc.ReadIDList(eukarseqidfile))
if len(g_params['eukarSeqIDSet']) < 1:
return 1
if pairlistwithpfamidFile == "":
pairlistwithpfamidFile = "%s/../../Pfam-.maxpair100.pairlistwithpfamid" % (
dirpath)
if myfunc.checkfile(
pairlistwithpfamidFile, "%s (line %d): pairlistwithpfamidFile" %
(__file__, inspect.currentframe().f_lineno)):
return 1
pfamid_2_seqidpair_Dict = ReadPairListWithFamID(pairlistwithpfamidFile)
usedPfamIDSet = set(
pfamid_2_seqidpair_Dict.keys()) # pfamids used in pair selection
if pairListFile != "":
li = myfunc.ReadPairList(pairListFile)
SPE_PAIR_LIST = []
for tup in li:
SPE_PAIR_LIST.append((int(tup[0]), int(tup[1])))
(pfamidDefDict, clanidDefDict) = ReadPfamDefFile(pfamDefFile)
signalpDict = lcmp.ReadSignalPDict(signalpFile)
seqid2clanidDict = myfunc.ReadFam2SeqidMap(seqid2clanidMapFile)
seqid2pfamidDict = myfunc.ReadFam2SeqidMap(seqid2pfamidMapFile)
clanid2seqidDict = myfunc.ReadFam2SeqidMap(clanid2seqidMapFile)
pfamid2seqidDict = myfunc.ReadFam2SeqidMap(pfamid2seqidMapFile)
tm_pfamidList = myfunc.ReadIDList(tm_pfamidListFile)
tm_clanidList = myfunc.ReadIDList(tm_clanidListFile)
tm_pfamidSet = set(tm_pfamidList)
tm_clanidSet = set(tm_clanidList)
hdl_predTM = myfunc.MyDB(dbname_predTM)
if not hdl_predTM.failure:
idSet_TMpro = set(hdl_predTM.indexedIDList)
else:
idSet_TMpro = set([])
#classList_TableNumTMHeatMap = ["ALL", "RMSP", "RMDUP"]
#alignrangeList = ['FULL_ALIGNED', 'all', 'PART_ALIGNED']
alignrangeList = ['FULL_ALIGNED']
if g_params['outpath'] != "" and not os.path.exists(g_params['outpath']):
cmd = ["mkdir", "-p", g_params['outpath']]
try:
subprocess.check_call(cmd)
except subprocess.CalledProcessError, e:
print e
return 1
def main(g_params): #{{{
argv = sys.argv
numArgv = len(argv)
if numArgv < 2:
PrintHelp()
return 1
outfile = ""
fileListFile = ""
fileList = []
pfamDefFile = "%s/data/pfam/pfam26.0/Pfam-A.clans.tsv" % (DATADIR3)
threshold_Fraction_Group_2 = 0.05
threshold_NumSeq_Group_2 = 2
tableinfoFile = ""
pdbtospFile = ""
sprotACListFile = ""
threshold_g12_seqidt = 20.0
topoalnFile = ""
aapath = ""
i = 1
isNonOptionArg = False
while i < numArgv:
if isNonOptionArg == True:
fileList.append(argv[i])
isNonOptionArg = False
i += 1
elif argv[i] == "--":
isNonOptionArg = True
i += 1
elif argv[i][0] == "-":
if argv[i] in ["-h", "--help"]:
PrintHelp()
return 1
elif argv[i] in ["-o", "--o", "-outfile"]:
(outfile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-l", "--l"]:
(fileListFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-seqidttype", "--seqidttype"]:
(g_params['seqidttype'], i) = myfunc.my_getopt_int(argv, i)
elif argv[i] in ["-tableinfo", "--tableinfo"]:
(tableinfoFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-topoaln", "--topoaln"]:
(topoalnFile, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-aapath", "--aapath"]:
(aapath, i) = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-thncls2", "--thncls2"]:
(threshold_NumSeq_Group_2, i) = myfunc.my_getopt_int(argv, i)
elif argv[i] in ["-thfrac2", "--thfrac2"]:
(threshold_Fraction_Group_2,
i) = myfunc.my_getopt_float(argv, i)
elif argv[i] in ["-pfamdef", "--pfamdef"]:
(pfamDefFile, i) = myfunc.my_getopt_str(argv, i)
elif (argv[i] in ["-pdbtosp", "--pdbtosp"]):
pdbtospFile, i = myfunc.my_getopt_str(argv, i)
elif (argv[i] in ["-sprot", "--sprot"]):
sprotACListFile, i = myfunc.my_getopt_str(argv, i)
elif argv[i] in ["-q", "--q"]:
g_params['isQuiet'] = True
i += 1
else:
print >> sys.stderr, "Error! Wrong argument:", argv[i]
return 1
else:
fileList.append(argv[i])
i += 1
if fileListFile != "":
fileList += myfunc.ReadIDList(fileListFile)
if len(fileList) < 1:
print >> sys.stderr, "No input set. exit"
return 1
if myfunc.checkfile(topoalnFile, "topoalnFile") != 0:
return 1
if myfunc.checkfile(aapath, "aapath") != 0:
return 1
if outfile == "":
print >> sys.stderr, "outfile not set. Exit"
return 1
outpath = myfunc.my_dirname(outfile)
if not os.path.exists(outpath):
cmd = ["mkdir", "-p", outpath]
try:
subprocess.check_output(cmd)
except subprocess.CalledProcessError, e:
print e
return 1
for tt in fracList:
fpout.write(" %7.2f" % (tt * 100))
fpout.write("\n")
myfunc.myclose(fpout)
print "file %s output" % (outfile1)
# make plot
cmd = ["%s/plotMaxFracFamilyWithTopoVariation.sh" %
(binpath)] + outfileList
try:
subprocess.check_output(cmd)
except subprocess.CalledProcessError, e:
print e
# output selected pairlist and draw pairwise topology alignment for selected
# pairs
outpath = myfunc.my_dirname(outfile)
ext_topoaln = ".topoaln.fa"
outfile_selected_pair = outfile + ".selected.pairlistwithfamid"
fpout = myfunc.myopen(outfile_selected_pair, None, "w", True)
print >> fpout, "#%-6s %6s %6s %7s %15s %10s %10s %6s %4s %4s %5s %5s" % (
"seqid1", "seqid2", "seqidt", "famid", "pfamdef", "numSeqCls1",
"numSeqCls2", "numSeq", "nTM1", "nTM2", "isSP", "isPDB")
for li in selectedPairList:
print >> fpout, "%-7s %6s %6.1f %7s %15s %10d %10d %6d %4d %4d %5d %5d" % (
li[0], li[1], li[2], li[3], li[4], li[5], li[6], li[7], li[8],
li[9], li[10], li[11])
myfunc.myclose(fpout)
cmd = [
"%s/selectPairaln.py" % (binpath), "-pairaln", topoalnFile, "-l",
outfile_selected_pair, "-split", "-outpath", outpath, "-ext",
cntf = 0
while i < numInput:
splitList = []
cntscore = 0
j = 0
while i+j < numInput:
pair = inputList[i+j]
j += 1
seqfilename = pair[0]
numseq = pair[1]
if numseq >= 2: # There should be at least two sequences to run msa
rootname = os.path.basename(os.path.splitext(seqfilename)[0])
if outpath != "":
cur_outpath = outpath
else:
cur_outpath = myfunc.my_dirname(seqfilename)
gzfile = cur_outpath + os.sep + "%s.%s.mfa.gz"%(rootname, prog)
msafile = cur_outpath + os.sep + "%s.%s.mfa"%(rootname, prog)
if ((g_params['isGzip'] == True and not os.path.exists(gzfile)) or
(g_params['isGzip'] == False and not os.path.exists(msafile))):
cntscore += pair[1]**2
splitList.append(pair[0])
if cntscore > threshold:
break
else:
print "%s already exist. Ignore"%(gzfile);
if len(splitList) > 0:
splitidlistfile = workdir + os.sep \
+ "splitidlist.%s.%d.idlist"%(prog, cntf)
splitscriptfile = workdir + os.sep \
def Build_seqid2pfamid(infile, g_outpath): #{{{
outpath = ""
dirname_infile = myfunc.my_dirname(infile)
if g_outpath != "":
outpath = g_outpath
else:
outpath = dirname_infile
rootname = os.path.basename(os.path.splitext(infile)[0])
domainfile = "%s/%s.domainlistperseq" % (outpath, rootname)
seqid2pfamidfile = "%s/%s.seqid2pfamid" % (outpath, rootname)
if os.path.exists(domainfile) and os.path.exists(
seqid2pfamidfile) and not g_params['isOverwrite']:
print >> sys.stderr, "result file %s and %s exist. Ignore" % (
domainfile, seqid2pfamidfile)
return 1
fpout_domain = myfunc.myopen(domainfile, None, "w", True)
fpout_table = myfunc.myopen(seqid2pfamidfile, None, "w", True)
evalue_threshold = g_params['evalue_threshold']
hdl = myfunc.ReadLineByBlock(infile)
queue = deque([])
if hdl.failure:
return 1
lines = hdl.readlines()
while lines != None:
for line in lines:
if not line or line[0] == "#":
continue
rd = ScanfHmmscanRecord(line)
# print rd
if rd == None:
print >> sys.stderr, "%s: bad record. line=\"%s\"" % (infile,
line)
else:
evalue = rd[2]
if evalue <= evalue_threshold:
queue.append(rd)
# scan queue and output result
idlist = [x[0] for x in queue]
# print idlist
# the top records is complete for one query if there are more than one unique seqids
idlist_unique = myfunc.uniquelist(idlist)
idlist_complete = idlist_unique[:-1] # remove the last item
idset_complete = set(idlist_complete)
if len(idset_complete) > 0:
recordDict = {}
cnt_used_rd = 0
for i in xrange(len(queue)):
seqid = queue[i][0]
if not seqid in idset_complete:
continue
if not seqid in recordDict:
recordDict[seqid] = []
recordDict[seqid].append(queue[i])
cnt_used_rd += 1
#output
for seqid in idlist_complete:
try:
li = recordDict[seqid]
except KeyError:
print "seqid=%s" % (
seqid), "idlist_complete=", idlist_complete
raise
famidlist = [x[1] for x in li]
famidlist = myfunc.uniquelist(famidlist)
fpout_table.write("%s %d" % (seqid, len(famidlist)))
for pfamid in famidlist:
fpout_table.write(" %s" % (pfamid))
fpout_table.write("\n")
fpout_domain.write("%s %d" % (seqid, len(li)))
for rd in li:
fpout_domain.write(" %s,%d,%d" % (rd[1], rd[5], rd[6]))
fpout_domain.write("\n")
# pop up queue
for i in xrange(cnt_used_rd):
queue.popleft()
lines = hdl.readlines()
if len(queue) > 0: # output the last item
seqid = queue[0][0]
li = queue
famidlist = [x[1] for x in li]
famidlist = myfunc.uniquelist(famidlist)
fpout_table.write("%s %d" % (seqid, len(famidlist)))
for pfamid in famidlist:
fpout_table.write(" %s" % (pfamid))
fpout_table.write("\n")
fpout_domain.write("%s %d" % (seqid, len(li)))
for rd in li:
fpout_domain.write(" %s,%d,%d" % (rd[1], rd[5], rd[6]))
fpout_domain.write("\n")
hdl.close()
myfunc.myclose(fpout_domain)
myfunc.myclose(fpout_table)
return 0
