''' Takes the file name with fasta sequences and make hmm, returns 0

'''

align_command = "clustalo --outfmt=clu -i " + fastafile + " > " + modelname + ".clustal"

print "Aligning sequences with ClustalO..."

os.system(align_command)

make_hmm_command = "hmmbuild -n " + modelname + " " + modelname + ".hmm " + modelname + ".clustal"

print "Building hmm..."

''' Takes hmm name and performs search through proteomes. The directory with proteomes is proteomes_dir assigned in Configuration section

'''

if out_filename == None:

out_filename = modelname + ".hits"

modelname = "data/" + modelname + ".hmm"

hmm_search_command = "hmmsearch --noali --tblout " + out_filename + " " + modelname + " " + proteomes_dir + organism + ".fasta"

print "Searching in all proteomes with built hmm ..."

''' Reads hmmer hits from standart hmmer output.

Returns hits with bitscore more than provided max bitscore percentage.

By default returns only hits that from unique loci.

'''

gene_regex = re.compile(r'GN=([^=]+)')

try:

hits = SearchIO.read(modelname + ".hits", "hmmer3-tab")

except:

hits = []

# select the longest isoform

filtered_hits = []

try:

max_bitscore = hits[0].bitscore

except:

max_bitscore = 0

unique_genes = []

na_gene_count = 0

for h in hits:

# try to guess gene from hit description

gene = None

for m in gene_regex.finditer(h.description):

gene = m.group(1)[:-3]

if not gene:

gene = "gene" + str(na_gene_count)

na_gene_count += 1

#print h.id, gene, na_gene_count

if gene not in unique_genes:

unique_genes.append(gene)

if h.bitscore > max_bitscore * forw_inc_bitscore_percentage:

filtered_hits.append(h)

if manual_mode:

print unique_genes

# raw_input("...")

''' Fetches the sequence from the proteome file.

Needs full sequence name in Uniprot format like sp|Q15811|ITSN1_HUMAN

'''

seq_id = full_seq_name.split("|")[2]

if not organism:

# if organism is not set, try to guess from seq name

organism = full_seq_name.split("|")[2].split("_")[1]

seq_filename = seq_id + ".fasta"

fetch_command = "esl-sfetch " + proteomes_dir + organism + ".fasta \"" + full_seq_name + "\" > " + seq_filename

try:

os.system(fetch_command)

return seq_filename

except:

print "Sequence was not found in the proteome of " + organism

''' Runs pfam_scan

'''

pfam_command = "pfam_scan.pl -fasta " + seq_filename + " -dir data > " + out_filename

# print pfam_command

try:

os.system(pfam_command)

return 0

except:

''' Runs prosite prediction with ps_scan

'''

prosite_command = "ps_scan.pl -d data/custom.dat -o pff " + seq_filename + " > " + out_filename

# print prosite_command

try:

os.system(prosite_command)

return 0

except:

''' Predicts coiled coil region with ncoils.

Currently results does not correspond to SMART, which also uses ncoils prediction.

On coils server they use ncoils with -nw parameter which is not described and not working both for 1999 and 2002 ver of ncoils.

Giving up as for now.

'''

ncoils_result_file = "ncoils.fasta"

ccr_regex = re.compile(r'x+')

ncoils_command = "ncoils -f -win 28 < " + seq_filename + " > " + ncoils_result_file

os.system(ncoils_command)

# parse ncoils results

ncoils_result_handle = open(ncoils_result_file, "rU")

ncoils_prediction = SeqIO.read(ncoils_result_handle, "fasta")

ncoils_result_handle.close()

out_handle = open(out_filename, "w")

for match in ccr_regex.finditer(str(ncoils_prediction.seq)):

out_handle.write("CCR" + " " + str(match.start()) + " " + str(match.end()) + " " + ncoils_prediction.id + "\n")

''' Predicts short motives using regular expressions

'''

motifs = {

'AP2Binding': re.compile(r'(DP[WF]|F.D.F|WV.F|F..F..[FL])'),

'ClathrinBinding': re.compile(r'(L[FILMV].[FILMV][DE]|L[FILMV].[DE][FILMV])'),

'NPF': re.compile(r'(NPF)'),

}

in_handle = open(seq_filename, "rU")

records = list(SeqIO.parse(in_handle, "fasta"))

in_handle.close()

out_handle = open(out_filename, "w")

for r in records:

for motif_name, motif_regex in motifs.iteritems():

for match in motif_regex.finditer(str(r.seq)):

out_handle.write(motif_name + " " + str(match.start()) + " " + str(match.end()) + " " + r.id + "\n")

''' Parses the file with coordinates of the predicted features

Type could be ncoils, prosite, pfam, regex.

'''

features_array = []

in_handle = open(coord_file, "rU")

for l in in_handle.readlines():

# if the line is not commented

if l[0] != "#" and len(l) > 1:

splitted = l.split()

if type == "pfam":

seq_name_index = 0

f_start_index = 3

f_end_index = 4

f_name_index = 6

elif type == "prosite":

f_name_index = 3

f_start_index = 1

f_end_index = 2

seq_name_index = 0

else:

# currently no differences with prosite

f_name_index = 0

f_start_index = 1

f_end_index = 2

seq_name_index = 3

try:

seq_name = splitted[seq_name_index].split("|")[2]

except:

seq_name = splitted[seq_name_index]

f_start = splitted[f_start_index]

f_end = splitted[f_end_index]

f_name = splitted[f_name_index]

features_array.append([f_name, f_start, f_end, seq_name])

in_handle.close()

''' Performs match of the sequence to the domain set.

Domains is the dictionary { "SH3": 1, "EF-hand": 1, "PH": 0} where domain has 1 if it is crucial for the protein and less otherwise. Domain names should correspond to the PFAM hmm. Motifs also could be included.

'''

pfam_coords_file = "pfam_coords"

prosite_coords_file = "prosite_coords"

ncoils_coords_file = "ncoils_coords"

regex_coords_file = "regex_coords"

make_pfam_prediction(seq_filename, pfam_coords_file)

make_prosite_prediction(seq_filename, prosite_coords_file)

make_ncoils_prediction(seq_filename, ncoils_coords_file)

make_regex_prediction(seq_filename, regex_coords_file)

features = []

features.extend(parse_coord_file(pfam_coords_file, "pfam"))

features.extend(parse_coord_file(prosite_coords_file, "prosite"))

features.extend(parse_coord_file(ncoils_coords_file, "ncoils"))

features.extend(parse_coord_file(regex_coords_file, "regex"))

f_dict = {}

for f in features:

if f[3] not in f_dict.keys():

f_dict[f[3]] = []

f_dict[f[3]].append(f[:3])

final_score = 0

print f_dict

print f_dict.keys()

# raw_input("...")

for seq_name, f in f_dict.iteritems():

not_found = ""

found = ""

seq_features = [ff[0].split("_")[0] for ff in f]

seq_features_set = set(seq_features)

score = 0

for d in domains:

if d in seq_features_set:

score += domains[d]

found += " " + str(score) + " " + d

else:

not_found += " " + d

if manual_mode:

print "Domains found in " + seq_filename + ": " + found

raw_input("Check the results of domain match...\n")

final_score = score / sum(domains.values())

''' Performs reciprocal hmmer search against the proteome of given organism

'''

print "# Reciprocal search..."

is_found = False

if out_filename == None:

out_filename = modelname + ".rechits_" + organism

reciprocal_search_command = "phmmer --noali --tblout " + out_filename + " " + filename + " " + proteomes_dir + organism + ".fasta > hmmer_res"

os.system(reciprocal_search_command)

try:

hits = SearchIO.read(out_filename, "hmmer3-tab")

max_bitscore = hits[0].bitscore

except:

hits = []

max_bitscore = 0

if len(hits) > 0:

if re.search(modelname_regex, hits[0].description):

is_found = True

# for h in hits:

# if h.bitscore > rev_inc_bitscore_percentage * max_bitscore:

# if manual_mode:

# print modelname_regex, h.description, re.search(modelname_regex, h.description)

# raw_input("...")

# if re.search(modelname_regex, h.description):

# is_found = True

if manual_mode:

print filename, is_found

raw_input("Check reciprocal search results...")

# "LEIBR", "LEIIN", "LEIMA", "TRYB2", "TRYCI", "TRYCC", "TRYVY", "THETR",

# "FONAL", "TRIVA", "NAEGR",

"GIAIB",

# "PLAF7", "THEPA", "CRYMR", "EIMAC",

# "EIMMA", "TOXGO", "GRENI", "TETTS", "PARTE", "PERM5", "THEAN", "BABBO",

# "NEOCL", "PHATC", "THAOC", "THAPS", "BLAHO", "HYAAE", "PHYIT", "PHYRM",

# "PHYSP",

# "AURAN", "ECTSI", "PYTUL", "EMIHU", "RETFI", "CHLRE", "CHLVA", "ARATH",

# "AUXPR", "SOLTU", "VITVI", "SOLLC", "SOYBN", "GALSU", "RICCO", "OSTTA",

# "WHEAT", "BRAOL", "EUTSA", "ORYSJ", "AMBTC", "SELML", "MEDTR", "OSTLU",

# "BRADI", "VOLCA",

# "ENTHI", "ENTDS", "DICPU", "DICDI", "ENTNP", "POLPA", "EDHAE",

# "VITCO", "SPRLO", "VAVCU",

# "BATDJ", "ZYMTI", "YARLI", "MAGO7", "GROCL", "OGAPD",

# "ARTOC", "GIBM7", "FUSO4", "BEAB2", "PARBP", "ASPNA", "CHATD", "NECH7",

# "TALSN", "COLSU", "BOTF1", "PENMQ", "PICPG", "CANAL", "YEAST", "MIXOS",

# "PHACS", "RHOT1", "RHOG2", "CRYNJ", "FIBRA", "WALI9", "CRYNH", "USTMA",

# "MALS4", "GLOTA", "RHIO9", "MUCC1", "CAPO3", "SPHAR", "SALR5", "MONBE",

# "CLOSI", "SCHMA", "TRIAD", "ONCVO", "TRISP", "PRIPA", "LOALO", "TRITR",

# "CAEEL", "PEDHC", "APIME", "DROME", "BOMMO", "CULQU", "SOLIN", "DENPD",

# "NASVI", "IXOSC", "STRPU", "AMPQE", "THEKT", "NEMVE", "LOTGI", "CRAGI",

# "CIOIN", "BRAFL", "OIKDI", "DANRE", "HETGA", "LEPOC", "TETNG", "ORYLA",

# "ASTMX", "ORENI", "GASAC", "CHEMY", "ANOCA", "ANAPL", "FICAL", "XENTR",

# "CHICK", "CRIGR", "MOUSE", "HUMAN"