def run_annotation(infile,
prefix=None,
ident=70,
threads=4,
kingdom='bacteria',
trusted=None,
trusted_format='uniprot',
callback=None,
**kwargs):
"""
Annotate nucelotide sequences (usually a draft assembly with contigs)
using prodigal and blast to prokka seqs. Writes a genbank file to the
same folder.
Args:
infile: input fasta file
prefix: optional prefix for locus_tags
ident: minimum percent identity for blast results, default 70
threads: cpu threads to use
kingdom: 'bacteria', 'viruses' or 'archaea'
trusted: optional fasta file of trusted protein sequences for blast
returns:
a dataframe of annotations and a list of SeqRecords with the features, one per contig
"""
#get simple name for contig
def get_contig(x):
return ('_').join(x.split('_')[:-1])
if prefix == None:
prefix = create_locus_tag(infile)
sprot = 'sprot_%s' % kingdom
dbs = ['IS', 'amr', trusted, sprot]
evalues = [1e-10, 1e-100, 1e-4, 1e-4]
#run prodigal
resfile = prodigal(infile)
#read in prodigal fasta to dataframe
df = tools.fasta_to_dataframe(resfile)
df[['start', 'end',
'strand']] = df.description.apply(get_prodigal_coords, 1)
df['feat_type'] = 'CDS'
df['contig'] = df['name'].apply(get_contig)
df['sequence'] = df.sequence.str.rstrip('*')
#get target seqs
seqs = list(SeqIO.parse(resfile, 'fasta'))
#remove temp prodigal file
os.remove(resfile)
#remove trailing asterisks
#seqs = [s.rstrip("*") for s in seqs]
#read input file nucleotide seqs
contigs = SeqIO.to_dict(SeqIO.parse(infile, 'fasta'))
#print (df[:5])
res = []
i = 0
for db in dbs:
if db is None:
i += 1
continue
fetch_sequence_from_url(db, path=prokkadbdir)
#make blast db of prokka proteins
if db in ['IS', 'amr', sprot]:
dbname = os.path.join(prokkadbdir, '%s.fa' % db)
else:
dbname = db
tools.make_blast_database(dbname, dbtype='prot')
print('blasting %s ORFs to %s' % (len(seqs), db))
bl = tools.blast_sequences(dbname,
seqs,
maxseqs=1,
evalue=evalues[i],
cmd='blastp',
show_cmd=True,
threads=threads,
**kwargs)
bl = bl[bl.pident > ident]
#print (bl)
if len(bl) == 0:
i += 1
continue
#get hit info from header (prokka format)
try:
bl[['protein_id', 'gene', 'product',
'cog']] = bl.stitle.apply(prokka_header_info, 1)
except:
bl[['protein_id', 'gene', 'product',
'cog']] = bl.stitle.apply(uniprot_header_info, 1)
cols = [
'qseqid', 'sseqid', 'pident', 'sstart', 'send', 'protein_id',
'gene', 'product'
]
bl = bl.sort_values(['qseqid', 'pident'],
ascending=False).drop_duplicates(['qseqid'])[cols]
#print (len(bl))
#merge blast result with prodigal sequences
found = df.merge(bl, left_on='name', right_on='qseqid', how='right')
#get remaining sequences with no hits to this db
df = df[~df.name.isin(bl.qseqid)]
#new sequences to blast in the next iteration
seqs = tools.dataframe_to_seqrecords(df, idkey='name')
#print (found)
res.append(found)
print('%s sequences unassigned' % len(df))
i += 1
#all results together
res = pd.concat(res)
#print (res)
#-------------------------------------------------------
#run hmmer on unassigned
if kingdom == 'bacteria':
print('running hmmer')
#write unknowns out
SeqIO.write(seqs, 'unknowns.fa', 'fasta')
hmmdf = hmmer('unknowns.fa', threads=threads)
if hmmdf is not None:
res = pd.concat([res, hmmdf], sort=False)
#get tRNAs with aragorn
print('running aragorn')
arag = aragorn(infile)
#print (arag)
res = pd.concat([res, arag], sort=False)
#remaining unknowns are hypothetical proteins
unknown = df[~df.name.isin(res.name)].copy()
unknown['product'] = 'hypothetical protein'
res = pd.concat([res, unknown], sort=False)
#post process dataframe
#res = res.fillna('')
res['translation'] = res.sequence
res['length'] = res.sequence.str.len()
#res['gene'] = res.gene.fillna('')
res = res.reset_index(drop=True)
#print (res['product'].value_counts())
#print (res.dtypes)
#-------------------------------------------------------
#we then write all found sequences to seqrecord/features
l = 1 #counter for assigning locus tags
recs = []
#print (res.iloc[])
#group by contig and get features for each protein found
for c, df in res.groupby('contig'):
contig = get_contig(c)
#truncated label for writing to genbank
label = ('_').join(c.split('_')[:2])
#print (c, len(df), label)
nucseq = contigs[c].seq
rec = SeqRecord(nucseq, annotations={"molecule_type": "DNA"})
#rec.seq.alphabet = generic_dna
rec.id = label
rec.name = label
rec.COMMENT = 'annotated with pathogenie'
df = df.sort_values('start')
qcols = ['gene', 'product', 'locus_tag', 'translation', 'length']
for i, row in df.iterrows():
row['locus_tag'] = '{p}_{l:05d}'.format(p=prefix, l=l)
row = row.dropna()
cols = [c for c in qcols if c in row.index]
quals = row[cols].to_dict()
#print (quals)
feat = SeqFeature(FeatureLocation(row.start, row.end, row.strand),
strand=row.strand,
type=row.feat_type,
qualifiers=quals)
rec.features.append(feat)
l += 1
recs.append(rec)
print('done')
return res, recs
