Source code for pydna.genbankfixer

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
# Copyright 2013-2020 by Björn Johansson.  All rights reserved.
# This code is part of the pydna distribution and governed by its
# license.  Please see the LICENSE.txt file that should have been included
# as part of this package.
# doctest: +NORMALIZE_WHITESPACE
# doctest: +SKIP
# doctest: +IGNORE_EXCEPTION_DETAIL
"""This module provides the :func:`gbtext_clean` function which can clean up broken Genbank files enough to
pass the BioPython Genbank parser

Almost all of this code was lifted from BioJSON (https://github.com/levskaya/BioJSON) by Anselm Levskaya.
The original code was not accompanied by any software licence. This parser is based on pyparsing.

There are some modifications to deal with fringe cases.

The parser first produces JSON as an intermediate format which is then formatted back into a
string in Genbank format.

The parser is not complete, so some fields do not survive the roundtrip (see below).
This should not be a difficult fix. The returned result has two properties,
.jseq which is the intermediate JSON produced by the parser and .gbtext
which is the formatted genbank string."""


import re as _re
import pyparsing as _pp

GoodLocus = (
    _pp.Literal("LOCUS")
    + _pp.Word(_pp.alphas + _pp.nums + "-_()." + "\\").setResultsName("name")
    + _pp.Word(_pp.nums).setResultsName("size")
    + _pp.Suppress(_pp.CaselessLiteral("bp"))
    + _pp.Word(_pp.alphas + "-").setResultsName("seqtype")
    + (_pp.CaselessLiteral("linear") | _pp.CaselessLiteral("circular")).setResultsName(
        "topology"
    )
    + _pp.Optional(_pp.Word(_pp.alphas), default="   ").setResultsName("divcode")
    + _pp.Regex(r"(\d{2})-(\S{3})-(\d{4})").setResultsName("date")
)

# Older versions of ApE don't include a LOCUS name! Need separate def for this case:
BrokenLocus1 = (
    _pp.Literal("LOCUS").setResultsName("name")
    + _pp.Word(_pp.nums).setResultsName("size")
    + _pp.Suppress(_pp.CaselessLiteral("bp"))
    + _pp.Word(_pp.alphas + "-").setResultsName("seqtype")
    + (_pp.CaselessLiteral("linear") | _pp.CaselessLiteral("circular")).setResultsName(
        "topology"
    )
    + _pp.Optional(_pp.Word(_pp.alphas), default="   ").setResultsName("divcode")
    + _pp.Regex(r"(\d{2})-(\S{3})-(\d{4})").setResultsName("date")
)

# LOCUS       YEplac181	5741 bp 	DNA	SYN
BrokenLocus2 = (
    _pp.Literal("LOCUS")
    + _pp.Word(_pp.alphas + _pp.nums + "-_()." + "\\").setResultsName("name")
    + _pp.Word(_pp.nums).setResultsName("size")
    + _pp.Suppress(_pp.CaselessLiteral("bp"))
    + _pp.Word(_pp.alphas + "-").setResultsName("seqtype")
    + _pp.Optional(
        _pp.CaselessLiteral("linear") | _pp.CaselessLiteral("circular"),
        default="linear",
    ).setResultsName("topology")
    + _pp.Optional(_pp.Word(_pp.alphas), default="   ").setResultsName("divcode")
    + _pp.Regex(r"(\d{2})-(\S{3})-(\d{4})").setResultsName("date")
)

BrokenLocus3 = (
    _pp.Literal("LOCUS")
    + _pp.Word(_pp.alphas + _pp.nums + "-_()." + "\\").setResultsName("name")
    + _pp.Word(_pp.nums).setResultsName("size")
    + _pp.Suppress(_pp.CaselessLiteral("bp"))
    + _pp.Word(_pp.alphas + "-").setResultsName("seqtype")
    + _pp.Optional(
        _pp.CaselessLiteral("linear") | _pp.CaselessLiteral("circular"),
        default="linear",
    ).setResultsName("topology")
    + _pp.Word(_pp.alphas).setResultsName("divcode")
    + _pp.Optional(
        _pp.Regex(r"(\d{2})-(\S{3})-(\d{4})").setResultsName("date"),
        default="19-MAR-1970",
    ).setResultsName("date")
)

LocusEntry = GoodLocus | BrokenLocus1 | BrokenLocus2 | BrokenLocus3

# ===============================================================================
# Generic Entry

# this catches everything but the FEATURES and SEQUENCE entries, really should add parsing code for
# ACCESSION, COMMENTS, REFERENCE, ORGANISM, etc.
# (Though these entries are generally useless when it comes to hacking on DNA)

# All entries in a genbank file headed by an all-caps title with no space between start-of-line and title
CapWord = _pp.Word("ABCDEFGHIJKLMNOPQRSTUVWXYZ")
# after titled line, all subsequent lines have to have at least one space in front of them
# this is how we split up the genbank record
SpacedLine = _pp.White(min=1) + _pp.CharsNotIn("\n") + _pp.LineEnd()
# HeaderLine = CapWord + CharsNotIn("\n") + LineEnd()
GenericEntry = _pp.Group(
    CapWord
    + _pp.Combine(_pp.CharsNotIn("\n") + _pp.LineEnd() + _pp.ZeroOrMore(SpacedLine))
).setResultsName("generics", listAllMatches=True)


# ===============================================================================
# Definition Entry
# SuppressedSpacedLine =  Suppress(White(min=1)) + CharsNotIn("\n") + LineEnd()
# DefinitionEntry =  Suppress(Literal("DEFINITION")) + Combine(CharsNotIn("\n") + LineEnd() + ZeroOrMore( SuppressedSpacedLine ))

# ===============================================================================
# GenBank Feature Table Parser

# ==== Genbank Location String Parser
#
# a string of slices w. functional modifiers that go at most two levels deep
# single position is just a number i.e. 23423
# slice is N1..N2  w. N1<N2
# i.e.
# 23..88  --> seq[23:89] in python syntax (genbank uses inclusive slicing)
# 234..555
# complement(234..434) --> rc(seq[234:435])
# join(23..343,454..666,777..999) --> seq[23:344]+seq[454:667]+seq[777:1000]
# complement(join(23..343,454..666,777..999))
# join(complement(34..123),complement(333..565))
#
# additionally the slices can have ambiguous locs like <454..999 or 232..>331
# also note the dumb 34.38 fuzzy slice notation
# i.e. <45..900  says the real feature starts "somewhere" before pos 45
#       45.48 says feature somewhere between bases 45->48
# lot of weird annotations best avoided because they connote ~useless knowledge for synthetic design
#
# if you don't know where something is, don't use it or guess and move on

LPAREN = _pp.Suppress("(")
RPAREN = _pp.Suppress(")")
SEP = _pp.Suppress(_pp.Literal(".."))

# recognize numbers w. < & > uncertainty specs, then strip the <> chars to make it fixed
gbIndex = _pp.Word(_pp.nums + "<>").setParseAction(
    lambda s, l, t: int(t[0].replace("<", "").replace(">", ""))
)
SimpleSlice = _pp.Group(gbIndex + SEP + gbIndex) | _pp.Group(gbIndex).setParseAction(
    lambda s, l, t: [[t[0][0], t[0][0]]]
)

# recursive def for nested function syntax:  f( g(), g() )
complexSlice = _pp.Forward()
complexSlice << (_pp.Literal("complement") | _pp.Literal("join")) + LPAREN + (
    _pp.delimitedList(complexSlice) | _pp.delimitedList(SimpleSlice)
) + RPAREN
featLocation = _pp.Group(SimpleSlice | complexSlice)


[docs]def parseGBLoc(s, l, t): """retwingles parsed genbank location strings, assumes no joins of RC and FWD sequences """ strand = 1 locationlist = [] # see if there are any complement operators for entry in t[0]: if entry == "complement": strand = -1 for entry in t[0]: if type(entry) != type("string"): locationlist.append([entry[0], entry[1]]) # return locationlist and strand spec return [["location", locationlist], ["strand", strand]]
featLocation.setParseAction(parseGBLoc) # ==== Genbank Feature Key-Value Pairs
[docs]def strip_multiline(s, l, t): whitespace = _re.compile("[\n]{1}[ ]+") return whitespace.sub(" ", t[0])
[docs]def toInt(s, l, t): return int(t[0])
# Quoted KeyVal: /key="value" QuoteFeaturekeyval = _pp.Group( _pp.Suppress("/") + _pp.Word(_pp.alphas + _pp.nums + "_-") + _pp.Suppress("=") + _pp.QuotedString('"', multiline=True).setParseAction(strip_multiline) ) # UnQuoted KeyVal: /key=value (I'm assuming it doesn't do multilines this way? wrong! ApE does store long labels this way! sigh.) # NoQuoteFeaturekeyval = Group(Suppress('/') + Word(alphas+nums+"_-") + Suppress('=') + OneOrMore(CharsNotIn("\n")) ) keyvalspacedline = ( _pp.White(exact=21) + _pp.CharsNotIn("/") + _pp.OneOrMore(_pp.CharsNotIn("\n")) + _pp.LineEnd() ) NoQuoteFeaturekeyval = _pp.Group( _pp.Suppress("/") + _pp.Word(_pp.alphas + _pp.nums + "_-") + _pp.Suppress("=") + _pp.Combine( _pp.CharsNotIn("\n") + _pp.LineEnd() + _pp.ZeroOrMore(keyvalspacedline) ) ) # Special Case for Numerical Vals: /bases=12 OR /bases="12" NumFeaturekeyval = _pp.Group( _pp.Suppress("/") + _pp.Word(_pp.alphas + _pp.nums + "_-") + _pp.Suppress("=") + (_pp.Suppress('"') + _pp.Word(_pp.nums).setParseAction(toInt) + _pp.Suppress('"')) | (_pp.Word(_pp.nums).setParseAction(toInt)) ) # Key Only KeyVal: /pseudo # post-parse convert it into a pair to resemble the structure of the first three cases i.e. [pseudo, True] FlagFeaturekeyval = _pp.Group( _pp.Suppress("/") + _pp.Word(_pp.alphas + _pp.nums + "_-") ).setParseAction(lambda s, l, t: [[t[0][0], True]]) Feature = _pp.Group( _pp.Word(_pp.alphas + _pp.nums + "_-").setParseAction( lambda s, l, t: [["type", t[0]]] ) + featLocation.setResultsName("location") + _pp.OneOrMore( NumFeaturekeyval | QuoteFeaturekeyval | NoQuoteFeaturekeyval | FlagFeaturekeyval ) ) FeaturesEntry = ( _pp.Literal("FEATURES") + _pp.Literal("Location/Qualifiers") + _pp.Group(_pp.OneOrMore(Feature)).setResultsName("features") ) # =============================================================================== # GenBank Sequence Parser # sequence is just a column-spaced big table of dna nucleotides # should it recognize full IUPAC alphabet? NCBI uses n for unknown region Sequence = _pp.OneOrMore( _pp.Suppress(_pp.Word(_pp.nums)) + _pp.OneOrMore(_pp.Word("ACGTacgtNn")) ) # Group( ) hides the setResultsName names def'd inside, such that one needs to first access this group and then access the dict of contents inside SequenceEntry = _pp.Suppress(_pp.Literal("ORIGIN")) + Sequence.setParseAction( lambda s, l, t: "".join(t) ).setResultsName("sequence") # =============================================================================== # Final GenBank Parser # GB files with multiple records split by "//" sequence at beginning of line GBEnd = _pp.Literal("//") # Begin w. LOCUS, slurp all entries, then stop at the end! GB = LocusEntry + _pp.OneOrMore(FeaturesEntry | SequenceEntry | GenericEntry) + GBEnd # NCBI often returns sets of GB files multipleGB = _pp.OneOrMore(_pp.Group(GB)) # =============================================================================== # End Genbank Parser # =============================================================================== # =============================================================================== # Main JSON Conversion Routine
[docs]def strip_indent(str): whitespace = _re.compile("[\n]{1}(COMMENT){0,1}[ ]+") return whitespace.sub("\n", str)
[docs]def concat_dict(dlist): """more or less dict(list of string pairs) but merges vals with the same keys so no duplicates occur """ newdict = {} for e in dlist: if e[0] in newdict.keys(): newdict[e[0]] = newdict[e[0]] + strip_indent(e[1]) else: newdict[e[0]] = strip_indent(e[1]) return newdict
[docs]def toJSON(gbkstring): parsed = multipleGB.parseString(gbkstring) jseqlist = [] for seq in parsed: # for item in seq.asList(): # print(item) # import sys;sys.exit(42) # Print to STDOUT some details (useful for long multi-record parses) # print(seq['name'], ": length:", len(seq['sequence']) , " #features:" , len(seq['features'].asList())) # build JSON object nl = [] if "features" in seq: for a in list(map(dict, seq["features"].asList())): dct = {} for key in a: val = a[key] # print(key, a[key]) dct[key] = a[key] if isinstance(val, str): dct[key] = a[key].strip() nl.append(dct) # import sys;sys.exit(42) # print(list(map(dict, hej))[2]["codon_start"]) jseq = { "__format__": "jseq v0.1", "name": seq["name"], "size": seq["size"], "seqtype": seq["seqtype"], "divcode": seq["divcode"], "date": seq["date"], "topology": seq["topology"], "sequence": seq["sequence"], "features": nl, "annotations": concat_dict(seq["generics"]), } jseqlist.append(jseq) return jseqlist
[docs]def wrapstring(str_, rowstart, rowend, padfirst=True): """ wraps the provided string in lines of length rowend-rowstart and padded on the left by rowstart. -> if padfirst is false the first line is not padded """ rowlen = rowend - rowstart leftpad = rowstart wrappedstr = "" # no wrapping needed, single line if len(str_) / rowlen < 1: if padfirst: return leftpad * " " + str_ + "\n" else: return str_ + "\n" # multiple lines so wrap: for linenum in range(1 + int(len(str_) / rowlen)): if linenum == 0 and not padfirst: wrappedstr += str_[linenum * rowlen : (linenum + 1) * rowlen] + "\n" else: wrappedstr += ( " " * leftpad + str_[linenum * rowlen : (linenum + 1) * rowlen] + "\n" ) # if str_.startswith("/translation="): # print(str_) # print(wrappedstr) # print(".................................") return wrappedstr
[docs]def locstr(locs, strand): "genbank formatted location string, assumes no join'd combo of rev and fwd seqs" # slice format is like: 1..10,20..30,101..200 locstr = ",".join(map((lambda x: str(x[0]) + ".." + str(x[1])), locs)) if len(locs) > 1: locstr = "join(" + locstr + ")" if int(strand) == -1: locstr = "complement(" + locstr + ")" return locstr
[docs]def originstr(sequence): "formats dna sequence as broken, numbered lines ala genbank" wordlen = 10 cols = 6 rowlen = wordlen * cols outstr = "" for linenum in range(int(len(sequence) / rowlen) + 1): pos = linenum * rowlen # position of string for this row, then six blocks of dna outstr += ( (" " * 9 + str(pos + 1))[-9:] + " " + sequence[pos : pos + 10] + " " + sequence[pos + 10 : pos + 20] + " " + sequence[pos + 20 : pos + 30] + " " + sequence[pos + 30 : pos + 40] + " " + sequence[pos + 40 : pos + 50] + " " + sequence[pos + 50 : pos + 60] + "\n" ) return outstr
[docs]def toGB(jseq): "parses json jseq data and prints out ApE compatible genbank" # construct the LOCUS header string # LOCUS format: # Positions Contents # --------- -------- # 00:06 LOCUS # 06:12 spaces # 12:?? Locus name # ??:?? space # ??:40 Length of sequence, right-justified # 40:44 space, bp, space # 44:47 Blank, ss-, ds-, ms- # 47:54 Blank, DNA, RNA, tRNA, mRNA, uRNA, snRNA, cDNA # 54:55 space # 55:63 Blank (implies linear), linear or circular # 63:64 space # 64:67 The division code (e.g. BCT, VRL, INV) # 67:68 space # 68:79 Date, in the form dd-MMM-yyyy (e.g., 15-MAR-1991) name = jseq["name"] or "default" size = jseq["size"] or "100" seqtype = jseq["seqtype"] or "DNA" prefix = "" for p in ["ds-", "ss-", "ms-"]: a, *b = seqtype.split(p) if b: prefix = p seqtype = b.pop() break prefix = prefix or "ds-" topology = jseq["topology"] or "linear" divcode = jseq["divcode"] or " " date = jseq["date"] or "19-MAR-1970" locusstr = "LOCUS {name:<24} {size:>4} bp {prefix}{seqtype:<4} {topology:<8} {divcode} {date}\n".format( name=name, size=size, prefix=prefix, seqtype=seqtype, topology=topology, divcode=divcode, date=date, ) # All these fields are left empty gbprops = ( "DEFINITION .\n" + "ACCESSION \n" + "VERSION \n" + "SOURCE .\n" + "ORGANISM .\n" + "COMMENT \n" + "COMMENT ApEinfo:methylated:1\n" + "FEATURES Location/Qualifiers\n" ) # build the feature table featuresstr = "" if "features" in jseq: for feat in jseq["features"]: fstr = ( " " * 5 + feat["type"] + " " * (16 - len(feat["type"])) + wrapstring(locstr(feat["location"], feat["strand"]), 21, 80, False) ) for k in feat.keys(): if k not in ["type", "location", "strand"]: # ApE idiosyncrasy: don't wrap val in quotation marks if k in [ "ApEinfo_label", "ApEinfo_fwdcolor", "ApEinfo_revcolor", "label", ]: fstr += wrapstring("/" + str(k) + "=" + str(feat[k]), 21, 80) # standard: wrap val in quotes else: fstr += wrapstring( "/" + str(k) + "=" + '"' + str(feat[k]) + '"', 21, 80 ) featuresstr += fstr # the spaced, numbered sequence gborigin = "ORIGIN\n" + originstr(jseq["sequence"]) + "//\n" return locusstr + gbprops + featuresstr + gborigin
[docs]def gbtext_clean(gbtext): """This function takes a string containing **one** genbank sequence in Genbank format and returns a named tuple containing two fields, the gbtext containing a string with the corrected genbank sequence and jseq which contains the JSON intermediate. Examples -------- >>> s = '''LOCUS New_DNA 3 bp DNA CIRCULAR SYN 19-JUN-2013 ... DEFINITION . ... ACCESSION ... VERSION ... SOURCE . ... ORGANISM . ... COMMENT ... COMMENT ApEinfo:methylated:1 ... ORIGIN ... 1 aaa ... //''' >>> from pydna.readers import read >>> read(s) # doctest: +SKIP /home/bjorn/anaconda3/envs/bjorn36/lib/python3.6/site-packages/Bio/GenBank/Scanner.py:1388: BiopythonParserWarning: Malformed LOCUS line found - is this correct? :'LOCUS New_DNA 3 bp DNA CIRCULAR SYN 19-JUN-2013\\n' "correct?\\n:%r" % line, BiopythonParserWarning) Traceback (most recent call last): File "/home/bjorn/python_packages/pydna/pydna/readers.py", line 48, in read results = results.pop() IndexError: pop from empty list <BLANKLINE> During handling of the above exception, another exception occurred: <BLANKLINE> Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/home/bjorn/python_packages/pydna/pydna/readers.py", line 50, in read raise ValueError("No sequences found in data:\\n({})".format(data[:79])) ValueError: No sequences found in data: (LOCUS New_DNA 3 bp DNA CIRCULAR SYN 19-JUN-2013 DEFINITI) >>> from pydna.genbankfixer import gbtext_clean >>> s2, j2 = gbtext_clean(s) >>> print(s2) LOCUS New_DNA 3 bp ds-DNA circular SYN 19-JUN-2013 DEFINITION . ACCESSION VERSION SOURCE . ORGANISM . COMMENT COMMENT ApEinfo:methylated:1 FEATURES Location/Qualifiers ORIGIN 1 aaa // >>> s3 = read(s2) >>> s3 Dseqrecord(o3) >>> print(s3.format()) LOCUS New_DNA 3 bp DNA circular SYN 19-JUN-2013 DEFINITION . ACCESSION New_DNA VERSION New_DNA KEYWORDS . SOURCE ORGANISM . . COMMENT ApEinfo:methylated:1 FEATURES Location/Qualifiers ORIGIN 1 aaa //""" jseqlist = toJSON(gbtext) jseq = jseqlist.pop() from collections import namedtuple as _namedtuple from pydna._pretty import pretty_str as _pretty_str Result = _namedtuple("Result", "gbtext jseq") result = Result(_pretty_str(toGB(jseq).strip()), jseq) return result
if __name__ == "__main__": import os as _os cached = _os.getenv("pydna_cached_funcs", "") _os.environ["pydna_cached_funcs"] = "" import doctest doctest.testmod(verbose=True, optionflags=doctest.ELLIPSIS) _os.environ["pydna_cached_funcs"] = cached