view shinefind.py @ 0:02569adc3290 draft default tip

Uploaded

#!/usr/bin/env python
import re
import sys
import argparse
import logging
from CPT_GFFParser import gffParse, gffWrite, gffSeqFeature
from Bio import SeqIO
from Bio.SeqRecord import SeqRecord
from Bio.SeqFeature import FeatureLocation
from gff3 import (
    feature_lambda,
    feature_test_type,
    feature_test_true,
    feature_test_quals,
    get_id,
    ensure_location_in_bounds,
)

logging.basicConfig(level=logging.INFO)
log = logging.getLogger()


class NaiveSDCaller(object):

    # TODO May make switch for different sequence sets
    SD_SEQUENCES = (
        "AGGAGGT",
        "GGAGGT",
        "AGGAGG",
        "GGGGGG",
        "AGGAG",
        "GAGGT",
        "GGAGG",
        "GGGGG",
        "AGGT",
        "GGGT",
        "GAGG",
        "GGGG",
        "AGGA",
        "GGAG",
        "GGA",
        "GAG",
        "AGG",
        "GGT",
        "GGG",
    )

    def __init__(self):
        self.sd_reg = [re.compile(x, re.IGNORECASE) for x in self.SD_SEQUENCES]

    def list_sds(self, sequence, sd_min=3, sd_max=17):
        hits = []
        for regex in self.sd_reg:
            for match in regex.finditer(sequence):
                spacing = len(sequence) - len(match.group()) - match.start()
                if sd_max >= spacing+sd_min and spacing+sd_min >= sd_min:
                    #if the spacing is within gap limits, add 
                    #(search space is [sd_max+7 .. sd_min] so actual gap is spacing+sd_min)
                    #print('min %d max %d - adding SD with gap %d' % (sd_min, sd_max, spacing+sd_min))
                    hits.append(
                        {
                            "spacing": spacing,
                            "hit": match.group(),
                            "start": match.start(),
                            "end": match.end(),
                            "len": len(match.group()),
                        }
                    )
        hits = sorted(hits, key= lambda x: (-x['len'],x['spacing']))
        return hits

    @classmethod
    def highlight_sd(cls, sequence, start, end):
        return " ".join(
            [
                sequence[0:start].lower(),
                sequence[start:end].upper(),
                sequence[end:].lower(),
            ]
        )

    @classmethod
    def to_features(cls, hits, strand, parent_start, parent_end, feature_id=None, sd_min=3, sd_max=17):
        results = []
        for idx, hit in enumerate(hits):
            # gene            complement(124..486)
            # -1      491     501     0       5       5
            # -1      491     501     0       4       5
            # -1      491     501     1       4       5
            # -1      491     501     2       3       5
            # -1      491     501     1       3       5
            # -1      491     501     0       3       5
            
            qualifiers = {
                "source": "CPT_ShineFind",
                "ID": "%s.rbs-%s" % (feature_id, idx),
            }

            if strand > 0:
                start = parent_end - hit["spacing"] - hit["len"]
                end = parent_end - hit["spacing"]
            else:
                start = parent_start + hit["spacing"]
                end = parent_start + hit["spacing"] + hit["len"]
            # check that the END of the SD sequence is within the given min/max of parent start/end

            # gap is either the sd_start-cds_end (neg strand) or the sd_end-cds_start (pos strand)
            # minimum absolute value of these two will be the proper gap regardless of strand
            tmp = gffSeqFeature(
                FeatureLocation(min(start, end), max(start, end), strand=strand),
                #FeatureLocation(min(start, end), max(start, end), strand=strand),
                type="Shine_Dalgarno_sequence",
                qualifiers=qualifiers,
            )
            results.append(tmp)
        return results

    def testFeatureUpstream(self, feature, record, sd_min=3, sd_max=17):
        # Strand information necessary to getting correct upstream sequence
        strand = feature.location.strand

        # n_bases_upstream (plus/minus 7 upstream to make the min/max define the possible gap position)
        if strand > 0:
            start = feature.location.start - sd_max - 7
            end = feature.location.start - sd_min
        else:
            start = feature.location.end + sd_min
            end = feature.location.end + sd_max + 7

        (start, end) = ensure_location_in_bounds(
            start=start, end=end, parent_length=len(record)
        )

        # Create our temp feature used to obtain correct portion of
        # genome
        tmp = gffSeqFeature(FeatureLocation(min(start, end), max(start, end), strand=strand), type="domain")
        seq = str(tmp.extract(record.seq))
        return self.list_sds(seq, sd_min, sd_max), start, end, seq

    def hasSd(self, feature, record, sd_min=3, sd_max=17):
        sds, start, end, seq = self.testFeatureUpstream(
            feature, record, sd_min=sd_min, sd_max=sd_max
        )
        return len(sds) > 0


# Cycle through subfeatures, set feature's location to be equal
# to the smallest start and largest end.
# Remove pending bugfix for feature display in Apollo
def fminmax(feature):
    fmin = None
    fmax = None
    for sf in feature_lambda([feature], feature_test_true, {}, subfeatures=True):
        if fmin is None:
            fmin = sf.location.start
            fmax = sf.location.end
        if sf.location.start < fmin:
            fmin = sf.location.start
        if sf.location.end > fmax:
            fmax = sf.location.end
    return fmin, fmax


def fix_gene_boundaries(feature):
    # There is a bug in Apollo whereby we have created gene
    # features which are larger than expected, but we cannot see this.
    # We only see a perfect sized gene + SD together.
    #
    # So, we clamp the location of the gene feature to the
    # contained mRNAs. Will remove pending Apollo upgrade.
    fmin, fmax = fminmax(feature)
    if feature.location.strand > 0:
        feature.location = FeatureLocation(fmin, fmax, strand=1)
    else:
        feature.location = FeatureLocation(fmin, fmax, strand=-1)
    return feature

def shinefind(
    fasta,
    gff3,
    gff3_output=None,
    table_output=None,
    lookahead_min=3,
    lookahead_max=17,
    top_only=False,
    add=False,
):
    table_output.write(
        "\t".join(
            [
                "ID",
                "Name",
                "Terminus",
                "Terminus",
                "Strand",
                "Upstream Sequence",
                "SD",
                "Spacing",
            ]
        )
        + "\n"
    )

    sd_finder = NaiveSDCaller()
    # Load up sequence(s) for GFF3 data
    seq_dict = SeqIO.to_dict(SeqIO.parse(fasta, "fasta"))
    # Parse GFF3 records
    for record in gffParse(gff3, base_dict=seq_dict):
        # Shinefind's gff3_output.
        gff3_output_record = SeqRecord(record.seq, record.id)
        # Filter out just coding sequences
        ignored_features = []
        for x in record.features:
            # If feature X does NOT contain a CDS, add to ignored_features
            # list. This means if we have a top level gene feature with or
            # without a CDS subfeature, we're catch it appropriately here.
            if (
                len(
                    list(
                        feature_lambda(
                            [x], feature_test_type, {"type": "CDS"}, subfeatures=True
                        )
                    )
                )
                == 0
            ):
                ignored_features.append(x)

        # Loop over all gene features
        for gene in feature_lambda(
            record.features, feature_test_type, {"type": "gene"}, subfeatures=True
        ):

            # Get the CDS from this gene.
            feature = sorted(
                list(
                    feature_lambda(
                        gene.sub_features,
                        feature_test_type,
                        {"type": "CDS"},
                        subfeatures=True,
                    )
                ),
                key=lambda x: x.location.start,
            )
            # If no CDSs are in this gene feature, then quit
            if len(feature) == 0:
                # We've already caught these above in our ignored_features
                # list, so we skip out on the rest of this for loop
                continue
            else:
                # Otherwise pull the first on the strand.
                feature = feature[0]

            # Three different ways RBSs can be stored that we expect.
            rbs_rbs = list(
                feature_lambda(
                    gene.sub_features,
                    feature_test_type,
                    {"type": "RBS"},
                    subfeatures=False,
                )
            )
            rbs_sds = list(
                feature_lambda(
                    gene.sub_features,
                    feature_test_type,
                    {"type": "Shine_Dalgarno_sequence"},
                    subfeatures=False,
                )
            )
            regulatory_elements = list(
                feature_lambda(
                    gene.sub_features,
                    feature_test_type,
                    {"type": "regulatory"},
                    subfeatures=False,
                )
            )
            rbs_regulatory = list(
                feature_lambda(
                    regulatory_elements,
                    feature_test_quals,
                    {"regulatory_class": ["ribosome_binding_site"]},
                    subfeatures=False,
                )
            )
            rbss = rbs_rbs + rbs_sds + rbs_regulatory

            # If someone has already annotated an RBS, we move to the next gene
            if len(rbss) > 0:
                log.debug("Has %s RBSs", len(rbss))
                ignored_features.append(gene)
                continue

            sds, start, end, seq = sd_finder.testFeatureUpstream(
                feature, record, sd_min=lookahead_min, sd_max=lookahead_max
            )

            feature_id = get_id(feature)
            sd_features = sd_finder.to_features(
                sds, feature.location.strand, start, end, feature_id=feature.id
            )

            human_strand = "+" if feature.location.strand == 1 else "-"

            # http://book.pythontips.com/en/latest/for_-_else.html
            log.debug("Found %s SDs", len(sds))
            for (sd, sd_feature) in zip(sds, sd_features):
                # If we only want the top feature, after the bulk of the
                # forloop executes once, we append the top feature, and fake a
                # break, because an actual break triggers the else: block
                table_output.write(
                    "\t".join(
                        map(
                            str,
                            [
                                feature.id,
                                feature_id,
                                feature.location.start,
                                feature.location.end,
                                human_strand,
                                sd_finder.highlight_sd(seq, sd["start"], sd["end"]),
                                sd["hit"],
                                int(sd["spacing"]) + lookahead_min,
                            ],
                        )
                    )
                    + "\n"
                )

                if add:
                    # Append the top RBS to the gene feature
                    gene.sub_features.append(sd_feature)
                    # Pick out start/end locations for all sub_features
                    locations = [x.location.start for x in gene.sub_features] + [
                        x.location.end for x in gene.sub_features
                    ]
                    # Update gene's start/end to be inclusive
                    gene.location._start = min(locations)
                    gene.location._end = max(locations)
                # Also register the feature with the separate GFF3 output
                sd_feature = fix_gene_boundaries(sd_feature)
                gff3_output_record.features.append(sd_feature)

                if top_only or sd == (sds[-1]):
                    break
            else:
                table_output.write(
                    "\t".join(
                        map(
                            str,
                            [
                                feature.id,
                                feature_id,
                                feature.location.start,
                                feature.location.end,
                                human_strand,
                                seq,
                                None,
                                -1,
                            ],
                        )
                    )
                    + "\n"
                )

        record.annotations = {}
        gffWrite([record], sys.stdout)

        gff3_output_record.features = sorted(
            gff3_output_record.features, key=lambda x: x.location.start
        )
        gff3_output_record.annotations = {}
        gffWrite([gff3_output_record], gff3_output)


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Identify shine-dalgarno sequences")
    parser.add_argument("fasta", type=argparse.FileType("r"), help="Fasta Genome")
    parser.add_argument("gff3", type=argparse.FileType("r"), help="GFF3 annotations")

    parser.add_argument(
        "--gff3_output",
        type=argparse.FileType("w"),
        help="GFF3 Output",
        default="shinefind.gff3",
    )
    parser.add_argument(
        "--table_output",
        type=argparse.FileType("w"),
        help="Tabular Output",
        default="shinefind.tbl",
    )

    parser.add_argument(
        "--lookahead_min",
        nargs="?",
        type=int,
        help="Number of bases upstream of CDSs to end search",
        default=3,
    )
    parser.add_argument(
        "--lookahead_max",
        nargs="?",
        type=int,
        help="Number of bases upstream of CDSs to begin search",
        default=17,
    )

    parser.add_argument("--top_only", action="store_true", help="Only report best hits")
    parser.add_argument(
        "--add",
        action="store_true",
        help='Function in "addition" mode whereby the '
        + "RBSs are added directly to the gene model.",
    )

    args = parser.parse_args()
    shinefind(**vars(args))