PROGRAM:        clean_genes

DESCRIPTION:    Given a GFF describing a set of genes and a corresponding 
                multiple alignment, output a new GFF with only those
                genes that meet certain "cleanliness" criteria. The
                coordinates in the GFF are assumed to correspond to
                the reference sequence in the alignment, which is
                assumed to be the first one listed.  Default behavior
                is simply to require that all annotated start/stop codons and
                splice sites are valid in the reference sequence (GT-AG, 
                GC-AG, and AT-AC splice sites are allowed).  This can
                be used with an "alignment" consisting of a single
                sequence to filter out incorrect annotations.  Options
                are available to impose additional criteria as well,
                mostly having to do with conservation across species
                (see the '--conserved' option in particular).

USAGE:          clean_genes [options] <gff_fname> <msa_fname>

OPTIONS:        

    --start, -s
        Require conserved start codons (all species)

    --stop, -t
        Require conserved stop codons (all species)

    --splice, -l    
        Require conserved splice sites (all species).  By default,
        only GT-AG, GC-AG, and AT-AC splice sites are allowed (see also
        --splice-strict)

    --fshift, -f 
        Require that no frame-shift gap is present in any species.  Frame 
        shifts are evaluated with respect to the reference sequence.  Gaps 
        that have non-multiple-of-three lengths are allowed if  
        compensatory gaps occur nearby (see source code for details).

    --nonsense, -n  
        Require that no premature stop codon is present in any species.

    --conserved, -c
        Implies --start, --stop, --splice, --fshift, and --nonsense.
        Recommended option for cross-species analysis.

    --N-limit, -N <f>
        Maximum fraction of bases aligned to CDSs that are Ns in any
        species (<f> must be between 0 and 1).  Default is 0.05.  Set to 1
        to allow any number of Ns.

    --clean-gaps, -e
        Require all cds gaps to be multiples of three in length.  Can be 
        used with --conserved.

    --indel-strict, -I
        Implies --clean_gaps, usually used with --conserved.  Prohibits
        overlapping cds gaps in different sequences, gaps near cds 
        boundaries, and gaps in the reference sequence within and between
        flanking features (splice sites, etc.; see code for details).
        Designed for use in training a phylo-HMM with an indel model.

    --splice-strict, -C
        Implies --splice.  Allow only GT-AG canonical splice sites.  Useful
        when training a gene finder with a simple model for splice sites.

    --groupby, -g <tag>
        Group features according to specified tag (default
        "transcript_id").  If any feature within a group fails, the
        entire group will be discarded.  By choosing to group features
        according to different criteria, you can make the program
        "clean" the data set at different levels.  For example, to
        clean at the level of individual exons, add a tag like
        "exon_id" to indicate exons (see the program "refeature"),
        and then invoke clean_genes with "--groupby exon_id".

    --msa-format, -i FASTA|PHYLIP|MPM|MAF|SS
        Alignment file format.  Default is to guess format from file 
        contents.

    --refseq, -r <seqfile.fa>
        (Required with --msa-format MAF)  Complete reference 
        sequence for alignment (FASTA format).

    --offset5, -o <n>
        (Default 0)  Offset of canonical "GT" with respect to boundary 
        on *intron side* of annotated 5' splice sites.  Useful with
        annotations that describe a window around the canonical splice site.

    --offset3, -p <n>
        (Default 0)  Offset of canonical "AG" with respect to boundary 
        on intron side of annotated 3' splice sites.

    --log, -L <fname>
        Write human-readable log to specified file.

    --machine-log, -M <fname>
        Like --log, but produces more concise, machine-readable log.

    --stats, -S <fname>
        Write statistics on retained and discarded features to specified file.

    --discards, -d <fname>
        Write discarded features to specified file.

    --no-output, -x
        Suppress output of "cleaned" features to stdout.  Useful if only
        log file and/or stats are of interest.

    --help, -h
        Print this help message.

NOTES:  Feature types are defined as follows.

               coding exon    <-> "CDS"
               start codon    <-> "start_codon"
               stop codon     <-> "stop_codon"
               5' splice site <-> "5'splice"
               3' splice site <-> "3'splice"

        In addition, splice sites in UTR can be separately designated as
        "5'splice_utr" and "3'splice_utr".  Errors in these sites will be
        given a different code in the log files, which can be useful for
        tracking purposes.

        If evaluation is done at the level of individual exons (see
        --groupby), then splice sites are considered independently
        rather than in the context of introns.  As a result, the exons flanking
        a GT-AC or AT-AG intron might (misleadingly) be considered "clean".

        With --fshift and --nonsense, it is possible for entries
        to pass through that have stop codons in the frame of the
        *reference* sequence, although they do not have any in their
        own frame.  Use --clean-gaps instead to guarantee that no stop
        codons occur in any sequence in the frame of the reference
        sequence.