Transfer GFF to GTF file

Basic Information

GFF3 (Generic Feature Format Version 3) file format represents the genomic features in a simple text-based tab-delimited file

GFF3 file has nine fields (seqid, source, feature, start, end, score, strand, phase, and attributes)

The lines which starts with ‘##’ provides the meta-information of the file and ‘#’ represents the human-readable comments

We sometimes need to transfer GFF3 to GTF format.

I used gffread before. It generally works well, but it has some problem when I need to deal with a GFF3 file downloaded from NCBI.

This file has some lines like:

chrxxx  Gnomon  exon    46964   46999   .   -   .   ID=id-LOC123327042;Parent=gene-LOC123327042;Dbxref=GeneID:123327042;gbkey=exon;gene=LOC123327042;model_evidence=Supporting evidence includes similarity to: 1 Protein%2C and 74%25 coverage of the annotated genomic feature by RNAseq alignments;pseudo=true
chrxxx  Gnomon  exon    47054   47468   .   -   .   ID=id-LOC123327042-2;Parent=gene-LOC123327042;Dbxref=GeneID:123327042;gbkey=exon;gene=LOC123327042;model_evidence=Supporting evidence includes similarity to: 1 Protein%2C and 74%25 coverage of the annotated genomic feature by RNAseq alignments;pseudo=true
chrxxx  Gnomon  exon    47542   47661   .   -   .   ID=id-LOC123327042-3;Parent=gene-LOC123327042;Dbxref=GeneID:123327042;gbkey=exon;gene=LOC123327042;model_evidence=Supporting evidence includes similarity to: 1 Protein%2C and 74%25 coverage of the annotated genomic feature by RNAseq alignments;pseudo=true

When I used gffread, it can not read there lines containing “ID=id*”, it will lose the gene_id attribute in the output GTF file.

I realized that I need to write a script by myself.

Script

import sys
import uuid
import re

def parse_gff_attributes(attributes_str):
    """Parse GFF attributes into a dictionary."""
    attributes = {}
    for attr in attributes_str.split(';'):
        if attr:
            key_value = attr.split('=', 1)
            if len(key_value) == 2:
                key, value = key_value
                attributes[key] = value
    return attributes

def convert_gff_to_gtf(gff_file, gtf_file):
    """Convert GFF3 to GTF format, retaining ID, gene_id, and transcript_id."""
    with open(gff_file, 'r') as gff, open(gtf_file, 'w') as gtf:
        for line in gff:
            if line.startswith('#'):
                gtf.write(line)
                continue
            
            fields = line.strip().split('\t')
            if len(fields) != 9:
                continue
                
            seqid, source, feature, start, end, score, strand, phase, attributes_str = fields
            attributes = parse_gff_attributes(attributes_str)
            
            # Skip features that are not relevant for GTF (e.g., region)
            if feature not in ['gene', 'mRNA', 'exon', 'CDS', 'lnc_RNA', 'pseudogene']:
                continue
                
            # Determine feature type for GTF
            gtf_feature = feature
            if feature == 'mRNA' or feature == 'lnc_RNA':
                gtf_feature = 'transcript'
            
            # Extract required attributes
            gene_id = attributes.get('gene', '')
            transcript_id = attributes.get('transcript_id', '')
            feature_id = attributes.get('ID', '')
            
            # Build GTF attributes string
            gtf_attributes = []
            if feature == 'gene':
                gtf_attributes.append(f'gene_id "{gene_id}"')
                if 'Name' in attributes:
                    gtf_attributes.append(f'gene_name "{attributes["Name"]}"')
            elif feature in ['mRNA', 'lnc_RNA', 'exon', 'CDS']:
                gtf_attributes.append(f'gene_id "{gene_id}"')
                gtf_attributes.append(f'transcript_id "{transcript_id}"')
                if 'Name' in attributes and feature in ['mRNA', 'lnc_RNA']:
                    gtf_attributes.append(f'gene_name "{attributes["Name"]}"')
                if feature == 'exon':
                    gtf_attributes.append(f'exon_id "{feature_id}"')
            elif feature == 'pseudogene':
                gtf_attributes.append(f'gene_id "{gene_id}"')
                if 'Name' in attributes:
                    gtf_attributes.append(f'gene_name "{attributes["Name"]}"')
            
            # Write GTF line
            gtf_fields = [seqid, source, gtf_feature, start, end, score, strand, phase, '; '.join(gtf_attributes)]
            gtf.write('\t'.join(gtf_fields) + '\n')

if __name__ == '__main__':
    if len(sys.argv) != 3:
        print("Usage: python gff_to_gtf.py input.gff output.gtf")
        sys.exit(1)
    
    gff_file = sys.argv[1]
    gtf_file = sys.argv[2]
    convert_gff_to_gtf(gff_file, gtf_file)
    print(f"Conversion complete. GTF file written to {gtf_file}")

Usage

It is very easy to use this script, just save it as gff2gtf.py and then run like

python gff2gtf.py input.gff output.gtf

The result looks like

chrxxx      Gnomon  exon    46964   46999   .       -       .       gene_id "LOC123327042"; transcript_id ""; exon_id "id-LOC123327042"
chrxxx      Gnomon  exon    47054   47468   .       -       .       gene_id "LOC123327042"; transcript_id ""; exon_id "id-LOC123327042-2"
chrxxx       Gnomon  exon    47542   47661   .       -       .       gene_id "LOC123327042"; transcript_id ""; exon_id "id-LOC123327042-3"