python_functions

# load GTF file into python and parse the attributes field
# --------------------------------------------------------

import gffutils
import pandas as pd

db = gffutils.create_db(gtf_file_name, ':memory:', disable_infer_transcripts=True, disable_infer_genes=True)

def process_attributes(feature):
    # Attributes are stored as lists in gffutils
    return {key: ','.join(value) for key, value in feature.attributes.items()}

# Iterate over all features
features_data = []
for feature in db.all_features():
    # Extract standard GFF columns
    data = {
        'seq_id': feature.seqid,
        'source': feature.source,
        'type': feature.featuretype,
        'start': feature.start,
        'end': feature.end,
        'score': feature.score,
        'strand': feature.strand,
    }
    # Parse attributes
    data.update(process_attributes(feature))
    features_data.append(data)

# Convert the list of dictionaries to a pandas DataFrame
gtf = pd.DataFrame(features_data)