Store GTF/GFF source, transcript_support_level and CDS

[davmlaw]

These all vary per transcript/build so should be in "genome builds"

• GTF/GFF column 2 - source (per feature - so collect up as list)
• transcript_support_level (Ensembl only)
• ccds - Ensembl = ccds_id, RefSeq = dbxref CCDS

[davmlaw]

Step 1: looks like source is unique, for the latest ones at least:

from HTSeq import GFF_Reader
from collections import defaultdict, Counter

filenames = [
    "/data/cdot_data/downloads/Homo_sapiens_GRCh38_RefSeq_RS_2025_08.gff.gz",
    "/data/cdot_data/downloads/Homo_sapiens_GRCh38_Ensembl_115.gtf.gz",
]

for filename in filenames:
    print(f"Processing {filename}")
    source_per_transcript = defaultdict(set)
    for feature in GFF_Reader(filename):
        if transcript_id := feature.attr.get("transcript_id"):
            source_per_transcript[transcript_id].add(feature.source)

    for transcript_id, source_set in source_per_transcript.items():
        if len(source_set) > 1:
            print(transcript_id, source_set)

[davmlaw]

A small amount did have dupes, so I ended up storing it in a list (as JSON has no set) like biotypes