Some HGNC identifiers do not resolve in /bioentityset/slimmer/function route

[kltm]

There seems to be an issue on the Alliance site wrt HGNC identifiers and GO data:

https://agr-jira.atlassian.net/browse/AGRHELP-531

@sierra-moxon has started to look at it.

@dustine32 , IIRC, you helped with the HGNC mapping for the widget way back when, right?

The issue seems to be that when going to the ribbon, even though there is data that appears when you click nothing comes up.

Works for mouse:

https://api.geneontology.org/api/bioentityset/slimmer/function?slim=GO:0003674&slim=GO:0008150&slim=GO:0005575&subject=MGI:3588192&rows=-1

not for this HGNC:

https://api.geneontology.org/api/bioentityset/slimmer/function?slim=GO:0003674&slim=GO:0008150&slim=GO:0005575&subject=HGNC:8729&rows=-1

but it does work for this one:

https://api.geneontology.org/api/bioentityset/slimmer/function?slim=GO:0003674&slim=GO:0008150&slim=GO:0005575&subject=HGNC:8725&rows=-1

Noting that we still depend on mygene.info as well: https://github.com/geneontology/go-fastapi/blob/main/app/routers/slimmer.py#L105 .

[cmungall]

Problem Diagnosis

The issue occurs in the reverse mapping step when converting UniProt IDs back to HGNC IDs in uniprot_to_gene_from_mygene() (app/routers/slimmer.py:170).

Root Cause

The function performs a generic text search against mygene.info:

results = mg.query(id, fields="HGNC")  # searches ALL fields for "P12004"

This causes two issues:

1. False matches on Ensembl protein IDs: A search for "P12004" returns 9 hits because it matches:

   • UniProt accession P12004 (the desired result)
   • Ensembl protein IDs with suffix .P12004 (e.g., ADAR2_007719.P12004, IPEI_012922.P12004)

   Example: https://mygene.info/v3/query?q=P12004&fields=all&size=10

2. Wrong gene selected: The code blindly takes results["hits"][0], which is often an irrelevant gene lacking an HGNC field, causing a KeyError.

3. Incomplete UniProt coverage: Only queries Swiss-Prot, but gene_to_uniprot_from_mygene() returns both Swiss-Prot and TrEMBL IDs.

Recommended Fix

Use field-specific queries to avoid false matches:

def uniprot_to_gene_from_mygene(id: str):
    """Query MyGeneInfo with a UniProtKB id and get its corresponding HGNC gene."""
    gene_id = None
    if id.startswith("UniProtKB"):
        id = id.split(":", 1)[1]

    mg = get_client("gene")
    try:
        # Query specifically in the uniprot fields to avoid false matches
        results = mg.query(f"uniprot.Swiss-Prot:{id} OR uniprot.TrEMBL:{id}", 
                          fields="HGNC,symbol,uniprot")
        if results["hits"]:
            # Verify the hit actually contains our UniProt ID
            for hit in results["hits"]:
                if "uniprot" in hit:
                    uniprot_data = hit["uniprot"]
                    # Check Swiss-Prot
                    has_id = False
                    if "Swiss-Prot" in uniprot_data:
                        swiss_prot = uniprot_data["Swiss-Prot"]
                        if (isinstance(swiss_prot, str) and swiss_prot == id) or \
                           (isinstance(swiss_prot, list) and id in swiss_prot):
                            has_id = True
                    # Check TrEMBL if not found
                    if not has_id and "TrEMBL" in uniprot_data:
                        trembl = uniprot_data["TrEMBL"]
                        if (isinstance(trembl, str) and trembl == id) or \
                           (isinstance(trembl, list) and id in trembl):
                            has_id = True

                    if has_id and "HGNC" in hit:
                        gene_id = hit["HGNC"]
                        if not gene_id.startswith("HGNC"):
                            gene_id = "HGNC:{}".format(gene_id)
                        break
    except ConnectionError:
        logging.error("ConnectionError while querying MyGeneInfo with {}".format(id))
    
    if not gene_id:
        raise DataNotFoundException(detail="No HGNC IDs found for {}".format(id))
    return [gene_id]

Compare queries:

• Generic (9 hits, wrong gene): https://mygene.info/v3/query?q=P12004&fields=all
• Field-specific (1 hit, correct): https://mygene.info/v3/query?q=uniprot.Swiss-Prot:P12004&fields=all

This fix also requires:

• Better error handling around the map2slim call
• Graceful fallback when UniProt-to-HGNC conversion fails
• Check for HGNC field existence before accessing

I have a working implementation with tests if you'd like me to submit a PR.

---

Comment generated with assistance from Claude Code

[sierra-moxon]

please submit a PR after updating with the current main branch

[kltm]

@sierra-moxon , I know we had some fixes that sound similar to this recently. The immediate expression seems to be:
https://api.geneontology.org/api/bioentityset/slimmer/function?slim=GO:0003824&subject=HGNC:8729&rows=-1
returns with a 500 code, when results are expected. The examples initially given above seem to work.

Perhaps an ongoing issue with mygene.info's completeness?

This is being tracked at https://agr-jira.atlassian.net/browse/KANBAN-865

[kltm]

(@pgaudet Just tagging you to let you know that we'll need to look at this.)