Genes to Transcripts misleading error messages

[leicray]

Describe the bug
The Genes to Transcripts tool expects the user to enter a gene symbol, or an HGNC ID, or a valid transcript identifier. Submitting more than one input type triggers an error message, but the messages vary according to the ORDER of the submitted search terms.

To Reproduce
Steps to reproduce the behaviour:

1. Go to https://variantvalidator.org/service/gene2trans/
2. Type both NM_000088.4 and COL1A1 into the data-entry box. Click the Submit button
3. The process completes and correctly returns information about the three transcript versions that are available. However, three error messages are displayed:

• EFAIL: Failed to recognize a gene symbol or identifier in your input.
• ISOURCE: https://api.lovd.nl/v2/checkHGVS/NM_000088.4 COL1A1
• LIBRARYVERSION: 0.5.2

4. However, if the order of the two search parameters is reversed, i.e. COL1A1 then NM_000088.4, there is a single error message:

• Error: No transcript definition for (tx_ac=COL1A1NM_000088)

5. The HGNC ID for COL1A1 is 2197. Submission of that ID on its own again displays the the three valid transcript versions but also triggers error messages:

• EFAIL: Failed to recognize a gene symbol or identifier in your input.
• ISOURCE: https://api.lovd.nl/v2/checkHGVS/2197
• LIBRARYVERSION: 0.5.2

Expected behaviour
The tool ought to trigger a simple warning that the expected input ought is a single string (gene symbol OR transcript identifier OR HGNC ID) if a user submits more that one string. Alternatively, the first string could be processed and subsequent strings ignored,

Improved prompts would help users.

Desktop (please complete the following information):

• OS: Windows 11
• Browser Chrome
• Version 140.0.7339.186

[ifokkema]

LOVD's checkHGVS endpoint expects a variant, and will recognize a reference sequence. LOVD's checkHGVS output for "NM_000088.4 COL1A1" is actually (simplified):

{
  "version": 2,
  "messages": [
    "Successfully received 1 variant description."
  ],
  "warnings": [],
  "errors": [],
  "data": [
    {
      "input": "NM_000088.4 COL1A1",
      "identified_as": "reference_sequence",
      "identified_as_formatted": "reference sequence",
      "valid": false,
      "messages": [],
      "warnings": {
        "WINPUTLEFT": "We stopped reading past \"NM_000088.4\". We could not interpret \" COL1A1\"."
      },
      "errors": {
        "EVARIANTREQUIRED": "This input requires a variant description; reference sequence given."
      },
      "data": [],
      "corrected_values": {
        "NM_000088.4": 0.1
      }
    }
  ]
}

However, when using the checkGene endpoint, the EFAIL will indeed occur. So, most likely, the printed LOVD API URL is not correct.

That said, I can definitely make https://api.lovd.nl/v2/checkGene/2197 return valid information, though. The checkGene endpoint can, of course, assume that any numeric input is an HGNC ID.

[leicray]

Thanks for the investigation. I will leave to others to decide what would be the best solution.

[Peter-J-Freeman]

2 issues here.

@ifokkema is correct, the URL we are creating for gene queries is wrong. I have updated

second is the way we handle the space here. What happens is that we just remove the whitespace. but for some reason it works. Will dig deeper

import json
import VariantValidator
vval = VariantValidator.Validator()
gene = 'NM_000088.4 COL1A1'
select_transcripts = None
g_and_t = vval.gene2transcripts(gene, select_transcripts=select_transcripts, bypass_genomic_spans=True, lovd_syntax_check=True)
print(json.dumps(g_and_t, sort_keys=True, indent=4, separators=(',', ': ')))

{
    "current_name": "collagen type I alpha 1 chain",
    "current_symbol": "COL1A1",
    "hgnc": "HGNC:2197",
    "lovd_corrections": {},
    "lovd_messages": {
        "EFAIL": "Failed to recognize a gene symbol or identifier in your input.",
        "ISOURCE": "https://api.lovd.nl/v2/checkGene/NM_000088.4 COL1A1",
        "LIBRARYVERSION": "0.5.6"
    },

[Peter-J-Freeman]

What I think is happening is that VV is concatenating to NM_000088.4COL1A1 then failing

LOVD is being passed the original query and finding that it is confident in COL1A1 then it is re-submitting COL1A1 back to VV.

So this is sort of a pass. Not sure what to do with it.

but this

 print(json.dumps(g_and_t, sort_keys=True, indent=4, separators=(',', ': ')))
{
    "error": "No transcript definition for (tx_ac=COL1A1NM_000088)",
    "requested_symbol": "COL1A1 NM_000088.4"
}

is certainly an error.

The easiest fix is to hard fail if a space is identified. There should never be one