为何Biomart提取的旁系同源基因数量与Ensembl网站不符？

Hey there, let's break down why you might be seeing that 1-paralogue gap between your code results (99 paralogues for AA792892/ENSMUSG00000073497) and the Ensembl web interface (100 paralogues)—this is a common gotcha with comparative genomics data, and there are a few key angles to check:

1. Paralogue Type Filtering Differences #

Ensembl's Compara pipeline sorts paralogues into distinct subtypes, and your code might be excluding a category that the web UI includes by default:

• Check if your query is filtering out low-confidence or retrieved paralogues (these are sometimes flagged separately but still displayed on the web).
• Verify if you're only fetching one-to-one paralogues, while the web interface includes one-to-many or many-to-many duplication relationships.

2. Ensembl Release Version Mismatch #

Ensembl updates their genome builds and Compara datasets regularly, and even minor version differences can change the paralogue set:

• Confirm your code uses the same Ensembl release as the web interface (look for the release number at the bottom of the Ensembl page—e.g., Release 110).
• If you're using the Ensembl REST API, BioMart, or Perl API, double-check that the release parameter is set to match the web version. Using an older API version could mean you're pulling from an outdated dataset.

3. Edge Cases: Pseudogenes or Partial Alignments #

The web interface might include genes your code is filtering out due to strict criteria:

• Check if the missing paralogue is a pseudogene (look for the "Gene type" column in the Ensembl table). Many scripts default to excluding pseudogene biotypes, but the web UI displays them.
• Verify if your code enforces a minimum sequence identity or alignment length threshold that the missing paralogue doesn't meet.

4. API Query Parameter Nuances #

Small differences in how you structure your API query can lead to missing entries:

• If using the Ensembl Perl API's get_all_paralogues method, make sure you aren't accidentally setting a species filter that narrows results too much (even for mouse, this can exclude cross-species paralogues if misconfigured).
• For BioMart users: Compare the filters in your script to the ones the web interface uses (the paralogue page relies on BioMart under the hood—you can export the web query's filter settings and match them in your code).

5. Counting Logic Errors #

It's easy to miscount if your code handles duplicates or nested relationships incorrectly:

• Ensure you aren't removing reciprocal paralogue pairs (e.g., if Gene A is a paralogue of your target, your code might count it once, but the web UI lists each unique gene only once—wait, no, actually Ensembl lists each distinct paralogue gene once, so check if your code is counting groups instead of individual genes).
• Double-check that you're iterating through the results correctly (e.g., not skipping an entry due to a null value or formatting issue).

Next Steps to Diagnose #

1. Identify the missing gene: Export your code's list of paralogue Ensembl IDs and compare it to the web interface's table. The missing gene's attributes (biotype, release status) will point you to the root cause.
2. Test a direct query: Use the Ensembl REST API to fetch paralogues for your target gene and inspect the raw output—this will show you if the missing gene is present in the API response, which tells you if your code is filtering it out.
3. Cross-reference documentation: Check the Ensembl Compara docs to confirm which paralogue types are included in the web interface vs. your code's query.

内容的提问来源于stack exchange，提问作者Jack Dean