Update consequence severity filtering #326
Conversation
| VEP_SHORT_QUERY_DISTANCE = 5000 | ||
| VEP_LONG_QUERY_DISTANCE = 500000 |
There was a problem hiding this comment.
What was the reason behind no longer querying VEP twice if the short distance doesn't find any gene? Just curious
| @@ -97,54 +88,58 @@ def load_consequence_severity_rank(): | |||
| return {term: index for index, term in enumerate(get_severity_ranking())} | |||
|
|
|||
|
|
|||
| def extract_consequences(vep_results, acceptable_biotypes, only_closest, results_by_variant, report_distance=False): | |||
| def extract_consequences(vep_results, acceptable_biotypes): | |||
| """Given VEP results, return a list of consequences matching certain criteria. | |||
|
|
|||
| Args: | |||
| vep_results: results obtained from VEP for a list of variants, in JSON format. | |||
| acceptable_biotypes: a list of transcript biotypes to consider (as defined in Ensembl documentation, see | |||
| https://www.ensembl.org/info/genome/genebuild/biotypes.html). Consequences for other transcript biotypes | |||
There was a problem hiding this comment.
Why, among all biotypes, we only consider miRNAs and protein coding genes? Sorry again, has not much to do with this PR but I'm curious.
There was a problem hiding this comment.
It's a good question and it seems to have been decided long ago (before the birth of this repo).
There are a few others that we could consider in the non-coding genes, the IG genes and the TR genes.
Probably. a topic to raise with OpenTarget
There was a problem hiding this comment.
Created #328 for this, we can discuss at our next meeting.
|
Actually that's just 87 new evidence strings for the set of 2000 variants used in the test for the original VEP pipeline (based on coordinates, mostly SNPs / short indels). You can see this test for a much smaller set used in testing the current structural pipeline. For your other questions, they're from before my time as well (and actually before the time of this repo, so not even in the history!) so we'd have to ask @tskir or @tcezard... I'm also curious about this. |
| @@ -97,54 +88,58 @@ def load_consequence_severity_rank(): | |||
| return {term: index for index, term in enumerate(get_severity_ranking())} | |||
|
|
|||
|
|
|||
| def extract_consequences(vep_results, acceptable_biotypes, only_closest, results_by_variant, report_distance=False): | |||
| def extract_consequences(vep_results, acceptable_biotypes): | |||
| """Given VEP results, return a list of consequences matching certain criteria. | |||
|
|
|||
| Args: | |||
| vep_results: results obtained from VEP for a list of variants, in JSON format. | |||
| acceptable_biotypes: a list of transcript biotypes to consider (as defined in Ensembl documentation, see | |||
| https://www.ensembl.org/info/genome/genebuild/biotypes.html). Consequences for other transcript biotypes | |||
There was a problem hiding this comment.
It's a good question and it seems to have been decided long ago (before the birth of this repo).
There are a few others that we could consider in the non-coding genes, the IG genes and the TR genes.
Probably. a topic to raise with OpenTarget
consequence_prediction/vep_mapping_pipeline/consequence_mapping.py
Outdated
Show resolved
Hide resolved
Closes #321
To get a sense of the additional consequences the new strategy gives us, you can look at this commit which shows the diff in the VEP test file for SNPs, as well as some of the other tests.