search

farhat-lab / gentb-site

The genTB project, the Django site, variant calling and prediciton pipeline, and mapping pipeline with hooks to two ravens
https://gentb.hms.harvard.edu
Other
8 stars 11 forks source link

Add detected resistance variants to the predict.JSON in WDNN pipeline #247

Open mgro opened 3 years ago

mgro commented 3 years ago

@eford8 The variants based on which the WDNN predicts resistance are currently not reported in the JSON predict output. We can add a step to the WDNN predict python script that parses these variants to add to the predict output.

mahafarhat commented 2 years ago

@mgro can you please review this issue, and post any progress and anything else pending to do

mgro commented 2 years ago

@mahafarhat yes this specific issue can be closed.

Resistance variants can be explored like in the RF pipelines (https://gentb.hms.harvard.edu/predict/9b809c212cd12a0b0b85baafc6cb3183/).

One problem is that as you know the WDNN not only uses the specific SNPs (defined by position, nucleotide/AA change), but also features like 'any SNP or any indel in gid'. This then looks as follows on GenTB:

Screenshot 2021-09-22 at 13 19 56

Maybe we need to add a sentence that if no specific SNP but only a gene locus is specified simply the presence of any mutation in that gene has been used as a feature in WDNN's prediction?

https://gentb.hms.harvard.edu/tb/media/pipeline/files/feature_list_222.txt