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

manual predict broken #97

Closed mahafarhat closed 7 years ago

mahafarhat commented 7 years ago

got the following error after selecting one mutation for prediciton:

Attaching package: ‘jsonlite’ The following object(s) are masked from ‘package:utils’: View Error in eval(expr, envir, enclos) : object 'INS_CF_4247350_i836TTCCAGAATATTGG_279_embB' not found Execution halted

doctormo commented 7 years ago

This error is because there are mutations which were removed from the database with the wg dataset, these were marked as predictors and thus entered the matrix.csv file that way. Not being in the database means they're not available to the randomForest process and it doesn't like missing columns.

mahafarhat commented 7 years ago

This seems nearly impossible as the predictor list was generated only from the targeted data set.

doctormo commented 7 years ago

It explains why the mutation that causes an error isn't in our database. When I add it to the matrix.csv the error changes to a new and different SNP name.

doctormo commented 7 years ago

This should be fixed and live now.

mahafarhat commented 7 years ago

still not fixed, I tried three different pncA mutations and I get the following error:

Genetic information: Mutations are not available in prediction: SNP_CN_2289103_T139C_T47A_pncA, SNP_CN_2289030_T212G_H71P_pncA ...

mahafarhat commented 7 years ago

and just tried mutations in a handful of other genes still with the same issue

doctormo commented 7 years ago

There seemed to be an issue with csv vs. line break format. That's fixed now. Can you try again?