Skip to content

Convolutional Neural Network to predict antibiotic resistance from M. tuberculosis sequence data; inspired by https://github.com/aggreen/MTB-CNN with the main difference of being independent of input dimensions; implemented in TensorFlow

Notifications You must be signed in to change notification settings

julibeg/TB-AMR-CNN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TB-AMR-CNN

Neural Network inspired by MTB-CNN to predict antibiotic resistance in M. tuberculosis. The defining feature of this implementation is that it is fully independent of the input dimensions while showing similar performance to the original. Therefore, new sequences don't need to be aligned to a fixed-length MSA (which would drop any insertions).

The trained model was packaged in a Docker container for easy use (or to be used in tb-ml). It takes one-hot-encoded sequences of 18 loci as input and predicts resistance against 13 drugs (amikacin, capreomycin, ciprofloxacin, ethambutol, ethionamide, isoniazid, kanamycin, levofloxacin, moxifloxacin, ofloxacin, pyrazinamide, rifampicin, streptomycin). There is an accompanying pre-processing container for exracting the one-hot-encoded sequences from a SAM/BAM/CRAM file with reads aligned against the H37Rv reference. For details on the containers please refer to their repositories (for the neural net or the pre-processing pipeline).

About

Convolutional Neural Network to predict antibiotic resistance from M. tuberculosis sequence data; inspired by https://github.com/aggreen/MTB-CNN with the main difference of being independent of input dimensions; implemented in TensorFlow

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published