Goby 3.0 supports using deep learning models for calling somatic variations in:

  • Matched RNA-Seq samples (one germline and one somatic sample, model trained with Nugen ovation v2 and Illumina sequencing)
  • Matched exome samples (one germline and one somatic sample, Haloplex exome kit, Illumina sequencing)
  • Trio exome samples (one father sample, one mother sample, and one somatic sample, Haloplex exome kit, Illumina sequencing)

To use these models, you need to follow the instructions provided on this Goby 2.0 tutorial, but add the following options to activate the correct model:

-x SomaticVariationOutputFormat:model-path=${GOBY_HOME}/models/somatic-variation/rna-seq-1472848302343/bestModel.bin \
-x SomaticVariationOutputFormat:model-p-mutated-threshold:0.99

The first line specifies the model to use, and the second line the threshold for reporting variations.

The Goby 3.0.2 distribution includes three  models under ${GOBY_HOME}/models:

  • rna-seq-1472848302343/   
  • pair_exome-1472682707130/
  • trio_exome-1472682734666/

Additional models may be provided in future releases.

If you wish to train models for additional types of assays, you can follow the variation analysis tutorial. The models obtained following variations of this tutorial can be used directly with Goby.