Classify somatic/germline mutations by deep neural networks

1. Download and install julia v0.5
2. Install Keras with TensorFlow by their guide
3. Install HDF5, StatsBase, PyCall and Libz in Julia

$ julia predict.jl predict ../model/model2.h5 input.bam input.txt > output.txt

By default it accept the txt file generated by AnnoVar.

First, run julia collect_data.jl cfDNA.bam gDNA.mpileup out.image out.txt for each sample to get a list of.image and .txt files, then call julia data_stat.jl clean to filter them, the result will be put on clean folder. After that, you can run julia ../model1.jl on the clean folder to train the corresponding model.