Conversion of Super-Silencers into Super-Enhancers Drives Transcriptional Reprogramming in Cancer Metastasis

We developed a deep learning framework to predict silencers. The following files are included here: A) README.txt. B) model.zip package includes: 1. all models 2. the programs as needed, C) predicted silencers (super silencers and typical silencers) for each tested sample. Command examples 1) phase-one model files phase_one_model.hdf5 and phase_one_weights.hdf5 are the built phase-one model structure and its weights. 2) Predicting using the phase-one model. Example: python phase_one_data_prediction.py temp.bed hg38.fa (Please download the genome sequence from https://hgdownload.gi.ucsc.edu/goldenPath/hg38/bigZips/hg38.fa.gz and unzip the file.) 3) Output of the phase-two model Example: python two_phase.pred.py ./CL-0000062model temp.bed.phase_one.pred.hdf5 temp.bed.CL-0000062.output.hdf5 For an input sequence, the output is a 1*3 vector. The first column represents the silencer prediction score. Architectures of deep learning models The phase-two model in the TREDNet consists of two convolutional layers and three fully connected layers arranged sequentially. The details of this model are: 1. 1-dimensional (1D) convolutional layer with 64 kernels, each having a window size of four and a step size of one. 2. Maxpooling layer with a window size of three and a step size of two. 3. Dropout layer with a dropout proportion of 0.2. 4. 1D convolutional layer 128 kernels, each having a window size of three and a step size of one. 5. Dropout layer with a dropout proportion of 0.2. 6. Fully connected layer of 100 neurons with the sigmoid activation function. 7. Fully connected layer of 50 neurons with the sigmoid activation function. 8. Fully connected output layer of 3 neurons with the SoftMax activation function.