csy99 / dna-nn-theory

Exploring data augmentation, natural language ideas and transfer learning in neural network for genomic sequences

Towards a robust out-of-the-box neural network model forgenomic data

Zhaoyi Zhang, Songyang Cheng, Claudia Solis-Lemus

Reproducible scripts

CNN

• histone.ipynb
• motif_discovery.ipynb
• splice.ipynb

LSTM-layer

v1

/v1_scripts

• supervised_chip(256_adam)_save_embeddings.ipynb
• supervised_chip(256_sgd)_save_embedding.ipynb
• supervised_UCI_adam256_save_embedding.ipynb
• supervised_UCI_sgd256_save_embedding.ipynb
• supervised_viridae_adam256_save_embedding.ipynb
• supervised_viridae_save_embedding.ipynb
• supervised_viridae(front)_sgd256_save_embedding.ipynb
• supervised_viridae(mdl)_sgd256_save_embedding.ipynb
• supervised_viridae(rear)_sgd256_save_embedding.ipynb

v2

• uci_baseline_adam256.ipynb
• histone_lstm_layer_adam.ipynb
• histone_lstm_layer_sgd.ipynb
• discovery_baseline.py

doc2vec+NN

v1

/v1_scripts

• doc2vec_chip_save_embedding.ipynb
• doc2vec_UCI_save_embedding.ipynb
• doc2vec_viridae_save_embedding.ipynb

v2

• uci_doc2vec_save_embedding.ipynb
• histone_doc2vec.ipynb
• discovery_doc2vec.py

LSTM-AE+NN

v1

/v1_scripts

• ae(1_layer)_chip_adam32_save_embedding.ipynb
• ae(1_layer)_chip_adam256_save_embedding.ipynb
• ae(1_layer)_chip_adam1024_save_embedding.ipynb
• ae(1_layer)_UCI_save_embedding.ipynb
• ae(1_layer)_viridae_save_embedding.ipynb

v2

• uci_ae_adam32.ipynb
• uci_ae_adam256.ipynb
• histone_ae_adam32.ipynb
• histone_ae_adam256.ipynb
• histone_ae_adam1024.ipynb
• discovery_ae.py

Figures in the manuscript

All figures were created with the R script in plots/final-plots.Rmd