Legacy Notes
Switch language: 旧版保留说明
This page preserves the essential writing from the original documentation rather than replacing it.
Original QuickStart highlights
CPU memory is recommended as
16GBGPU memory is recommended as
8GBPython 3.8
Keras == 2.4.0
TensorFlow == 2.3.0
numpy >= 1.15.4
scipy >= 1.2.1
scikit-learn >= 0.20.3
seaborn >=0.9.0
matplotlib >=3.1.0
Original motivation
Though deep learning methods have been widely developed for predicting chromatin interactions using flanking DNA sequence in identified chromatin interaction regions, a comprehensive software toolkit to integrate and evaluate different deep learning architectures are under-developed.
Original model inventory
onehot_cnn_one_branchonehot_cnn_two_branchonehot_embedding_denseonehot_embedding_cnn_one_branchonehot_embedding_cnn_two_branchonehot_denseonehot_resnet18embedding_cnn_one_branchembedding_cnn_two_branch
Original data organization highlights
The original project centered on:
DeepChrInteract.pyas the main entry file;data_preprocessing.pyfor preprocessing;model.pyfor model definitions;train.pyfor training;test.pyfor evaluation;embedding_matrix.npyfor pretrained DNA2Vec features;data/for labeled sequence folders;h5_weights/andresult/for saved outputs.
Preservation policy
The current documentation keeps these legacy concepts visible and extends them with the modern PyTorch redesign rather than erasing them.