DeepCSeqSite (DCS-SI) is a toolkit for protein-ligand binding sites prediction.
The current version is a demo for DCS-SI.
The formal version will be released later.
Platform = Linux
Python = 2.7.x
1.5.0 <= TensorFlow <= 1.10.0
We strongly recommend you execute DCS-SI with GPU.
Enter the root dir. If needed, you can get help information by this command:
python dcs_si.py -hThe demo contains three versions of model which differ in their network.
The versions include DCS-SI-std, DCS-SI-k9, DCS-SI-k9a and DCS-SI-en.
For example, you can load DCS-SI-std by:
python dcs_si.py --model DCS-SI-stdDCS-SI-std is the default version of DCS-SI. The kernel width in DCS-SI is k = 5.
k9 means the kernel width k = 9, and 'a' in k9a means the model is trained on the augmented training set.
DCS-SI-en is the enhanced version of DCS-SI, which executes forward propagation twice and takes the previous output into consideration.
We provides all the test sets used in our paper.
The test sets include SITA, SITA-EX1, SITA-EX2 and SITA-EX3.
After loading a version of the model, you can easily test the model on the test sets with the guide of program.
The source code and notes of the network architecture can be found in Models directory.
Complete source code of training will be released later.
DeepCSeqSite is GPL 3.0-licensed