A machine learning method for disulfide bond engineering site prediction based on structures
This program include the following functions:
(1) select/download structure files;
(2) extract disulfide bonds and generate negative samples;
(3) convert the coordinates to distance information;
(4) training and testing;
(5) disulfide bond engineering site prediction;
(6) scripts for web-server setup.
SSBONDPredict is a project use a computational method based on neural network to predict residue pairs that can form disulfide bonds after cysteine mutations.The neural network was trained with atomic structures curated from the Protein Data Bank. The webserver are available at PredDisufideBond,when you click website you can see the detail source code of this webserver,you can get the detail usage about this project in PreDisulfideBond folder.Beside predicting the residue pairs which can form disulfide bonds after mutations,it also can calculate the change of entropy and energy due to mutations. The predicted result will show you this:
CYSA4-ARGA10 0.997 -24.4450 -1.8942
from left to right, the columes are:
- Residue pairs that are predicted to form disulfide bonds after mutations.
- The probability for this residue pairs to form disulfide bonds after mutations.
- The change of entropy after mutations
- The change of energy after mutations
see the README in the Source code pages.
If you don't want to spend energy on building complex python operating environment,you can download Executable file by click Executable file download.
SSBONDPredict is created by liulab of Beijing Compulational Science Research Center.