Siamese neural network for protein remote homology
The following architecture was built using Pytorch:
The final model consisted of twin four-layer convolutional neural networks:
Our model was validated using stratified 5-fold cross-validation:
SiameseNN Model performance on Saripella PFAM, GENE3D, and SUPFAM datasets compared to existing methods:
SiameseNN Model was found to have higher precision and lower FPR.


