Prediction Mode
Users are interested to assess the performance of novel most informative residue distribution based neural network on unseen data.
- Users can provide viral protein sequence and host protein sequence
in respective text boxes to get the prediction whether provided protein sequence pairs interact or not.
- Users can also upload a csv file of viral protein and host protein sequences and perform inference using one of the 3 pre-trained model.
- At the end of inference, csv artifact containing model predictions can be downloaded to analyze the performance of chosen model.
Training Mode
In training module, users have the freedom to choose custom distribution of lncRNA sequences and hyper-parameters to improve the convergence and generalizeability of neural network (batch size, epochs, learning rate, dropout). To keep the researchers and practitioners informed about the processing of neural network, performance of neural network in terms of different evaluation measures will be communicated.
- Users need to provide a csv file containing viral protein and host protein sequence pairs and class information
- They have the freedom to choose protein sequence encoding scheme such as local encoding, global encoding,
or local-global encoding scheme, and data-split strategy (Standard, K-fold) to train the deep forest model from scratch
- Sign up preferably using organizational email account with providing the required data and purpose of experimentation
- After the completion of SignUp process, one need to wait for approval of account and permission for training
- If the request is approved, you will be able to login just for one time training.
- On successful activation of processing command, exploratory model training engine will process the data shortly in order to train the model.
- At the end of training, users can download performance related artifacts to analyze the deep forest model behavior.