Server Detail Resource Contact
Resource page


The program is wrote by Tensorflow in python in our paper. And the result can be downloaded here and the code can be downloaded here


In our study, we focus on the recognition of antiviral peptides. Here, we evaluate on the well-established dataset, which is proposed by Nishant et al. The peptide sequences are collected with a reported antiviral activity against human viruses like HIV, HCV, SARS and Influenza, etc. More than 90% of antiviral peptides are extracted from natural source, and remaining peptides have synthetic source.

We make use of 604 highly effective antiviral peptides and 452 least or non-effective antiviral peptides, as one training set T 544p+407n (544 positive and 407 negative) and one validation set V60p+45n (60 positive and 45 negative). Also, we take non-experimental negative peptides, as one training set T544p+544n* and one validation set V60p+60n*. The negative peptides have been employed in earlier antimicrobial peptide prediction method AntiBP2. And the dataset can be downloaded here.

With the development of antiviral peptides (AVPs) research, in addition to the previous dataset, the AVPs databases have emerged in large numbers. We summarize the commonly used databases including antiviral peptides, and list the relevant information in Table 1. To update and expand the size of dataset, we extract 916 highly effective antiviral peptides from four different datasets (AVPdb, APD3, CAMPR3, LAMP) and 452 non-effective antiviral peptides from one dataset (AVPdb). The homologous sequences are removed by CD-hit if they shared a high sequence identity (greater than 90%) with any sequence in the dataset. Finally, we obtain 413 AVPs and 348 non-AVPs as a novel non-redundant AVPs dataset. And the dataset can be download here