Algorithm - AntiBP3


Antibacterial peptides are small oligopeptides that play a major role in an organism’s innate immune system, exhibiting diverse protective activity against pathogens.ABPs are also effective against multi-drug-resistant bacterial species. To our knowledge, no method has been developed to predict antibacterial peptides that cover all bacterial groups. We have tried an attempt to develop a method for predicting ABPs against three different groups of bacteria, i.e., gram-positive, gram-negative and gram-variable bacteria. This tool uses an RF , ET and SVC-based classifiers for GP, GN and GV, respectively, to predict peptides with antibacterial activity. Here, on this page, the user can get the details of all feature generation techniques, algorithms and methods used to develop prediction models.


Figure

Dataset Used

Gram-positive ABPs dataset
It comprises of 744 antibacterial peptides against gram-positive bacteria and 744 non-antibacterial peptides.

Gram-negative ABPs dataset
It comprises of 1164 antibacterial peptides against gram-negative bacteria and 1164 non-antibacterial peptides.

Gram-variable ABPs dataset
It comprises of 7188 antibacterial peptides against gram-variable bacteria and 7188 non-antibacterial peptides.

Validation dataset
It comprises a total of 2274 positive and 2274 negative peptide sequences (data not used in training or testing) obtained from AntiBP2, DRAMP, dbAMP and CAMP databases. In order to avoid any biases, common peptide sequences found in any other dataset (used for developing the models) and in ABP-Finder were removed. Therefore, this validation dataset finally contains 186, 291 and 1797 unique experimentally validated balanced antibacterial and 186, 291 and 1797 non-antibacterial peptide sequences against gram-positive, gram-negative and gram-variable bacteria, respectively.