Antimicrobial Peptide Scanner #

Introduction #

According to the evolutionary theory, each species is fighting for their survival, at the end those have a potential to combat to the given environment is the successor. In this principal, the array of pathogens fighting with array of antibiotics and keep evolving with the power of resistance. It’s good for pathogen, but trickier for human immune system to resolve the problem. However, in this event the host immune system uses several weapons, one of those is the small peptides to combat the pathogens. Those mechanisms are captured and used in drug developments. Similarly, in machine learning models, the recent evolved machine is “Deep Learning”, which adopted the neural network principle and improved for multi-layer neural networks to resolve more complex problems. Here, the antimicrobial peptides were predicted through Deep Learning and the characteristic of the peptides are understood more precisely through the binary mode classifications, i.e., AMPs and Non-AMPs. This proposed ML model only focused for gram-negative and gram-positive bacteria, and which out-performed the state-of-art AMP prediction methods.

Improvements #

1. Updated new datasets (Training and test) for AMP prediction 2.Implemented the Deep Neural Networks (DNN), which out-perform the existing state of art methods
2. Eliminated the a priori feature construction and dependency of the domain experts
3. Opened the dataset to public via AMP Scanner server (hhtp://www.ampscanner.com)
4. Proposed the minimum size of peptides, i.e. Nine amino-acids to the experimentalist.

Use #

Name of the Model : AMP Scanner

Webserver : https://www.dveltri.com/ascan/v2/ascan.html

Analysis: Submit the multi fasta (1 to 50,000) to the server and download the table as result.

Reference: #

Veltri, D.; Kamath, U.; Shehu, A. Deep learning improves antimicrobial peptide recognition. Bioinformatics 2018, 34, 2740-2747, doi:10.1093/bioinformatics/bty179.