AMP Sequence Analysis

We explore two directions in our research on AMPs. First, seek novel features and test them in a machine learning framework that discriminate between AMPs and non-AMPs. A preliminary study has been published: Daniel Veltri and Amarda Shehu. Physicochemical Determinants of Antimicrobial Activity. BICOB, Hawai, 2013. The study proposes local position-dependent features to provide highest resolution on specific positions and their role in antimicrobial activity. The goal of this line of investigation is to provide further understanding so that experimental techniques can be guided in their modification or design of novel more powerful AMPs for treating bacterial infections.

A second line of investigation concerns exploring interactions among features. A preliminary study, focusing on global features detected in experimentation and already shown to perform well in the context of classification in machine learning, employs regression models to discriminate among features and propose a best model built on subsets of cross-interacting features. This study has appeared in: Elena G. Randou, Daniel Veltri, and Amarda Shehu. "Systematic Analysis of Global Features and Model Building for Recognition of Antimicrobial Peptides". IEEE International Conference on Computational Advances in Bio and Medical Sciences (ICCABS), New Orleans, LA, 2013. A related publication is in: Elena G. Randou, Daniel Veltri, and Amarda Shehu. "Binary Response Models for Recognition of Antimicrobial Peptides." ACM Bioinf and Comp Biol (BCB), Washington, D. C. 2013.

The goal of this work is to help both the computational and experimental communities to make advances in understanding of AMPs for the ultimate objective of employing these small peptides as effective templates for novel antimicrobial compounds. With this objective in mind, we have put three web servers in place. The first one, available at http://binf.gmu.edu/dveltri/bicob2013/, provides detailed access to researchers on the local features discovered in our BICOB 2013 paper. The server is Cathelicidin Physicochemical Profile Explorer. The second one, PASS for Predictive AMP Statistical Server, available at http://binf.gmu.edu/dveltri/cgi-bin/iccabs2013.cgi, allows model building on any kind of features, even novel ones proposed by a researcher. The web server accompanies our paper at ICCABS 2013. A more recent and expanded version is available at http://binf.gmu.edu/dveltri/cgi-bin/AMP-PASS.cgi and accompanies our BCB 2013 paper.

This is Daniel's presentation at BiCOB 2013.


This is Daniel's presentation at ICCABS 2013.



On this Project:

  • Daniel Veltri (Ph.D. student, School of Systems Biology, GMU)

    Elena Randou (Dept. of Mathematics, GMU, now at FDA)

    Anand Vidyashankar (Dept. of Statistics, GMU)>/h4>

    Barney Bishop (Dept. of Biochemistry, GMU)

    Amarda Shehu

    See the list of publications for publications resulting from this project.