Refinamento manual e automático de modelos tridimensionais de proteínas para o workflow científico MHOLline

Abstract

MHOLline is a scientific workflow designed to model and analyze proteins, reaching researchers in domains of Bioinformatics, Biophysics, Computational Chemists and Computational Biologists. This project started in 2004 as a local software and became a web service in 2010 (available at http : //www .mholline . lncc . br), through the partnership between the Universidade Federal do Rio de Janeiro (UFRJ) and Laboratório Nacional de Computação Científica (LNCC). In 2013, a new partnership with Universidade Federal de Juiz de Fora started the development of MHOLline 2.0, now available at http : //www .mholline2 . lncc . br. This version presents a new interface and a refinement ama to logged users, offering the possibility to add or modify the template of the protein, remove signal peptides and restrict secondary structures and optimize protein loops on Modeller. All can be done in an automatic way, dispensing the user to manually generate any script. The user can also refine the protein automatically trough the use of artificial intelligence tools classifying the generated results with a set of restrictions in groups, aiming to reduce the effort to analyze the final refinement results. In this work, we also present a new proposal for clustering protein models based on a set of attributes related to their quality (i.e., energy and structural quality). To the user, in addition to the groups of structures with similar qualities, is also returned the representative structure of each group, in order to assist in the decision making of which model or models will follow for the future studies.