Projects

The SBCB laboratory carries out research in machine learning, optimization/metahuristics and high performance computing for Bioinformatics problems. We apply computational methods to solve complex problems with focus on Structural Bioinformatics and Natural Sciences. We are engaged in the following scientific-technical projects:.

2019-2022: Mesoscopic Molecular Dynamics Simulations: Development of Models and Computational Strategies for Complex Structural Bioinformatics Problems
Description: There is a wide range of unanswered scientific questions which cannot be answered neither by experiments nor by classical modeling and simulation approaches. One question is how to consider microscopic and macroscopic effects in one model. On the one hand, available microscopic approaches (Molecular Dynamics (MD) type) are precise but computational too complex for large or multi-scale problems. On the other hand, several macroscopic methods (Computational Fluid Dynamics (CFD) type) cannot show crucial microscopic effects. Therefore, a mesoscopic scale model is required that closes this gap. For such a Mesoscopic Molecular Dynamics (MMD) model, we propose to take advantage of the Lattice Boltzmann Method (LBM), a modern CFD method, and state-of-the-art MD methods in order to combine both. This MMD model resolves certain microscopic effects important for the considered macroscopic application. In this project, two complementary academic expert groups from Brazil and Germany join together in order to provide an interdisciplinary research and academic education in the fields of mathematics, computer science, biology, chemistry and engineering to PhD students as well as to researchers.
Finantial Support:       Partners:    

2017-2018: Massively Parallel Computing for Personalized Medicine: Bioinformatics and the Discovery of Novel Biomarkers
Description: In this research we are interested in developing and applying strategies to integrate, explore and analyze genomic data to identify new biomarkers with diagnostic or prognostic value, and potential therapeutic targets. Through bioinformatics approaches and the union of the efforts of researchers with training and experience in Computer Science, Biology, and Health, we intend to address research questions in this line related to cardiovascular diseases and cancer, focusing on the translational potential of the in silico findings. That is, it is intended through the realization of this project to generate scientific knowledge with potential clinical use, that can bring significant benefits to the care actions and preventive activities in the Brazilian Unified Health System (SUS). The idea is to improve the process of diagnosis, prognosis and planning Therapeutic use of the patients affected by these pathologies. Within this theme, this project will specifically address the following research questions: i) detection of molecular differences between physiological and pathological cardiac hypertrophy with potential therapeutic application in the treatment of heart failure, ii) identification of biomarkers of unstable atherosclerotic disease, iii ) Characterization of molecular alterations in breast, ovarian and colon solid tumors in patients with germline mutations in genes predisposing to cancer for investigation of hereditary cancer biomarkers, iv) Evaluation of the pathogenicity of variants of uncertain significance in genes predisposing to cancer. GPU computing is used to accelerate the analyze of the amount of biological data (genomic, transcriptomic, and proteomic, etc.), to accelerate heuristic algorithms developed for structural bioinformatics problems, to accelerate molecular dynamics simulations and molecular docking experiments. Click here to see more details.
Finantial Support:         Partners:        

2017-2019: PadMetBio: Parallel and Distributed Metaheuristics for Structural Bioinformatics
Description: Structural bioinformatics deals with problems where the rules that govern the biochemical processes and relations are partially known which makes hard to design efficient computational strategies for these problems. There is a wide range of unanswered questions, which cannot be answered neither by experiments nor by classical modeling and simulation approaches. Specifically, there are several problems that still do not have a computational method that can guarantee a minimum quality of solution. The general goal of this project is to develop metaheuristics models based on robust and scalable parallel and distributed computing for structural bioinformatic problems. International Project (Stic-Amsud 2016) between the Universidade Federal do Rio Grande do Sul (UFRGS, Brazil), Universidad Nacional de San Luis (UNSL, Argentine), Universidad de Santiago de Chile (USACH, Chile), University of Pierre et Marie Currie (UPMC, LIP6, INRIA, France).Click here to see more details.
Finantial Support:    Partners:                 

2017-2018: Mesoscopic Molecular Simulations for Structural Bioinformatics Problems
Description: There is a broad range of unanswered scientific questions which cannot be answered neither by experiments nor by classical modeling and simulation approaches. One question is how to consider microscopic and macroscopic effects in one model. On the one hand, available microscopic methods (Molecular Dynamics (MD) type) are precise but computational too complex for large or multi-scale problems. On the contrary, several macroscopic methods (Computational Fluid Dynamics (CFD) type) cannot show severe microscopic effects. Therefore, a mesoscopic scale model is required that closes this gap. For such a Mesoscopic Molecular Dynamics (MMD) model, we propose to take advantage of the Lattice Boltzmann Method (LBM), a modern CFD method, and state-of-the-art MD methods to combine both. This MMD model resolves certain microscopic effects necessary for the considered macroscopic application. In this interdisciplinary research, we will investigate the development of such model and its applicability in Structural Bioinformatics problems such as protein structure prediction and molecular docking.
Finantial Support:     Partners:    

2016-2017: Parallel Metaheuristics for Structural Bioinformatics
Description: The main goal of this project is to gather together the expertise and current work of researchers in the areas of structural bioinformatics, metaheuristics and parallel computing, in order to build new interdisciplinary and high quality solutions for these hot research area using parallel computing. In this project, we will explore collaborative work for the design and implementation of knowledge-based hybrid population metaheuristics, like genetic algorithms and memetic algorithms and its implementation for parallel enviroments. There are several research opportunities to be explored in this field, with relevant multidisciplinary applications in Computer Science, Bioinformatics, Biochemistry, and the Medical Sciences. Support: Azure cloud under the Microsoft Azure for Research Award program.
Finantial Support:  

2015-2016: Massive Parallel Metaheuristics for Structural Bioinformatics
Description: The main goal of this project is to gather together the expertise and current work of researchers in the areas of structural bioinformatics, metaheuristics and parallel computing, in order to build new interdisciplinary and high quality solutions for the Molecular Docking and Protein Structure Prediction problems using GPU-accelerated computing. In this project, we will explore collabo- rative work for the design and implementation of knowledge-based hybrid population metaheuristics, like genetic algorithms and memetic algorithms and its implementation for GPUs.
Finantial Support:  

2013-2016: Desenvolvimento de métodos e estratégias computacionais para a predição da estrutura tridimensional de polipeptídeos
Description: Este projeto de pesquisa visa o estudo e o desenvolvimento de estratégias computacionais para a predição in-silico da estrutura tridimensional de polipeptídeos. Este projeto de pesquisa tem os seguintes objetivos específicos: (a) Estudo do problema da predição da estrutura 3D de proteínas in-silico e o desenvolvimento de estratégias computacionais para extração e manipulação de grandes quantidades de dados estruturais de polipeptídeos/proteínas determinados por meios experimentais (NMR e Raios X); (b) Estudo do problema da glicosilação de proteínas. Desenvolvimento de estratégias computacionais para manipulação de estruturas de carboidratos e simulação computacional da formação de gliconjugados; (c) Desenvolvimento de métodos e ferramentas computacionais que permitam o estudo do impacto na estrutura 3D de uma proteína quando da presença de carboidratos; (d) Estudo, desenvolvimento e aplicação de técnicas de otimização (metaheurísticas) e de aprendizagem de máquina no desenvolvimento de estratégias computacionais para a predição da estrutura 3D de proteínas; (e) Investigação e desenvolvimento de técnicas mais robustas para busca do espaço conformacional de polipeptídeos. Melhoria de métodos atuais que realizam a predição da estrutura 3D aproximada de polipeptídeos; e (f) Desenvolvimento e disponibilização a comunidade científica de ferramentas computacionais para o estudo e manipulação de dados biológicos (proteínas e gliconjugados).
Finantial Support:  

2013 - 2015: Aplicação de Técnicas de Aprendizagem de Máquina e Otimização no Desenvolvimento de Estratégias Computacionais para a Predição em Silico da Estrutura Tridimensional de Proteínas
Description: As proteínas ou polipeptídeos são macromoléculas que estão envolvidas na maior parte das transformações moleculares que ocorrem nos seres vivos. Estas transformações representam as funções que uma dada proteína exerce no organismo (transporte, de catálise, entre outras). Uma proteína é constituída por uma seqüência de aminoácidos, comumente denominada estrutura primária, que forma uma cadeia polipeptídica por meio da polimerização representada por uma reação de condensação. A ligação CO-NH (grupo carboxílico grupo amino) resultante, entre aminoácidos contíguos, é conhecida como ligação peptídica. A cadeia polipeptídica de uma proteína, em seu estado nativo, enovela-se assumindo uma conformação única (estrutura terciária). Esta conformação ou estrutura tridimensional (3D) determina a função que a proteína irá exercer na célula ou organismo. Este conhecimento é de fundamental importância no desenvolvimento de compostos químicos (fármacos) que possam inibir ou ativar a função de desta proteína no organismo. Experimentalmente, a estrutura tridimensional de uma proteína pode ser obtida através de técnicas de cristalografia por difração de raios X ou por espectroscopia de ressonância magnética nuclear (NMR, sigla em inglês). Entretanto, devido às diversas dificuldades, incluindo o alto custo e o elevado tempo demandado por estas técnicas, a determinação da estrutura 3D de proteínas ainda é um problema que desafia os cientistas. Cientistas de áreas como a ciência da computação, engenharia, bioinformática, matemática, física e química estão empenhados na construção de métodos e estratégias computacionais que possam determinar a estrutura tridimensional de polipeptídeos a partir da seqüência primária da proteína. Este projeto de pesquisa tem como objetivo o estudo, desenvolvimento e aplicação de técnicas aprendizagem de máquina e otimização na construção de estratégias computacionais para predição em sílico da estrutura tridimensional de polipeptídeos.
Finantial Support: