Recent Publications   

 

Image 1



Research


Editor for the Following Journals

Walter F. de Azevedo Jr., PhD.


"Computational scientists solve tomorrow's problems with yesterday's computers; computer scientists seem to do it the other way around."--anonymous.

I graduated in Physics (BSc in Physics) at University of Sao Paulo (USP) in 1990. I completed a Master Degree in Applied Physics also at USP (1992), working under the supervision of Prof. Yvonne P. Mascarenhas, the founder of crystallography in Brazil. My dissertation was about X-ray crystallography applied to organometallics compounds (De Azevedo et al., 1995). During my Ph.D., I worked under the supervision of Prof. Sung-Hou Kim (University of California, Berkeley. Department of Chemistry), on a split Ph.D. program with a fellowship from Brazilian Research Council (CNPq)(1993-1996). My Ph.D. was about the crystallographic structure of CDK2 (Cyclin-Dependent Kinase 2) (De Azevedo et al., 1996). In 1996, I returned to Brazil. In April 1997, I finished my Ph.D. and moved to Sao Jose do Rio Preto (SP, Brazil) (UNESP) and worked there from 1997 to 2005. In 1997, I started the Laboratory of Biomolecular Systems- Department of Physics-UNESP - São Paulo State University. In 2005, I moved to Porto Alegre/RS (Brazil), where I am now. My current position is the coordinator of the Laboratory of Computational Systems Biology at Pontifical Catholic University of Rio Grande do Sul (PUCRS). My research interests are interdisciplinary with two major emphases: Bio-inspired computing and Computational Systems Biology. I published over 170 scientific papers about protein structures and computer simulation methods applied to the study of biological systems (H-index: 38). These publications have over 5000 citations.   

Research Interests 

 

Prof. Walter F. de Azevedo Jr. and the clusters Pandora and LaBoheme,UNESP-2003.

  

           I focus my research on the protein-ligand interaction problem. As a scientific problem, we may address it from different perspectives. In the later 1990s and early 2000s, my goal was on the application of experimental techniques such as X-ray diffraction crystallography and nuclear magnetic resonance (NMR) to determine the three-dimensional structures of biological macromolecules and investigate their interactions with potential binders. In the recent years, I have moved forward to simulate the interactions using molecular docking and creating machine-learning models to assess binding affinity using the atomic coordinates of protein-ligand complexes. My focus is now on the development of programs to evaluate protein-ligand interactions. The most recent progress was the creation of the programs SAnDReS and Taba. SAnDReS is a suite of computational tools to carry out docking simulations and to generate machine-learning models to predict binding affinity. The program Taba was developed to create machine-learning models based on an ensemble of protein structures for complexes involving protein and ligand. Taba considers that we may approach the protein-ligand problem as a spring-mass system, where we approximate the binding affinity using a pseudo-energy equation similar to the equation to calculate the potential energy of a spring-mass system. I successfully applied SAnDReS to study cyclin-dependent kinase, HIV-1 protease, and estrogen receptor. Taba is in the final stage of development, and I use it to study protein targets to generate enzyme-targeted scoring functions for prediction of binding affinity.

            I envisage the problem of protein-ligand interaction as an intersection of the protein sequence space and the chemical space, and I propose to approach these sets as a unique complex system, where the application of computational methodologies could contribute to the creation of specific scoring functions to predict binding affinity. To establish a robust mathematical framework to address this problem, I developed the concept of scoring function space that I use to find an adequate model to predict binding affinity for a biological system of interest. I designed the program SAnDReS as a tool to explore this scoring function space, where through machine learning techniques I build new models to predict binding affinity. One way to think about this approach is considering the experimental binding data and the structures available for a given protein as a system, where through the application of machine learning techniques I generate a computational model tailored to this biological system. In doing so, I give up to find a fit-all-model for all proteins; I address this problem creating a fine-tuned model, that seems reasonable considering the limited amount of experimental data, especially considering the complex structures for which experimental binding affinity is available.

            Also, I expect that as being proteins subject the evolution and being inserted in a complex chemical environment, as found in the biological systems, the application of a targeted machine-learning model is adequate to predict binding affinity.  


                                                            Walter F. de Azevedo Jr., PhD.