In our lab we are integrated in multidisciplinary projects centered on two major themes for which we established a network of international and national collaborations, both experimental and theoretical. In fact, we lead teams of geographically distributed researchers (Bonn U, Medipol U, Weill Cornell M, Mount Sinai, Utrecht U, IST – Lisbon U, CMAT, INESC and LQAV from Porto U, CNC at UC) with very diverse scientific backgrounds to work at the interface of Computer Science and Structural Biology.


Contrary to the reductionist approach aimed at understanding individual components, the new data revolution allows the understanding of complicated interactions and  pathways through the use of statistical Data Mining/Machine-Learning (ML) techniques. This is an active area of research in computer science with the increasing availability of big data collections of all sorts prompting interest in the development of novel tools for data mining. It is expected that continuous improvement of software infrastructure will make ML applicable to a growing range of biological problems. We envision, plan and manage large and completely independent projects that merged knowledge from chemistry, bioinformatics, biophysics, statistics and data-mining to build innovative and completely new methods and computational solutions. Striking examples are the application of ML to:

a) the detection of Hot-Spots (Protein, 2014, 82;479);

b) to the understanding of protein-protein interfacial (PPI) characteristics;

c) to the study of co-evolution in PPIs (MOL2NET, 2016, doi:10.3390/mol2net-02-03889);

d) to the construction of web-servers such as SBHD (J Chem Info Model, 2015, 55, 1077) and SPOTON (Sci Rep, 2017, 7, 8007);

e) to the application to small-ligand binding IC50 prediction (J Comput Aided Mol Des, 2018, 32, 175). We are now involved in the development of new methodologies to be applied to membrane proteins and drug design.


We focused not only in the individual characterization of molecular systems but as well in the development of new and crucial methodologies to tackle specific problems in the area of structural computational biology. In the past we have developed a new methodological approach to calculate the binding free energy upon alanine mutation of residues at a protein-protein interface – compASM (Computational Alanine Scanning Mutagenesis, J Comput Chem, 2007, 28, 644). It has been applied since then very successfully to interfaces in a variety of systems (e.g. MDM2/P53, HEL/antibody FvD1.3, HEL/HyHEL-10, ZipA/FtsZ, IgG1/Streptococcal Protein G) allowing to highlight key features of complex formation.  We compared our approach against Thermodynamic Integration (TI), and our results demonstrated that the much faster compASM protocol gives results at the same level of accuracy as the TI method, but at a fraction of the computational time required to run TI (J Chem Theor Comput, 2013, 9, 1311). We studied in detail the physic-chemical characteristics of the aromatic residues due to their high prevalence as HS at protein-protein interfaces (Biochim Biophys.Acta, 2013, 183, 404). A new VMD plugin was built and is now available for the scientific community (CompASM, Theor Chem Acc, 2012, 131, 1271). We also successfully extended the approach to protein-nucleic acids systems (J Chem Theor Comput, 2013, 9, 4243). Several Solvent Accessible Surface Area (SASA) features were measured and their role in the determination of the binding Hot-Spots statistically evaluated. The combination of these features was also analyzed by a support vector machine learning (SVM) algorithm, which led to an accurate new model for predicting these crucial residues: SBHD (J Chem Info Model, 2015, 55, 1077). More recently we launched a new web-server to predict HS with a very high accuracy and specificity (SPOTON, Sci Rep, 2017, 7, 8007).

We are also working in PPIs involving G-Protein Coupled Receptors (GPCRs). In our previous research we have contributed to the discovery of asymmetrical signaling through GPCR dimers and determined models for oligomeric GPCRs assemblies (Nature Chem Biol, 2009, 5, 688), produced comprehensive classifications of GPCR/G-protein interfaces that determine subtype selectivity (Biochem Biophys Acta, 2014, 1840, 16), identified functionally critical region on arrestin structure that can be targeted with drugs or chemical tools for functional modulation (ACS Chem Neuro, 2016, 7, 1212–1224) and produced computational protocols to study GPCR function and mechanism (Methods Cell Biol, 2017, 205). Based on this experience and the unprecedented progress in the determination of detailed molecular structures of GPCRs, G-proteins, and members of the Arr family, we are now involved in work that aims to illuminate the molecular mechanisms of signaling selectivity with powerful computational methods by exploring the dynamic properties of the complex systems in their natural environment.

In addition to GPCRs function and molecular mechanism research, we are also working in the design of HIV-based Virus-Like Particles (VLP) that express specific antibodies on the surface as potential target-specific drug/radioactivity delivery platforms to receptors overexpressed in cancer cells, in particular HER2-positive breast cancer cells. We have taken advantage of the newly settled computational methods (Int J Mol Sci, 2016, 27, 17; Sci Rep, 2017, 7, 8007) to study the key interactions of the antibodies with the HIV capsid as well with the receptors. The understanding of the established structural interactions is of fundamental as a guideline to monitor the experimental work paths.

In the area of protein-small ligand interactions, we have been working on various systems: the PICK1 PDZ domain (PNAS, 2010, 107, 413), the nitric oxide synthase isoforms (Chem Biol Drug Design, 2015, 86, 1072), the TEM family of β-lactamases (J Chem Info Model, 2013, 53, 2648) and the Dopamine 2 Receptor. All these studies result from an interchange of ideas with experimental groups in order to develop better and more specific small-molecules.