Section: New Results

Structural Systems Biology and Docking

Participants : Thomas Bourquard, Marie-Dominique Devignes, Anisah Ghoorah, Bernard Maigret, Lazaros Mavridis, Violeta Pérez-Nueno, Dave Ritchie, Malika Smaïl-Tabbone, Vishwesh Venkatraman.

Structural systems biology aims to describe and analyze the many components and interactions within living cells in terms of their three-dimensional (3D) molecular structures. Much of our work in this area has been funded by the ANR Chaires d'Excellence project entitled “High Performance Algorithms for Structural Systems Biology” (HPASSB) which was awarded to Dave Ritchie (January 2009 – September 2011). A related follow-on ANR project entitled “Polynomial Expansions of Protein Structures and Interactions” (PEPSI) has recently started (November 2011). The HPASSB project complements existing competencies in the Orpailleur team represented by Marie-Dominique Devignes (CR CNRS) who is coordinating the MBI project (Modelling Biomolecules and their Interactions, http://bioinfo.loria.fr ), Malika Smaïl-Tabbone (MCU Nancy University) who is working on data integration and relational data-mining approaches, and Bernard Maigret (DR CNRS) who has an extensive experience of molecular dynamics and virtual screening. We are currently developing advanced computing techniques for molecular shape representation, protein-protein docking, protein-ligand docking, high-throughput virtual drug screening, and knowledge discovery in databases dedicated to protein-protein interactions. The PEPSI project is a collaboration with Sergei Grudinin at INRIA Grenoble (project Nano-D) and Valentin Gordeliy at the Institut de Biologie Structurale in Grenoble. This new project will involve developing further the above techniques and using them to help solve the structures of large molecular systems experimentally.

Accelerating protein docking calculations using graphics processors

We have recently adapted the Hex protein docking software to use modern graphics processors (GPUs) to carry out the expensive FFT part of a docking calculation [115] . Compared to using a single conventional central processor (CPU), a high-end GPU gives a speed-up of 45 or more. This software is publicly available at http://hex.loria.fr . A public GPU-powered server has also been created (http://hexserver.loria.fr ) [99] . These advances have facilitated further work on modeling the assembly of multi-component molecular structures using a particle swarm optimization technique [69] .

Eigen-Hex: Modeling protein flexibility during docking

Although the Hex protein docking software can often make reasonably good predictions about how two proteins might fit together, a major limitation of many current algorithms, including Hex, is that that they assume that proteins are rigid objects. In fact, proteins can be highly flexible, and the internal conformations of their atoms often change on going from the unbound forms in the free proteins to the bound conformations in the complex. We have developed a novel approach to model such flexibility using a principal component analysis (PCA) technique to identify and predict the main atomic motions during a docking calculation. Our approach gives better results than rigid body docking, although the flexible docking problem is still by no means solved. A journal article describing this work has been submitted.

3D-Blast: A new approach for protein structure alignment and clustering

We recently developed a new sequence-independent protein structure alignment approach called 3D-Blast [102] , which exploits the spherical polar Fourier (SPF) correlation technique used in the Hex protein docking software [114] . This approach recently performed very well in a blind shape comparison experiment organized by Orpailleur as part of Eurographics Workshop on 3D Object Retrieval [103] . The utility of this approach has been demonstrated by clustering subsets of the CATH protein structure classification database [106] for each of the four main CATH fold types, and by searching the entire CATH database of some 12,000 structures using several protein structures as queries. Overall, the automatic SPF clustering approach agrees very well with the expert-curated CATH classification, and ROC-plot analysis of database searches show that the approach has very high precision and recall. We recently proposed that the 3D-Blast approach could ultimately provide a novel way to enumerate and index protein fold space (major [7] ).

KBDOCK: Protein docking using Knowledge-Based approaches

Protein docking is the difficult computational task of predicting how a pair of three-dimensional protein structures come together to form a complex. Historically, there has been considerable interest in developing ab initio docking algorithms such as the Hex docking program developed by Dave Ritchie. However, as structural genomics initiatives continue to populate the space of protein 3D structures, and as several on-line databases of protein interactions have recently become available, using structural database systems to perform docking by homology will become an increasingly powerful approach to predicting protein interactions. In order to explore such possibilities, Anisah Ghoorah has recently developed the KDBOCK system as part of her doctoral thesis project. KDBOCK combines residue contact information from the 3DID database [117] with the Pfam protein domain family classification [89] together with coordinate data from the Protein Data Bank [86] in order to describe and analyze all known protein-protein interactions for which the 3D structures are available. In a recent publication [24] we demonstrated the utility of this approach for template-based docking using 73 complexes from the Protein Docking Benchmark [94] . KBDOCK is available at http://kbdock.loria.fr .

V-Dock: scoring protein-protein interactions using Voronoi fingerprints

There is growing interest in using machine learning techniques to analyze and populate protein-protein interaction (PPI) networks [104] . The aim of this project is to investigate the use of Voronoi fingerprints [16] as a way to distinguish cognate and non-cognate pairs of protein-protein interfaces. In collaboration with colleagues in the INRIA AMIB and INRA Bios teams, we recently applied our Voronoi fingerprint representation (V-Dock) to re-score rigid body docking predictions from Hex [60] , and we demonstrated that it could be used to improve the ranking of 7 out of 9 docking targets from the CAPRI protein docking experiment [60] . This approach was also used to predict the stability of engineered protein structures for another recent CAPRI target [21] .

DOVSA: Developing new algorithms for virtual screening

In 2010, Violeta Pérez-Nueno joined the Orpailleur team thanks to a Marie Curie Intra-European Fellowship (IEF) award to develop new virtual screening algorithms (DOVSA). The aim of this project is to advance the state of the art in computational virtual drug screening by developing a novel consensus shape clustering approach based on spherical harmonic (SH) shape representations [110] . The main disease target in this project is the acquired immune deficiency syndrome (AIDS), caused by the human immuno-deficiency virus (HIV) [109] . However, the approach will be quite generic and will be broadly applicable to many other diseases. So far, good progress has been made on calculating and clustering spherical harmonic “consensus shapes” which represent rather well the essential features of groups of active molecules [30] . Recent progress on this project has been presented orally at the 5th Journée Nationale de Chémoinformatique in Cabourg, the 9th International Conference on Chemical Structures in Noordwijkerhout, and at 3rd International Conference on Drug Discovery and Therapy in Dubai. A review of the state of the art in drug promiscuity was also recently published [29] .