Data-driven clustering : new methods and applications

Abstract : This work is concerned with the development and application of novel unsupervised learning methods, having in mind two target applications: the analysis of forensic case data and the classification of remote sensing images. First, a method based on a symbolic optimization of the inter-sample distance measure is proposed to improve the flexibility of spectral clustering algorithms, and applied to the problem of forensic case data. This distance is optimized using a loss function related to the preservation of neighborhood structure between the input space and the space of principal components, and solutions are found using genetic programming. Results are compared to a variety of state-of--the-art clustering algorithms. Subsequently, a new large-scale clustering method based on a joint optimization of feature extraction and classification is proposed and applied to various databases, including two hyperspectral remote sensing images. The algorithm makes uses of a functional model (e.g., a neural network) for clustering which is trained by stochastic gradient descent. Results indicate that such a technique can easily scale to huge databases, can avoid the so-called out-of-sample problem, and can compete with or even outperform existing clustering algorithms on both artificial data and real remote sensing images. This is verified on small databases as well as very large problems. Résumé : Ce travail de recherche porte sur le développement et l'application de méthodes d'apprentissage dites non supervisées. Les applications visées par ces méthodes sont l'analyse de données forensiques et la classification d'images hyperspectrales en télédétection. Dans un premier temps, une méthodologie de classification non supervisée fondée sur l'optimisation symbolique d'une mesure de distance inter-échantillons est proposée. Cette mesure est obtenue en optimisant une fonction de coût reliée à la préservation de la structure de voisinage d'un point entre l'espace des variables initiales et l'espace des composantes principales. Cette méthode est appliquée à l'analyse de données forensiques et comparée à un éventail de méthodes déjà existantes. En second lieu, une méthode fondée sur une optimisation conjointe des tâches de sélection de variables et de classification est implémentée dans un réseau de neurones et appliquée à diverses bases de données, dont deux images hyperspectrales. Le réseau de neurones est entraîné à l'aide d'un algorithme de gradient stochastique, ce qui rend cette technique applicable à des images de très haute résolution. Les résultats de l'application de cette dernière montrent que l'utilisation d'une telle technique permet de classifier de très grandes bases de données sans difficulté et donne des résultats avantageusement comparables aux méthodes existantes.

Organisational reforms in active welfare states: A comparative analysis of the turn to 'single gateways' in Western Europe

Based on the case of reforms aimed at integrating the provision of income protection and employment services for jobless people in Europe, this thesis seeks to understand the reasons which may prompt governments to engage in large-scale organisational reforms. Over the last 20 years, several European countries have indeed radically redesigned the organisational structure of their welfare state by merging or bundling existing front-line offices in charge of benefit payment and employment services together into 'one-stop' agencies. Whereas in academic and political debates, these reforms are generally presented as a necessary and rational response to the problems and inconsistencies induced by fragmentation in a context of the reorientation of welfare states towards labour market activation, this thesis shows that the agenda setting of these reforms is in fact the result of multidimensional political dynamics. More specifically, the main argument of this thesis is that these reforms are best understood not so such from the problems induced by organisational compartmentalism, whose political recognition is often controversial, but from the various goals that governments may simultaneously achieve by means of their adoption. This argument is tested by comparing agenda-setting processes of large-scale reforms of coordination in the United Kingdom (Jobcentre Plus), Germany (Hartz IV reform) and Denmark (2005 Jobcentre reform), and contrasting them with the Swiss case where the government has so far rejected any coordination initiative involving organisational redesign. This comparison brings to light the importance, for the rise of organisational reforms, of the possibility to couple them with the following three goals: first, goals related to the strengthening of activation policies; second, institutional goals seeking to redefine the balance of responsibilities between the central state and non-state actors, and finally electoral goals for governments eager to maintain political credibility. The decisive role of electoral goals in the three countries suggests that these reforms are less bound by partisan politics than by the particular pressures facing governments arrived in office after long periods in opposition.

Resting-State Functional Connectivity Emerges from Structurally and Dynamically Shaped Slow Linear Fluctuations.

Brain fluctuations at rest are not random but are structured in spatial patterns of correlated activity across different brain areas. The question of how resting-state functional connectivity (FC) emerges from the brain's anatomical connections has motivated several experimental and computational studies to understand structure-function relationships. However, the mechanistic origin of resting state is obscured by large-scale models' complexity, and a close structure-function relation is still an open problem. Thus, a realistic but simple enough description of relevant brain dynamics is needed. Here, we derived a dynamic mean field model that consistently summarizes the realistic dynamics of a detailed spiking and conductance-based synaptic large-scale network, in which connectivity is constrained by diffusion imaging data from human subjects. The dynamic mean field approximates the ensemble dynamics, whose temporal evolution is dominated by the longest time scale of the system. With this reduction, we demonstrated that FC emerges as structured linear fluctuations around a stable low firing activity state close to destabilization. Moreover, the model can be further and crucially simplified into a set of motion equations for statistical moments, providing a direct analytical link between anatomical structure, neural network dynamics, and FC. Our study suggests that FC arises from noise propagation and dynamical slowing down of fluctuations in an anatomically constrained dynamical system. Altogether, the reduction from spiking models to statistical moments presented here provides a new framework to explicitly understand the building up of FC through neuronal dynamics underpinned by anatomical connections and to drive hypotheses in task-evoked studies and for clinical applications.

Enhanced biodegradation of hexachlorocyclohexane (HCH) in contaminated soils via inoculation with Sphingobium indicum B90A.

Soil pollution with hexachlorocyclohexane (HCH) has caused serious environmental problems. Here we describe the targeted degradation of all HCH isomers by applying the aerobic bacterium Sphingobium indicum B90A. In particular, we examined possibilities for large-scale cultivation of strain B90A, tested immobilization, storage and inoculation procedures, and determined the survival and HCH-degradation activity of inoculated cells in soil. Optimal growth of strain B90A was achieved in glucose-containing mineral medium and up to 65% culturability could be maintained after 60 days storage at 30 degrees C by mixing cells with sterile dry corncob powder. B90A biomass produced in water supplemented with sugarcane molasses and immobilized on corncob powder retained 15-20% culturability after 30 days storage at 30 degrees C, whereas full culturability was maintained when cells were stored frozen at -20 degrees C. On the contrary, cells stored on corncob degraded gamma-HCH faster than those that had been stored frozen, with between 15 and 85% of gamma-HCH disappearance in microcosms within 20 h at 30 degrees C. Soil microcosm tests at 25 degrees C confirmed complete mineralization of [(14)C]-gamma-HCH by corncob-immobilized strain B90A. Experiments conducted in small pits and at an HCH-contaminated agricultural site resulted in between 85 and 95% HCH degradation by strain B90A applied via corncob, depending on the type of HCH isomer and even at residual HCH concentrations. Up to 20% of the inoculated B90A cells survived under field conditions after 8 days and could be traced among other soil microorganisms by a combination of natural antibiotic resistance properties, unique pigmentation and PCR amplification of the linA genes. Neither the addition of corncob nor of corncob immobilized B90A did measurably change the microbial community structure as determined by T-RFLP analysis. Overall, these results indicate that on-site aerobic bioremediation of HCH exploiting the biodegradation activity of S. indicum B90A cells stored on corncob powder is a promising technology.

Field/Laboratory Testing of Damaged Prestressed Concrete Girder Bridges, HR-397, 1999

Due to frequent accidental damage to prestressed concrete (P/C) bridges caused by impact from overheight vehicles, a project was initiated to evaluate the strength and load distribution characteristics of damaged P/C bridges. A comprehensive literature review was conducted. It was concluded that only a few references pertain to the assessment and repair of damaged P/C beams. No reference was found that involves testing of a damaged bridge(s) as well as the damaged beams following their removal. Structural testing of two bridges was conducted in the field. The first bridge tested, damaged by accidental impact, was the westbound (WB) I-680 bridge in Beebeetown, Iowa. This bridge had significant damage to the first and second beams consisting of extensive loss of section and the exposure of numerous strands. The second bridge, the adjacent eastbound (EB) structure, was used as a baseline of the behavior of an undamaged bridge. Load testing concluded that a redistribution of load away from the damaged beams of the WB bridge was occurring. Subsequent to these tests, the damaged beams in the WB bridge were replaced and the bridge retested. The repaired WB bridge behaved, for the most part, like the undamaged EB bridge indicating that the beam replacement restored the original live load distribution patterns. A large-scale bridge model constructed for a previous project was tested to study the changes in behavior due to incrementally applied damage consisting initially of only concrete removal and then concrete removal and strand damage. A total of 180 tests were conducted with the general conclusion that for exterior beam damage, the bridge load distribution characteristics were relatively unchanged until significant portions of the bottom flange were removed along with several strands. A large amount of the total applied moment to the exterior beam was redistributed to the interior beam of the model. Four isolated P/C beams were tested, two removed from the Beebeetown bridge and two from the aforementioned bridge model. For the Beebeetown beams, the first beam, Beam 1W, was tested in an "as removed" condition to obtain the baseline characteristics of a damaged beam. The second beam, Beam 2W, was retrofit with carbon fiber reinforced polymer (CFRP) longitudinal plates and transverse stirrups to strengthen the section. The strengthened Beam was 12% stronger than Beam 1W. Beams 1 and 2 from the bridge model were also tested. Beam 1 was not damaged and served as the baseline behavior of a "new" beam while Beam 2 was damaged and repaired again using CFRP plates. Prior to debonding of the plates from the beam, the behavior of both Beams 1 and 2 was similar. The retrofit beam attained a capacity greater than a theoretically undamaged beam prior to plate debonding. Analytical models were created for the undamaged and damaged center spans of the WB bridge; stiffened plate and refined grillage models were used. Both models were accurate at predicting the deflections in the tested bridge and should be similarly accurate in modeling other P/C bridges. The moment fractions per beam were computed using both models for the undamaged and damaged bridges. The damaged model indicates a significant decrease in moment in the damaged beams and a redistribution of load to the adjacent curb and rail as well as to the undamaged beam lines.

Contribution of electrostatic interactions, compactness and quaternary structure to protein thermostability: lessons from structural genomics of Thermotoga maritima.

Studies of the structural basis of protein thermostability have produced a confusing picture. Small sets of proteins have been analyzed from a variety of thermophilic species, suggesting different structural features as responsible for protein thermostability. Taking advantage of the recent advances in structural genomics, we have compiled a relatively large protein structure dataset, which was constructed very carefully and selectively; that is, the dataset contains only experimentally determined structures of proteins from one specific organism, the hyperthermophilic bacterium Thermotoga maritima, and those of close homologs from mesophilic bacteria. In contrast to the conclusions of previous studies, our analyses show that oligomerization order, hydrogen bonds, and secondary structure play minor roles in adaptation to hyperthermophily in bacteria. On the other hand, the data exhibit very significant increases in the density of salt-bridges and in compactness for proteins from T.maritima. The latter effect can be measured by contact order or solvent accessibility, and network analysis shows a specific increase in highly connected residues in this thermophile. These features account for changes in 96% of the protein pairs studied. Our results provide a clear picture of protein thermostability in one species, and a framework for future studies of thermal adaptation.

Influence of biocontrol strain Pseudomonas fluorescens CHA0 and its antibiotic overproducing derivative on the diversity of root colonizing pseudomonads

Non-target effects of biocontrol strains of Pseudomonas on the population of resident pseudomonads should be assessed prior to their large scale application in the environment. The rifampicin resistant bacterium P. fluorescens CHA0-Rif and its antibiotic overproducing derivative CHA0-Rif/pME3424 were introduced into soil microcosms and the population of resident pseudomonads colonizing cucumber roots was investigated after 10 and 52 days. Both CHA0-Rif and CHA0-Rif/pME3424 displaced a part of the resident pseudomonad population after 10 days. To investigate the population structure, utilization of 10 carbon sources and production of two exoenzymes was assessed for 5600 individual pseudomonad isolates and 1700 isolates were subjected to amplified ribosomal DNA restriction analysis of the spacer region (spacer-ARDRA). After 10 days, only the proportion of pseudomonads able to degrade -tryptophan was reduced in treatments inoculated with either biocontrol strain. In parallel the phenotypic diversity was reduced. These effects were only observed 10 days after inoculation, and they were similar for inoculation with CHA0-Rif and CHA0-Rif/pME3424. Changes in the population structure of resident pseudomonads on cucumber roots during plant growth were more pronounced than changes due to the inoculants. The inoculants did not affect the genotypic diversity detected with spacer-ARDRA, but the genotypic fingerprints corresponded only partially to the phenotypic profiles. Overall CHA0-Rif had a small and transient impact on the population of resident pseudomonads and the effect was essentially the same for the genetically engineered derivative CHA0-

Comparative phylogeography in a specific and obligate pollination antagonism.

In specific and obligate interactions the nature and abundance of a given species can have important effects on the survival and population dynamics of associated organisms. In a phylogeographic framework, we therefore expect that the fates of organisms interacting specifically are also tightly interrelated. Here we investigate such a scenario by analyzing the genetic structures of species interacting in an obligate plant-insect pollination lure-and-trap antagonism, involving Arum maculatum (Araceae) and its specific psychodid (Diptera) visitors Psychoda phalaenoides and Psycha grisescens. Because the interaction is asymmetric (i.e., only the plant depends on the insect), we expect the genetic structure of the plant to be related with the historical pollinator availability, yielding incongruent phylogeographic patterns between the interacting organisms.Using insect mtDNA sequences and plant AFLP genome fingerprinting, we inferred the large-scale phylogeographies of each species and the distribution of genetic diversities throughout the sampled range, and evaluated the congruence in their respective genetic structures using hierarchical analyses of molecular variances (AMOVA). Because the composition of pollinator species varies in Europe, we also examined its association with the spatial genetic structure of the plant.Our findings indicate that while the plant presents a spatially well-defined genetic structure, this is not the case in the insects. Patterns of genetic diversities also show dissimilar distributions among the three interacting species. Phylogeographic histories of the plant and its pollinating insects are thus not congruent, a result that would indicate that plant and insect lineages do not share the same glacial and postglacial histories. However, the genetic structure of the plant can, at least partially, be explained by the type of pollinators available at a regional scale. Differences in life-history traits of available pollinators might therefore have influenced the genetic structure of the plant, the dependent organism, in this antagonistic interaction.

Structural, metamorphic and geochronological relations between the Zanskar Shear Zone and the Miyar Shear Zone (NW Indian Himalaya): Evidence for two distinct tectonic structures and implications for the evolution of the High Himalayan Crystalline of Zanskar

The geologic structures and metamorphic zonation of the northwestern Indian Himalaya contrast significantly with those in the central and eastern parts of the range, where the high-grade metamorphic rocks of the High Himalayan Crystalline (HHC) thrust southward over the weakly metamorphosed sediments of the Lesser Himalaya along the Main Central Thrust (MCT). Indeed, the hanging wall of the MCT in the NW Himalaya mainly consists of the greenschist facies metasediments of the Chamba zone, whereas HHC high-grade rocks are exposed more internally in the range as a large-scale dome called the Gianbul dome. This Gianbul dome is bounded by two oppositely directed shear zones, the NE-dipping Zanskar Shear Zone (ZSZ) on the northern flank and the SW-dipping Miyar Shear Zone (MSZ) on the southern limb. Current models for the emplacement of the HHC in NW India as a dome structure differ mainly in terms of the roles played by both the ZSZ and the MSZ during the tectonothermal evolution of the HHC. In both the channel flow model and wedge extrusion model, the ZSZ acts as a backstop normal fault along which the high-grade metamorphic rocks of the HHC of Zanskar are exhumed. In contrast, the recently proposed tectonic wedging model argues that the ZSZ and the MSZ correspond to one single detachment system that operates as a subhorizontal backthrust off of the MCT. Thus, the kinematic evolution of the two shear zones, the ZSZ and the MSZ, and their structural, metamorphic and chronological relations appear to be diagnostic features for discriminating the different models. In this paper, structural, metamorphic and geochronological data demonstrate that the MSZ and the ZSZ experienced two distinct kinematic evolutions. As such, the data presented in this paper rule out the hypothesis that the MSZ and the ZSZ constitute one single detachment system, as postulated by the tectonic wedging model. Structural, metamorphic and geochronological data are used to present an alternative tectonic model for the large-scale doming in the NW Indian Himalaya involving early NE-directed tectonics, weakness in the upper crust, reduced erosion at the orogenic front and rapid exhumation along both the ZSZ and the MSZ.

Oxygen-isotope thermometry of quartz-calcite veins - Unraveling the thermal-tectonic history of the subgreenschist facies Morcles nappe (Swiss Alps)

Combined structural analysis and oxygen isotope thermometry of syntectonic quartz-calcite fibrous veins can be used to correlate the thermal history of deformed rocks,vith specific structural and tectonic events. Results are presented for the Mercies nappe in the western Helvetic Alps, Switzerland, where mineral parageneses, illite `'crystallinity,'' and fluid inclusion chemistry record an apparent peak metamorphic temperature gradient that increased across the Morcles nappe from anchizonal conditions in the foreland to epizonal conditions in its hinterland root zone. Twenty-seven quartz-calcite veins were analyzed in this study in order to determine the temperatures of veining during formation and deformation of the nappe, Peak metamorphic temperatures ranged from approximate to 260 to 290 degrees C in the shallower, foreland localities and to approximate to 330 to 350 degrees C in the deeper, more hinterland localities at the end of S1-foliation formation, related to large-scale folding. Temperatures gradually decreased throughout the nappe during subsequent development of the S2 foliation and S3 crenulation cleavage, Uplift and erosion of the overlying nappe pile resulted in slow cooling of the Morcles nappe during the waning stages of the Alpine Orogeny. The dominant foliation-forming deformation of the Morcles nappe occurred at elevated temperatures over the course of 10 to 15 Ma. Combined structure-oxygen isotope analyses of quartz-calcite veins yield better temperature and temporal constraints on the thermal histories of subgreenschist vein-bearing tectonites than do other geothermometers.

The Besançon Galaxy Model renewed. I. Constraints on the Galactic thin disc evolution from Tycho data

Context. The understanding of Galaxy evolution can be facilitated by the use of population synthesis models, which allow to test hypotheses on the star formation history, star evolution, as well as chemical and dynamical evolution of the Galaxy. Aims. The new version of the Besanc¸on Galaxy Model (hereafter BGM) aims to provide a more flexible and powerful tool to investigate the Initial Mass Function (IMF) and Star Formation Rate (SFR) of the Galactic disc. Methods. We present a new strategy for the generation of thin disc stars which assumes the IMF, SFR and evolutionary tracks as free parameters. We have updated most of the ingredients for the star count production and, for the first time, binary stars are generated in a consistent way. We keep in this new scheme the local dynamical self-consistency as in Bienayme et al (1987). We then compare simulations from the new model with Tycho-2 data and the local luminosity function, as a first test to verify and constrain the new ingredients. The effects of changing thirteen different ingredients of the model are systematically studied. Results. For the first time, a full sky comparison is performed between BGM and data. This strategy allows to constrain the IMF slope at high masses which is found to be close to 3.0, excluding a shallower slope such as Salpeter"s one. The SFR is found decreasing whatever IMF is assumed. The model is compatible with a local dark matter density of 0.011 M pc−3 implying that there is no compelling evidence for significant amount of dark matter in the disc. While the model is fitted to Tycho2 data, a magnitude limited sample with V<11, we check that it is still consistent with fainter stars. Conclusions. The new model constitutes a new basis for further comparisons with large scale surveys and is being prepared to become a powerful tool for the analysis of the Gaia mission data.

Aurinkokeräimien käyttö turpeen kenttäkuivauksessa

Tämä tutkimus tehtiin osana Vapo Oy:n uuden turvetuotantotekniikan kehitystä. Kihniön Aitonevalle on rakennettu uuden turvetuotantotekniikan tutkimusalue, johon kuuluu muun muassa yksi lämmittämätön kuivatuskenttä sekä yksi aurinkolämmöllä lämmitetty kuivatuskenttä aurinkokeräimineen ja putkistoineen. Työn tavoitteena oli selvittää aurinkolämmöllä lämmitetyn kuivatuskentän tuotannon teho verrattuna lämmittämättömään kenttään. Toinen tavoite oli selvittää Aitonevan tutkimusalueella käytössä olevista aurinkokeräimistä turpeen kuivaustarkoitukseen parhaiten soveltuva keräin. Tuotantoa uudella menetelmällä tehtiin vuoden 2005 kesän ajan. Tuotantotehon ero pyrittiin selvittämään seuraamalla yksittäisten turvetuotantoerien eli satokiertojen kuivumista kosteusnäyttein ja toisaalta vertaamalla koko kesän aikana saatua tuotantoa. Aurinkokeräimien vertailu toteutettiin energiamäärä- ja hyötysuhdemittauksin. Lisäksi kuivatuskenttien lämpötiloja mitattiin kentässä tapahtuvan lämmönsiirron selvittämiseksi. Mittausten perusteella havaittiin, ettälämmitetyn ja lämmittämättömän kentän välillä on tutkimuksen aikaisella kenttärakenteella 6-8 % ero satokierron aikana haihdutetussa vesimäärässä. Tätä voidaanpitää odotuksia pienempänä. Kenttien lämpötilamittausten perusteella osoittautui, että kentän pintarakenne tulisi eristää maaperästä, koska kentän alle siirrettyä lämpöä siirtyy häviöinä kylmään pohjamaahan. Käytössä olleista aurinkokeräimistä parhaaksi osoittautui katettu kumimattokeräin niin hyötysuhteen kuin tehokkuudenkin puolesta. Työn aikana todettiin, että tutkimusta keräimien ja varsinkinkenttärakenteen suhteen tulee jatkaa tulevaisuudessa ennen aurinkokeräinkentän laajamittaisen käytön aloittamista.

Volcanic debris avalanche deposits of the upper Maronne valley (Cantal Volcano, France): evidence for contrasted formation and transport mechanisms

The deposits of two volcanic debris avalanches (VDA I and II) that occur in the upper Maronne valley, northwest sector of Cantal Volcano, France, were studied to establish their mechanisms of formation, transport and deposition. These two volcanic debris avalanches that clearly differ with regard to their structures, textures and extensions, exemplify the wide spectrum of events associated with large-scale sector collapse. VDA I is voluminous (similar to1 km(3) in the upper Maronne valley) and widespread. The deposits comprise two distinct facies: the block facies that forms the intermediate and upper part of the unit and the mixed facies that crops out essentially at the base of the unit. The block facies consists of more or less brecciated lava, block-and-ash-flow breccia and pumice-flow tuff megablocks set in breccias resulting from block disaggregation. Mixing and differential movements are almost absent in this part of the VDA. The mixed facies consists of breccias rich in fine particles that originate from block disagregation, as well as being picked up from the substratum during movement. Mixing and differential movements are predominant in this zone. Analysis of fractures on lava megablocks suggests that shear stress during the initial sliding is the principal cause of fracture. These data strongly indicate that VDA I is purely gravitational and argue for a model in which the initial sliding mass transforms into a flow due to differential in situ fragmentation caused by the shear stress. VDA II is restricted to low-topography areas. Its volume, in the studied area, is about 0.3 km(3). The deposits consist of brecciated, rounded blocks and megablocks set in a fine-grained matrix composed essentially of volcanic glass. This unit is stratified, with a massive layer that contains all the megablocks at the base and in the intermediate part, and in the upper part a normally graded layer that contains only blocks <1 m in size. The different lithologies present are totally mixed. These observations suggest that VDA II may be of the Bezymianny-type and that it underwent a flow transformation from a turbulent to a stratified flow consisting of a basal hyperconcentrated laminar body overlain by a dilute layer. (C) 2000 Elsevier Science B.V. All rights reserved.

Predicting human resting-state functional connectivity from structural connectivity

In the cerebral cortex, the activity levels of neuronal populations are continuously fluctuating. When neuronal activity, as measured using functional MRI (fMRI), is temporally coherent across 2 populations, those populations are said to be functionally connected. Functional connectivity has previously been shown to correlate with structural (anatomical) connectivity patterns at an aggregate level. In the present study we investigate, with the aid of computational modeling, whether systems-level properties of functional networks-including their spatial statistics and their persistence across time-can be accounted for by properties of the underlying anatomical network. We measured resting state functional connectivity (using fMRI) and structural connectivity (using diffusion spectrum imaging tractography) in the same individuals at high resolution. Structural connectivity then provided the couplings for a model of macroscopic cortical dynamics. In both model and data, we observed (i) that strong functional connections commonly exist between regions with no direct structural connection, rendering the inference of structural connectivity from functional connectivity impractical; (ii) that indirect connections and interregional distance accounted for some of the variance in functional connectivity that was unexplained by direct structural connectivity; and (iii) that resting-state functional connectivity exhibits variability within and across both scanning sessions and model runs. These empirical and modeling results demonstrate that although resting state functional connectivity is variable and is frequently present between regions without direct structural linkage, its strength, persistence, and spatial statistics are nevertheless constrained by the large-scale anatomical structure of the human cerebral cortex.

Blind docking of 260 protein-ligand complexes with EADock 2.0.

Molecular docking softwares are one of the important tools of modern drug development pipelines. The promising achievements of the last 10 years emphasize the need for further improvement, as reflected by several recent publications (Leach et al., J Med Chem 2006, 49, 5851; Warren et al., J Med Chem 2006, 49, 5912). Our initial approach, EADock, showed a good performance in reproducing the experimental binding modes for a set of 37 different ligand-protein complexes (Grosdidier et al., Proteins 2007, 67, 1010). This article presents recent improvements regarding the scoring and sampling aspects over the initial implementation, as well as a new seeding procedure based on the detection of cavities, opening the door to blind docking with EADock. These enhancements were validated on 260 complexes taken from the high quality Ligand Protein Database [LPDB, (Roche et al., J Med Chem 2001, 44, 3592)]. Two issues were identified: first, the quality of the initial structures cannot be assumed and a manual inspection and/or a search in the literature are likely to be required to achieve the best performance. Second the description of interactions involving metal ions still has to be improved. Nonetheless, a remarkable success rate of 65% was achieved for a large scale blind docking assay, when considering only the top ranked binding mode and a success threshold of 2 A RMSD to the crystal structure. When looking at the five-top ranked binding modes, the success rate increases up to 76%. In a standard local docking assay, success rates of 75 and 83% were obtained, considering only the top ranked binding mode, or the five top binding modes, respectively.