Pawel Daniluk

Interaction model based on local protein substructures generalizes to the entire structural enzyme-ligand space.

Chemogenomics is a new strategy in in silico drug discovery, where the ultimate goal is to understand molecular recognition for all molecules interacting with all proteins in the proteome. To study such cross interactions, methods that can generalize over proteins that vary greatly in sequence, structure, and function are needed. We present a general quantitative approach to protein-ligand binding affinity prediction that spans the entire structural enzyme-ligand space. The model was trained on a data set composed of all available enzymes cocrystallized with druglike ligands, taken from four publicly available interaction databases, for which a crystal structure is available. Each enzyme was characterized by a set of local descriptors of protein structure that describe the binding site of the cocrystallized ligand. The ligands in the training set were described by traditional QSAR descriptors. To evaluate the model, a comprehensive test set consisting of enzyme structures and ligands was manually curated. The test set contained enzyme-ligand complexes for which no crystal structures were available, and thus the binding modes were unknown. The test set enzymes were therefore characterized by matching their entire structures to the local descriptor library constructed from the training set. Both the training and the test set contained enzyme-ligand complexes from all major enzyme classes, and the enzymes spanned a large range of sequences and folds. The experimental binding affinities (p K i) ranged from 0.5 to 11.9 (0.7-11.0 in the test set). The induced model predicted the binding affinities of the external test set enzyme-ligand complexes with an r (2) of 0.53 and an RMSEP of 1.5. This demonstrates that the use of local descriptors makes it possible to create rough predictive models that can generalize over a wide range of protein targets.

Protein structure prediction center in CASP8

We present an outline of the Critical Assessment of Protein Structure Prediction (CASP) infrastructure implemented at the University of California, Davis, Protein Structure Prediction Center. The infrastructure supports selection and validation of prediction targets, collection of predictions, standard evaluation of submitted predictions, and presentation of results. The Center also supports information exchange relating to CASP experiments and structure prediction in general. Technical aspects of conducting the CASP8 experiment and relevant statistics are also provided. Proteins 2009.

Using multi-data hidden Markov models trained on local neighborhoods of protein structure to predict residue–residue contacts

MOTIVATION Correct prediction of residue-residue contacts in proteins that lack good templates with known structure would take ab initio protein structure prediction a large step forward. The lack of correct contacts, and in particular long-range contacts, is considered the main reason why these methods often fail. RESULTS We propose a novel hidden Markov model (HMM)-based method for predicting residue-residue contacts from protein sequences using as training data homologous sequences, predicted secondary structure and a library of local neighborhoods (local descriptors of protein structure). The library consists of recurring structural entities incorporating short-, medium- and long-range interactions and is general enough to reassemble the cores of nearly all proteins in the PDB. The method is tested on an external test set of 606 domains with no significant sequence similarity to the training set as well as 151 domains with SCOP folds not present in the training set. Considering the top 0.2 x L predictions (L = sequence length), our HMMs obtained an accuracy of 22.8% for long-range interactions in new fold targets, and an average accuracy of 28.6% for long-, medium- and short-range contacts. This is a significant performance increase over currently available methods when comparing against results published in the literature. AVAILABILITY http://predictioncenter.org/Services/FragHMMent/.

CASP6 Data Processing and Automatic Evaluation at the Protein Structure Prediction Center

We present a short overview of the system governing data processing and automatic evaluation of predictions in CASP6, implemented at the Livermore Protein Structure Prediction Center. The system incorporates interrelated facilities for registering participants, collecting prediction targets from crystallographers and NMR spectroscopists and making them available to the CASP6 participants, accepting predictions and providing their preliminary evaluation, and finally, storing and visualizing results. We have automatically evaluated predictions submitted to CASP6 using criteria and methods developed over the successive CASP experiments. Also, we have tested a new evaluation technique based on non‐rigid‐body type superpositions. Approximately the same number of predictions has been submitted to CASP6 as to all previous CASPs combined, making navigation through and understanding of the data particularly challenging. To facilitate this, we have substantially modernized all data handling procedures, including implementation of a dedicated relational database. An overview of our redesigned website is also presented (http://predictioncenter.org/casp6/). Proteins 2005;Suppl 7:19–23.

New tools and expanded data analysis capabilities at the protein structure prediction center

We outline the main tasks performed by the Protein Structure Prediction Center in support of the CASP7 experiment and provide a brief review of the major measures used in the automatic evaluation of predictions. We describe in more detail the software developed to facilitate analysis of modeling success over and beyond the available templates and the adopted Java‐based tool enabling visualization of multiple structural superpositions between target and several models/templates. We also give an overview of the CASP infrastructure provided by the Center and discuss the organization of the results web pages available through http://predictioncenter.org. Proteins 2007.

Protein Model Database

The phenomenal success of the genome sequencing projects reveals the power of completeness in revolutionizing biological science. Currently it is possible to sequence entire organisms at a time, allowing for a systemic rather than fractional view of their organization and the various genome-encoded functions. There is an international plan to move towards a similar goal in the area of protein structure. This will not be achieved by experiment alone, but rather by a combination of efforts in crystallography, NMR spectroscopy, and computational modeling. Only a small fraction of structures are expected to be identified experimentally, the remainder to be modeled. Presently there is no organized infrastructure to critically evaluate and present these data to the biological community. The goal of the Protein Model Database project is to create such infrastructure, including (1) public database of theoretically derived protein structures; (2) reliable annotation of protein model quality, (3) novel structure analysis tools, and (4) access to the highest quality modeling techniques available.

WeBIAS: a web server for publishing bioinformatics applications

BackgroundOne of the requirements for a successful scientific tool is its availability. Developing a functional web service, however, is usually considered a mundane and ungratifying task, and quite often neglected. When publishing bioinformatic applications, such attitude puts additional burden on the reviewers who have to cope with poorly designed interfaces in order to assess quality of presented methods, as well as impairs actual usefulness to the scientific community at large.ResultsIn this note we present WeBIAS—a simple, self-contained solution to make command-line programs accessible through web forms. It comprises a web portal capable of serving several applications and backend schedulers which carry out computations. The server handles user registration and authentication, stores queries and results, and provides a convenient administrator interface. WeBIAS is implemented in Python and available under GNU Affero General Public License. It has been developed and tested on GNU/Linux compatible platforms covering a vast majority of operational WWW servers. Since it is written in pure Python, it should be easy to deploy also on all other platforms supporting Python (e.g. Windows, Mac OS X). Documentation and source code, as well as a demonstration site are available at http://bioinfo.imdik.pan.pl/webias.ConclusionsWeBIAS has been designed specifically with ease of installation and deployment of services in mind. Setting up a simple application requires minimal effort, yet it is possible to create visually appealing, feature-rich interfaces for query submission and presentation of results.

Theoretical and Computational Aspects of Protein Structural Alignment

Computing alignments of proteins based on their structure is one of the fundamental tasks of bioinformatics. It is crucial in all kinds of comparative analysis as well as in performing evolutionary and functional classification. Whereas determination of sequence relationships is well founded in statistical models, there is still considerable uncertainty over how to describe geometric relationships between proteins. Continuous growth of structural databases calls for fast and reliable algorithmic methods, enabling one to effectively compute alignments of pairs and larger sets of protein molecules. Although such methodologies have been developed over the past two decades, there exist so-called “difficult similarities” which may include repeats, insertions or deletions, permutations and conformational changes. A brief overview of existing methodologies with emphasis on the different approaches to decomposition of structures into smaller fragments is followed by a presentation of a formalism of local descriptors of protein structures. A formal definition of the problem of computing optimal alignments accommodating aforementioned difficulties is presented along with an analysis of the computational complexity of its important variants. Examples of “difficult similarities” and practical aspects of protein structure comparison are discussed.

Implementation of a maximum clique search procedure on CUDA

We present a novel implementation of a Motzkin–Straus-based iterative clique-finding algorithm for GPUs. The well-known iterative approach is enhanced by an annealing variant, where better convergence is obtained by introducing an additional parameter that eliminates certain local maxima, and by an attenuation variant, where the search is repeated several times and known cliques are attenuated by reducing the edge weights. The proposed solution belongs to a global optimization class of methods. It is particularly well suited to large and/or dense graphs, and outperforms other popular clique-finding methods. Therefore, it could be useful in many practical applications related to graph representations of the structures and/or dynamics of complex systems. The proposed algorithm was chosen on the basis of its portability to GPUs. Our implementation includes optimizations aimed at maximizing utilization of GPU cores by delaying some auxiliary computations and performing them simultaneously with the main matrix-vector multiplication. It achieves an average speedup of up to