Sander B. Nabuurs

Making Optimal Use of Empirical Energy Functions: Force-Field Parameterization in Crystal Space

Todays energy functions are not able yet to distinguish reliably between correct and almost correct protein models. Improving these near‐native models is currently a major bottle‐neck in homology modeling or experimental structure determination at low resolution. Increasingly accurate energy functions are required to complete the “last mile of the protein folding problem,” for example during a molecular dynamics simulation. We present a new approach to reach this goal. For 50 high resolution X‐ray structures, the complete unit cell was reconstructed, including disordered water molecules, counter ions, and hydrogen atoms. Simulations were then run at the pH at which the crystal was solved, while force‐field parameters were iteratively adjusted so that the damage done to the structures was minimal. Starting with initial parameters from the AMBER force field, the optimization procedure converged at a new force field called YAMBER (Yet Another Model Building and Energy Refinement force field), which is shown to do significantly less damage to X‐ray structures, often move homology models in the right direction, and occasionally make them look like experimental structures. Application of YAMBER during the CASP5 structure prediction experiment yielded a model for target 176 that was ranked first among 150 submissions. Due to its compatibility with the well‐established AMBER format, YAMBER can be used by almost any molecular dynamics program. The parameters are freely available from www.yasara.org/yamber. Proteins 2004.

RECOORD: A Recalculated Coordinate Database of 500 Proteins from the PDB Using Restraints from the BioMagResBank

State‐of‐the‐art methods based on CNS and CYANA were used to recalculate the nuclear magnetic resonance (NMR) solution structures of 500+ proteins for which coordinates and NMR restraints are available from the Protein Data Bank. Curated restraints were obtained from the BioMagResBank FRED database. Although the original NMR structures were determined by various methods, they all were recalculated by CNS and CYANA and refined subsequently by restrained molecular dynamics (CNS) in a hydrated environment. We present an extensive analysis of the results, in terms of various quality indicators generated by PROCHECK and WHAT_CHECK. On average, the quality indicators for packing and Ramachandran appearance moved one standard deviation closer to the mean of the reference database. The structural quality of the recalculated structures is discussed in relation to various parameters, including number of restraints per residue, NOE completeness and positional root mean square deviation (RMSD). Correlations between pairs of these quality indicators were generally low; for example, there is a weak correlation between the number of restraints per residue and the Ramachandran appearance according to WHAT_CHECK (r = 0.31). The set of recalculated coordinates constitutes a unified database of protein structures in which potential user‐ and software‐dependent biases have been kept as small as possible. The database can be used by the structural biology community for further development of calculation protocols, validation tools, structure‐based statistical approaches and modeling. The RECOORD database of recalculated structures is publicly available from http://www.ebi.ac.uk/msd/recoord. Proteins 2005.

Gaussian interaction profile kernels for predicting drug–target interaction

MOTIVATION The in silico prediction of potential interactions between drugs and target proteins is of core importance for the identification of new drugs or novel targets for existing drugs. However, only a tiny portion of all drug-target pairs in current datasets are experimentally validated interactions. This motivates the need for developing computational methods that predict true interaction pairs with high accuracy. RESULTS We show that a simple machine learning method that uses the drug-target network as the only source of information is capable of predicting true interaction pairs with high accuracy. Specifically, we introduce interaction profiles of drugs (and of targets) in a network, which are binary vectors specifying the presence or absence of interaction with every target (drug) in that network. We define a kernel on these profiles, called the Gaussian Interaction Profile (GIP) kernel, and use a simple classifier, (kernel) Regularized Least Squares (RLS), for prediction drug-target interactions. We test comparatively the effectiveness of RLS with the GIP kernel on four drug-target interaction networks used in previous studies. The proposed algorithm achieves area under the precision-recall curve (AUPR) up to 92.7, significantly improving over results of state-of-the-art methods. Moreover, we show that using also kernels based on chemical and genomic information further increases accuracy, with a neat improvement on small datasets. These results substantiate the relevance of the network topology (in the form of interaction profiles) as source of information for predicting drug-target interactions. AVAILABILITY Software and Supplementary Material are available at http://cs.ru.nl/~tvanlaarhoven/drugtarget2011/. CONTACT tvanlaarhoven@cs.ru.nl; elenam@cs.ru.nl. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Two divergent leptin paralogues in zebrafish (Danio rerio) that originate early in teleostean evolution.

We describe duplicate leptin genes in zebrafish (Danio rerio) that share merely 24% amino acid identity with each other and only 18% with human leptin. We were also able to retrieve a second leptin gene in medaka (Oryzias latipes). The presence of duplicate leptin genes in these two distantly related teleosts suggests that duplicate leptin genes are a common feature of teleostean fishes. Despite low primary sequence conservation, we are confident in assigning orthology between mammalian and zebrafish leptins for several reasons. First, both zebrafish leptins share their characteristic gene structure and display key features of conserved synteny with mammalian leptin genes. Secondly, the cysteine residues that make up leptins single disulphide bridge are equally spaced in mammalian and zebrafish leptins and are unique among all members of the class-I helical cytokine family. Thirdly, the zebrafish leptins cluster with other fish leptins and mammalian leptins in phylogenetic analysis, supported by high bootstrap values. Within the leptin cluster, leptin-b forms a separate clade with the leptin-b orthologue from medaka. Finally, our prediction of the tertiary structures shows that both leptins conform to the typical four alpha-helix bundle structure of the class-I alpha-helical cytokines. The zebrafish leptins are differentially expressed; the liver shows high leptin-a expression (in concordance with what we observed for carp leptins), while leptin-b is expressed at much lower levels, which are downregulated further upon fasting. The finding of duplicate leptin genes in teleosts adds to our understanding of the evolution of leptin physiology in the early vertebrate lineage.

DRESS: a database of refined solution NMR structures

Several studies have shown that biomolecular NMR structures are often of lower quality when compared to crystal structures, and consequently they are often excluded from structural analyses. We present a publicly available database of re‐refined NMR structures, exhibiting significantly improved quality. This database (available at http://www.cmbi.kun.nl/dress/) presents a uniformly refined and validated set of structural models that improves the value of these NMR structures as input for experimental and theoretical studies in many fields of research. Proteins 2004.

Basal laminar drusen caused by compound heterozygous variants in the CFH gene.

Age-related macular degeneration (AMD) is a multifactorial disease that is strongly associated with the Tyr402His variant in the complement factor H (CFH) gene. Drusen are hallmark lesions of AMD and consist of focal-inflammatory and/or immune-mediated depositions of extracellular material at the interface of the retinal pigment epithelium (RPE) and the Bruch membrane. We evaluated the role of CFH in 30 probands with early-onset drusen and identified heterozygous nonsense, missense, and splice variants in five families. The affected individuals all carried the Tyr402His AMD risk variant on the other allele. This supports an autosomal-recessive disease model in which individuals who carry a CFH mutation on one allele and the Tyr402His variant on the other allele develop drusen. Our findings strongly suggest that monogenic inheritance of CFH variants can result in basal laminar drusen in young adults, and this can progress to maculopathy and severe vision loss later in life.

ZNF674: A New Krüppel-Associated Box–Containing Zinc-Finger Gene Involved in Nonsyndromic X-Linked Mental Retardation

Array-based comparative genomic hybridization has proven to be successful in the identification of genetic defects in disorders involving mental retardation. Here, we studied a patient with learning disabilities, retinal dystrophy, and short stature. The family history was suggestive of an X-linked contiguous gene syndrome. Hybridization of full-coverage X-chromosomal bacterial artificial chromosome arrays revealed a deletion of ~1 Mb in Xp11.3, which harbors RP2, SLC9A7, CHST7, and two hypothetical zinc-finger genes, ZNF673 and ZNF674. These genes were analyzed in 28 families with nonsyndromic X-linked mental retardation (XLMR) that show linkage to Xp11.3; the analysis revealed a nonsense mutation, p.E118X, in the coding sequence of ZNF674 in one family. This mutation is predicted to result in a truncated protein containing the Kruppel-associated box domains but lacking the zinc-finger domains, which are crucial for DNA binding. We characterized the complete ZNF674 gene structure and subsequently tested an additional 306 patients with XLMR for mutations by direct sequencing. Two amino acid substitutions, p.T343M and p.P412L, were identified that were not found in unaffected individuals. The proline at position 412 is conserved between species and is predicted by molecular modeling to reduce the DNA-binding properties of ZNF674. The p.T343M transition is probably a polymorphism, because the homologous ZNF674 gene in chimpanzee has a methionine at that position. ZNF674 belongs to a cluster of seven highly related zinc-finger genes in Xp11, two of which (ZNF41 and ZNF81) were implicated previously in XLMR. Identification of ZNF674 as the third XLMR gene in this cluster may indicate a common role for these zinc-finger genes that is crucial to human cognitive functioning.

Stress and innate immunity in carp: corticosteroid receptors and pro-inflammatory cytokines.

The stress hormone cortisol is deeply involved in immune regulation in all vertebrates. Common carp (Cyprinus carpio L.) express four corticoid receptors that may modulate immune responses: three glucocorticoid receptors (GR); GR1, with two splice variants (GR1a and GR1b), GR2 and a single mineralocorticoid receptor (MR). All receptors are expressed as of 4 days post-fertilization and may thus play a critical role in development and functioning of the adult immune system. Immune tissues and cells predominantly express mRNA for GRs compared to mRNA for the MR. Three-dimensional protein structure modeling predicts, and transfection assays confirm that alternative splicing of GR1 does not influence the capacity to induce transcription of effector genes. When tested for cortisol activation, GR2 is the most sensitive corticoid receptor in carp, followed by the MR and GR1a and GR1b. Lipopolysacharide (LPS) treatment of head kidney phagocytes quickly induces GR1 expression and inhibits GR2 expression. Cortisol treatment in vivo enhances GR1a and MR mRNA expression, but only mildly, and cortisol treatment in vitro does not affect receptor expression of phagocytes. Cortisol has no direct effect on the LPS-induced receptor profile. Therefore, an immune rather than a stress stimulus regulates GR expression. Cortisol administered at stress levels to phagocytes in vitro significantly inhibits LPS-induced expression of the pro-inflammatory cytokines tumor necrosis factor alpha (TNF-alpha) and interleukin-12 (IL-12) (subunit p35) and of inducible nitric oxide synthase (iNOS) expression. A physiologically differential function for GR1 and GR2 in the immune response of fish to infection is indicated.

Traditional biomolecular structure determination by NMR spectroscopy allows for major errors.

One of the major goals of structural genomics projects is to determine the three-dimensional structure of representative members of as many different fold families as possible. Comparative modeling is expected to fill the remaining gaps by providing structural models of homologs of the experimentally determined proteins. However, for such an approach to be successful it is essential that the quality of the experimentally determined structures is adequate. In an attempt to build a homology model for the protein dynein light chain 2A (DLC2A) we found two potential templates, both experimentally determined nuclear magnetic resonance (NMR) structures originating from structural genomics efforts. Despite their high sequence identity (96%), the folds of the two structures are markedly different. This urged us to perform in-depth analyses of both structure ensembles and the deposited experimental data, the results of which clearly identify one of the two models as largely incorrect. Next, we analyzed the quality of a large set of recent NMR-derived structure ensembles originating from both structural genomics projects and individual structure determination groups. Unfortunately, a visual inspection of structures exhibiting lower quality scores than DLC2A reveals that the seriously flawed DLC2A structure is not an isolated incident. Overall, our results illustrate that the quality of NMR structures cannot be reliably evaluated using only traditional experimental input data and overall quality indicators as a reference and clearly demonstrate the urgent need for a tight integration of more sophisticated structure validation tools in NMR structure determination projects. In contrast to common methodologies where structures are typically evaluated as a whole, such tools should preferentially operate on a per-residue basis.

The precision of NMR structure ensembles revisited

Biomolecular structures provide the basis for many studies in research areas such as structure-based drug design and homology modeling. In order to use molecular coordinates it is important that they are reliable in terms of accurate description of the experimental data and in terms of the overall and local geometry. Besides these primary quality criteria an indication is needed for the uncertainty in the atomic coordinates that may arise from the dynamic behavior of the considered molecules as well as from experimental- and computational procedures.In contrast to the crystallographic B-factor, a good measure for the uncertainty in NMR-derived atomic coordinates is still not available. It has become clear in recent years that the widely used atomic Root Mean Square Deviation (RMSD), which is a measure for the precision of the data, overestimates the accuracy of NMR structure ensembles and therefore is a problematic measure for the uncertainty in the atomic coordinates.In this study we report a method that yields a more realistic estimate of the uncertainty in the atomic coordinates by maximizing the RMSD of an ensemble of structures, while maintaining the accordance with the experimentally derived data. The results indicate that the RMSD of most NMR structure ensembles can be significantly increased compromising neither geometric quality nor NMR data. This maximized RMSD therefore seems a better estimate of the true uncertainty in the atomic coordinates.