Search

Unfavorable physicochemical properties often cause drug failures. It is therefore important to take lipophilicity and water solubility into account early on in lead discovery. This study presents log D7 models built using Gaussian Process regression, Support Vector Machines, decision trees and ridge regression algorithms based on 14556 drug discovery compounds of Bayer Schering Pharma. A blind test was conducted using 7013 new measurements from the last months. We also present independent evaluations using public data. Apart from accuracy, we discuss the quality of error bars that can be computed by Gaussian Process models, and ensemble and distance based techniques for the other modelling approaches.

Mass spectrometry (MS) has become today's de‐facto standard for high‐throughput analysis in proteomics research. Its applications range from toxicity analysis to MS‐based diagnostics. Often, the time spent on the MS experiment itself is significantly less than the time necessary to interpret the measured signals, since the amount of data can easily exceed several gigabytes. In addition, automated analysis is hampered by baseline artifacts, chemical as well as electrical noise, and an irregular spacing of data points. Thus, filtering techniques originating from signal and image analysis are commonly employed to address these problems. Unfortunately, smoothing, base‐line reduction, and in particular a resampling of data points can affect important characteristics of the experimental signal. To overcome these problems, we propose a new family of wavelet functions based on the isotope wavelet, which is hand‐tailored for the analysis of mass spectrometry data. The resulting technique is theoretically well‐foun...

The theory of gene assembly in ciliates has much in common with the theory of sorting by reversal. Both model processes that are based on splicing, and have a fixed begin and end product. The main difference is the type of splicing operations used to obtain the end product from the begin product. In this overview paper we show that the concept of breakpoint graph, known from the theory of sorting by reversal, has many uses in the theory of gene assembly. Our aim is to present the material in an intuitive and informal manner to allow for an efficient introduction into the subject.

The electrical activity of our brain's neurons is controlled by voltage‐gated ion channels. Mutations in these ion channels have been recently associated with clinical conditions, so called channelopathies. The involved ion channels have been well characterised at a molecular and biophysical level. However, the impact of these mutations on neuron function have been only rudimentary studied. It remains unclear how operation and performance (in terms of input‐output characteristics and reliability) are affected. Here, I show how system identification techniques provide neuronal performance measures which allow to quantitatively asses the impact of channelopathies by comparing whole cell input‐output relationships. I illustrate the feasibility of this approach by comparing the effects on neuronal signalling of two human sodium channel mutations (NaV 1.1 W1204R, R1648H), linked to generalized epilepsy with febrile seizures, to the wild‐type NaV 1.1 channel.

Reverse‐engineering of genetic interaction networks is one of the primary goals of current bio‐informatics research. Accurate knowledge of these networks is of crucial concern for both biological and medical understanding of biological function, and is relevant to the investigation of diseases with genetic origin. In this study, five different measures of genetic interactions are compared with regard to their ability to detect genetic interactions. The results are validated by means of comparison with five databases of known interactions: KEGG, Reactome, BIND, HPRD, and IntAct. A striking finding is that there are substantial differences between the interaction databases with regard to the correspondence with interactions found from gene expression data. These difference overshadow the relatively minor differences between the different interaction measures. These results lead to two conclusions. First, the choice between different interaction measures does not appear do be a crucial factor, given the relatively minor impact of this choice compared to the differences between the results obtained with different interaction databases. Moreover, these findings suggest that gene expression by itself is insufficient as a data source for the detection of genetic interactions. This implies that the identification of other data sources and the integration of multiple data sources are vital issues in improving the quality of bio‐informatics approaches to detect genetic interactions.

In this paper we present a method for modelling gene regulatory networks by forming a consensus Bayesian network model from multiple microarray gene expression datasets. Our method is based on combining Bayesian network graph topologies and does not require any special pre‐processing of the datasets, such as re‐normalisation. We evaluate our method on a synthetic regulatory network and part of the yeast heat‐shock response regulatory network using publicly available yeast microarray datasets. Results are promising; the consensus networks formed provide a broader view of the potential underlying network, obtaining an increased true positive rate over networks constructed from a single data source.

Assuming that the dynamics of peptides and proteins can be described by Markov transitions between different configurational states, we present a method which can calculate the mean first passage time (MFPT) between sets of initial and final states. The method is described in detail and differences between this and a method commonly employed [1] are explained. It is shown that the proposed method is (i) more general in allowing sets of final states, (ii) significantly faster, (iii) more accurate since it does not involve the calculation of infinite summations. Particular attention is given to the biologically important case of multiple final states.

We present a new and completely parallel method for protein ligand docking. The potential of the docking target structure is obtained directly from the electron density derived through an ab initio computation. A large subregion of the crystal structure of Isocitrate Lyase, was selected as docking target. To allow the full ab initio treatment of this region special care was taken to assign optimal basis functions. The electrostatic potential is tested by docking a small charged molecule (succinate) into the binding site. The ab initio grid yields a superior result by producing the best binding orientation and position, and by recognizing it as the best. In contrast the same docking procedure, but using a classical point‐charge based potential, produces a number of additional incorrect binding poses, and does not recognize the correct pose as the best solution.

Minimal cut sets (MCSs) have been introduced to define perturbations in a metabolic network needed to repress the activity of a given target reaction. While this approach gives interesting insight into the structure of metabolic networks, it suffers from two major drawbacks that are limiting in practice. First, using MCSs becomes inadequate when we are interested in inhibiting only one direction of a reversible target reaction. Second, and more importantly, the computation of MCSs by currently available methods is limited by the combinatorial explosion of the number of elementary modes and hence is not practical for genomescale networks. To overcome these problems, we introduce the concept of minimal direction cuts. A minimal forward (resp. backward) direction cut (MFC) (resp. MBC) is a minimal set of reactions whose removal prevents the target reaction from carrying a flux in the forward (resp. backward) direction. Computing MFCs and MBCs does not require to determine beforehand the elementary modes of the network. Instead it can be based on the Farkas lemma for equality and inequality constraints. This leads to a new algorithm to identify minimal direction cuts, which can also be used for a direct computation of MCSs.

With many of the major technological issues overcome, the remaining hurdle to the widespread use of microarrays is analytical. With so many genes and often very few replicates, most commonly used statistical techniques are inadequate to the task. Here we present a new technique for identifying differentially expressed genes based on linear mixed effect models and cluster analysis, that combines the advantages of univariate and multivariate methods whilst beginning to take into account the correlation structure of the data. Our results with the Golub leukemia data set have been promising, with our method selecting not only the commonly identified important genes, but several other biologically significant genes as well.

International audience; Regulatory networks are not randomly connected. They are modular, scale-free networks and some motifs distribution is clearly different from random distribution. However, the evolutionary causes and consequences of this specific connectivity are mainly unknown. In this paper we propose Raevol, an integrative model to study the evolution of regulatory networks. While most existing models consider direct evolution of the regulatory network, Raevol integrates a realistic genotype-phenotype mapping where the genome undergo mutations that indirectly modify the genetic network. Moreover, the organisms are selected at the phenotype level (which is produced by the genome via the regulation network). Thus, in Raevol, the network only indirectly evolve and it can only be selected if its activity influences the phenotype. We plan to use this model to better understand the network evolution and to study the influence of networks topology on evolution.

Most of the existing methods for the identification of biologically interesting genes in a temporal expression profiling dataset do not fully exploit the temporal ordering in the dataset and are based on normality assumptions for the gene expression.In this paper, we introduce a Bayesian median regression model to detect genes whose temporal profile is significantly different across a number of biological conditions. The regression model is defined by a polynomial function where both time and condition effects as well as interactions between the two are included. MCMC‐based inference returns the posterior distribution of the polynomial coefficients. From this a simple Bayes factor test is proposed to test for significance.The estimation of the median rather than the mean, and within a Bayesian framework, increases the robustness of the method compared to a Hotelling T2‐test previously suggested. This is shown on simulated data and on muscular dystrophy gene expression data.

We present in this article a Mueller polarimeter for the assessment of superficial skin lesions. The system is capable of capturing polarized light images non‐invasively. There are two objectives in this paper. The first is to describe how to obtain a polarimetric image through a Mueller imaging system. The second objective is using the Markov random field model to segment theses polarimetric images. We are choosing the ICM algorithm for the optimisation of probability because it is speed and can converge in few numbers of iterations. The use of the Markovian model gives efficient results separating normal and pathologic skin (detection of lesions).

Many diseases can be traced to point mutations in the DNA coding for specific enzymes. These point mutations result in the change of one amino acid residue in the enzyme. We have developed a model using simple molecular orbital calculations which can be used to quantitatively determine the change in interaction between the enzyme's active site and necessary ligands upon mutation. We have applied this model to three hydroxylase proteins: phenylalanine hydroxylase, tyrosine hydroxylase, and tryptophan hydroxylase, and we have obtained excellent correlation between our results and observed disease symptoms. Furthermore, we are able to use this agreement as a baseline to screen other mutations which may also cause onset of disease symptoms. Our focus is on systems where the binding is due largely to dispersion, which is much more difficult to model inexpensively than pure electrostatic interactions. Our calculations are run in parallel on a sixteen processor cluster of 64‐bit Athlon processors.