Search

The identification of genes involved in host-pathogen interactions is important for the elucidation of mechanisms of disease resistance and host susceptibility. A traditional way to classify the origin of genes sampled from a pool of mixed cDNA is through sequence similarity to known genes from either the pathogen or host organism or other closely related species. This approach does not work when the identified sequence has no close homologues in the sequence databases. In our previous studies, we classified genes using their codon frequencies. This method, however, explicitly required the prediction of CDS regions and thus could not be applied to sequences composed from the non-coding regions of genes. In this study, we show that the use of sliding-window triplet frequencies extends the application of the algorithm to both coding and non-coding sequences and also increases the prediction accuracy of a Support Vector Machine classifier from 95.6+/-0.3 to 96.5+/-0.2. Thus the use of the triplet frequencies increased the prediction accuracy of the new method by more than 20% compared to our previous approach. A functional analysis of sequences detected gene families having significantly higher or lower probability to be correctly classified compared to the average accuracy of the method is described. The server to perform classification of EST sequences using triplet frequencies is available at (URL: http://mips.gsf.de/proj/est3).

We investigated the role of the two calcium-binding proteins parvalbumin (PV) and calbindin D-28k (CB) in the locomotor activity and motor coordination using null-mutant mice for PV (PV-/-), CB (CB-/-) or both proteins (PV-/-CB-/-). These proteins are expressed in distinct, mainly non-overlapping populations of neurons of the central and peripheral nervous system and PV additionally in fast-twitch muscles. In a test measuring repeated locomotor activity during 18-20 days, the analysis revealed a slightly increased activity in mice lacking either protein, while the lack of both decreased the number of beams crossed during active periods. An increase in the characteristic speed during the first 8 days could be attributed to PV-deficiency, while the elimination of CB in CB-/- and double-KO mice decreased the percentage of fast movements at all time points. In the latter, additionally a reduction of the fastest speed was observed. The alterations in locomotor activity (fast movements, fastest speed) strongly correlate with the impairment in locomotor coordination in mice deficient for CB evidenced in the runway assay and the rotarod assay. The graded locomotor phenotype (CB>PV) is qualitatively correlated with alterations in Purkinje cell firing reported previously in these mice. The presence or absence of either protein did not affect the spontaneous locomotor activity when animals were placed in a novel environment and tested only once for 30 min. In summary, the lack of these calcium-binding proteins yields characteristic, yet distinct phenotypes with respect to locomotor activity and coordination.

There have recently been developments in the methods used to access the accuracy of the prediction and applicability domain of absorption, distribution, metabolism, excretion and toxicity models, and also in the methods used to predict the physicochemical properties of compounds in the early stages of drug development. The methods are classified into two main groups: those based on the analysis of similarity of molecules, and those based on the analysis of calculated properties. An analysis of octanol-water distribution coefficients is used to exemplify the consistency of estimated and calculated accuracy of the ALOGPS program (http://www.vcclab.org) to predict in-house and publicly available datasets.

Scaffold/matrix attachment regions (S/MARs) are essential for structural organization of the chromatin within the nucleus and serve as anchors of chromatin loop domains. A significant fraction of genes in Arabidopsis thaliana contains intragenic S/MAR elements and a significant correlation of S/MAR presence and overall expression strength has been demonstrated. In this study, we undertook a genome scale analysis of expression level and spatiotemporal expression differences in correlation with the presence or absence of genic S/MAR elements. We demonstrate that genes containing intragenic S/MARs are prone to pronounced spatiotemporal expression regulation. This characteristic is found to be even more pronounced for transcription factor genes. Our observations illustrate the importance of S/MARs in transcriptional regulation and the role of chromatin structural characteristics for gene regulation. Our findings open new perspectives for the understanding of tissue- and organ-specific regulation of gene expression.

A benchmark of several popular methods, Associative Neural Networks (ANN), Support Vector Machines (SVM), k Nearest Neighbors (kNN), Maximal Margin Linear Programming (MMLP), Radial Basis Function Neural Network (RBFNN), and Multiple Linear Regression (MLR), is reported for quantitative-structure property relationships (QSPR) of stability constants logK1 for the 1:1 (M:L) and logbeta2 for 1:2 complexes of metal cations Ag+ and Eu3+ with diverse sets of organic molecules in water at 298 K and ionic strength 0.1 M. The methods were tested on three types of descriptors: molecular descriptors including E-state values, counts of atoms determined for E-state atom types, and substructural molecular fragments (SMF). Comparison of the models was performed using a 5-fold external cross-validation procedure. Robust statistical tests (bootstrap and Kolmogorov-Smirnov statistics) were employed to evaluate the significance of calculated models. The Wilcoxon signed-rank test was used to compare the performance of methods. Individual structure-complexation property models obtained with nonlinear methods demonstrated a significantly better performance than the models built using multilinear regression analysis (MLRA). However, the averaging of several MLRA models based on SMF descriptors provided as good of a prediction as the most efficient nonlinear techniques. Support Vector Machines and Associative Neural Networks contributed in the largest number of significant models. Models based on fragments (SMF descriptors and E-state counts) had higher prediction ability than those based on E-state indices. The use of SMF descriptors and E-state counts provided similar results, whereas E-state indices lead to less significant models. The current study illustrates the difficulties of quantitative comparison of different methods: conclusions based only on one data set without appropriate statistical tests could be wrong.

The development of high-throughput technologies has generated the need for bioinformatics approaches to assess the biological relevance of gene networks. Although several tools have been proposed for analysing the enrichment of functional categories in a set of genes, none of them is suitable for evaluating the biological relevance of the gene network. We propose a procedure and develop a web-based resource (BIOREL) to estimate the functional bias (biological relevance) of any given genetic network by integrating different sources of biological information. The weights of the edges in the network may be either binary or continuous. These essential features make our web tool unique among many similar services. BIOREL provides standardized estimations of the network biases extracted from independent data. By the analyses of real data we demonstrate that the potential application of BIOREL ranges from various benchmarking purposes to systematic analysis of the network biology.

MfunGD (http://mips.gsf.de/genre/proj/mfungd/) provides a resource for annotated mouse proteins and their occurrence in protein networks. Manual annotation concentrates on proteins which are found to interact physically with other proteins. Accordingly, manually curated information from a protein-protein interaction database (MPPI) and a database of mammalian protein complexes is interconnected with MfunGD. Protein function annotation is performed using the Functional Catalogue (FunCat) annotation scheme which is widely used for the analysis of protein networks. The dataset is also supplemented with information about the literature that was used in the annotation process as well as links to the SIMAP Fasta database, the Pedant protein analysis system and cross-references to external resources. Proteins that so far were not manually inspected are annotated automatically by a graphical probabilistic model and/or superparamagnetic clustering. The database is continuously expanding to include the rapidly growing amount of functional information about gene products from mouse. MfunGD is implemented in GenRE, a J2EE-based component-oriented multi-tier architecture following the separation of concern principle.

The solubility of drugs and drug-like compounds has been the subject of extensive studies aimed at finding a way to predict solubility from molecular structure. The aqueous solubility of a drug is an important factor that influences its absorption, distribution and elimination in the body. Poor aqueous solubility often causes a drug to appear inactive and may cause other biological problems. Compound solubility in DMSO represents another serious problem in early stages of drug discovery. An appreciation of the factors affecting a compound's DMSO solubility could help in predicting the storage conditions and appropriateness of compounds for primary bioscreening programs. In silico procedures for estimation of water and DMSO solubility represent extremely useful tools for the drug discovery practitioners. In this review, we provide a critical discussion of in silico models for the prediction of DMSO and water solubility of drug-like compounds used for virtual screening. We describe the main tendencies in the field, "booming" approaches and unsolved problems. A critical analysis of the accuracy and applicability of methods is provided.

The development of on-line software tools is changing the way we traditionally perform our analysis in drug design, but will chemoinformatics be forever behind bioinformatics in this development?

The privacy of chemical structure is of paramount importance for the industrial sector, in particular for the pharmaceutical industry. At the same time, companies handle large amounts of physico-chemical and biological data that could be shared in order to improve our molecular understanding of pharmacokinetic and toxicological properties, which could lead to improved predictivity and shorten the development time for drugs, in particular in the early phases of drug discovery. The current study provides some theoretical limits on the information required to produce reverse engineering of molecules from generated descriptors and demonstrates that the information content of molecules can be as low as less than one bit per atom. Thus theoretically just one descriptor can be used to completely disclose the molecular structure. Instead of sharing descriptors, we propose to share surrogate data. The sharing of surrogate data is nothing else but sharing of reliably predicted molecules. The use of surrogate data can provide the same information as the original set. We consider the practical application of this idea to predict lipophilicity of chemical compounds and we demonstrate that surrogate and real (original) data provides similar prediction ability. Thus, our proposed strategy makes it possible not only to share descriptors, but also complete collections of surrogate molecules without the danger of disclosing the underlying molecular structures.

Most studies concerning expression data analyses usually exploit information on the variability of gene intensity across samples. This information is sensitive to initial data processing, which affects the final conclusions. However expression data contains scale-free information, which is directly comparable between different samples. We propose to use the pairwise ratio of gene expression values rather than their absolute intensities for a classification of expression data. This information is stable to data processing and thus more attractive for classification analyses. In proposed schema of data analyses only information on relative gene expression levels in each sample is exploited. Testing on publicly available datasets leads to superior classification results.

The identification of the genes that participate at the biological interface of two species remains critical to our understanding of the mechanisms of disease resistance, disease susceptibility and symbiosis. The sequencing of complementary DNA (cDNA) libraries prepared from the biological interface between two organisms provides an inexpensive way to identify the novel genes that may be expressed as a cause or consequence of compatible or incompatible interactions. Sequence classification and annotation of species origin typically use an orthology-based approach and require access to large portions of either genome, or a close relative. Novel species- or clade-specific sequences may have no counterpart within existing databases and remain ambiguous features. Here we present a web-service, Eclair, which utilizes support vector machines for the classification of the origin of expressed sequence tags stemming from mixed host cDNA libraries. In addition to providing an interface for the classification of sequences, users are presented with the opportunity to train a model to suit their preferred species pair. Eclair is freely available at http://eclair.btk.fi.

Internet technology offers an excellent opportunity for the development of tools by the cooperative effort of various groups and institutions. We have developed a multi-platform software system, Virtual Computational Chemistry Laboratory, http://www.vcclab.org, allowing the computational chemist to perform a comprehensive series of molecular indices/properties calculations and data analysis. The implemented software is based on a three-tier architecture that is one of the standard technologies to provide client-server services on the Internet. The developed software includes several popular programs, including the indices generation program, DRAGON, a 3D structure generator, CORINA, a program to predict lipophilicity and aqueous solubility of chemicals, ALOGPS and others. All these programs are running at the host institutes located in five countries over Europe. In this article we review the main features and statistics of the developed system that can be used as a prototype for academic and industry models.

Motivation: Any development of new methods for automatic functional annotation of proteins according to their sequences requires high-quality data (as benchmark) as well as tedious preparatory work to generate sequence parameters required as input data for the machine learning methods. Different program settings and incompatible protocols make a comparison of the analyzed methods difficult., Results: The MIPS Bacterial Functional Annotation Benchmark dataset (MIPS-BFAB) is a new, high-quality resource comprising four bacterial genomes manually annotated according to the MIPS functional catalogue (FunCat). These resources include precalculated sequence parameters, such as sequence similarity scores, InterPro domain composition and other parameters that could be used to develop and benchmark methods for functional annotation of bacterial protein sequences. These data are provided in XML format and can be used by scientists who are not necessarily experts in genome annotation., Availability: BFAB is available at http://mips.gsf.de/proj/bfab

Motivation: Discovery of host and pathogen genes expressed at the plant-pathogen interface often requires the construction of mixed libraries that contain sequences from both genomes. Sequence identification requires high-throughput and reliable classification of genome origin. When using single-pass cDNA sequences difficulties arise from the short sequence length, the lack of sufficient taxonomically relevant sequence data in public databases and ambiguous sequence homology between plant and pathogen genes., Results: A novel method is described, which is independent of the availability of homologous genes and relies on subtle differences in codon usage between plant and fungal genes. We used support vector machines (SVMs) to identify the probable origin of sequences. SVMs were compared to several other machine learning techniques and to a probabilistic algorithm (PF-IND) for expressed sequence tag (EST) classification also based on codon bias differences. Our software (Eclat) has achieved a classification accuracy of 93.1% on a test set of 3217 EST sequences from Hordeum vulgare and Blumeria graminis, which is a significant improvement compared to PF-IND (prediction accuracy of 81.2% on the same test set). EST sequences with at least 50 nt of coding sequence can be classified using Eclat with high confidence. Eclat allows training of classifiers for any host-pathogen combination for which there are sufficient classified training sequences., Availability: Eclat is freely available on the Internet (http://mips.gsf.de/proj/est) or on request as a standalone version., Contact: friedel@informatik.uni-muenchen.de.

Background: Detection of sequence homologues represents a challenging task that is important for the discovery of protein families and the reliable application of automatic annotation methods. The presence of domains in protein families of diverse function, inhomogeneity and different sizes of protein families create considerable difficulties for the application of published clustering methods., Results: Our work analyses the Super Paramagnetic Clustering (SPC) and its extension, global SPC (gSPC) algorithm. These algorithms cluster input data based on a method that is analogous to the treatment of an inhomogeneous ferromagnet in physics. For the SwissProt and SCOP databases we show that the gSPC improves the specificity and sensitivity of clustering over the original SPC and Markov Cluster algorithm (TRIBE-MCL) up to 30%. The three algorithms provided similar results for the MIPS FunCat 1.3 annotation of four bacterial genomes, Bacillus subtilis, Helicobacter pylori, Listeria innocua and Listeria monocytogenes. However, the gSPC covered about 12% more sequences compared to the other methods. The SPC algorithm was programmed in house using C++ and it is available at http://mips.gsf.de/proj/spc. The FunCat annotation is available at http://mips.gsf.de., Conclusion: The gSPC calculated to a higher accuracy or covered a larger number of sequences than the TRIBE-MCL algorithm. Thus it is a useful approach for automatic detection of protein families and unsupervised annotation of full genomes.

A DNA microarray can track the expression levels of thousands of genes simultaneously. Previous research has demonstrated that this technology can be useful in the classification of cancers. Cancer microarray data normally contains a small number of samples which have a large number of gene expression levels as features. To select relevant genes involved in different types of cancer remains a challenge. In order to extract useful gene information from cancer microarray data and reduce dimensionality, feature selection algorithms were systematically investigated in this study. Using a correlation-based feature selector combined with machine learning algorithms such as decision trees, naïve Bayes and support vector machines, we show that classification performance at least as good as published results can be obtained on acute leukemia and diffuse large B-cell lymphoma microarray data sets. We also demonstrate that a combined use of different classification and feature selection approaches makes it possible to select relevant genes with high confidence. This is also the first paper which discusses both computational and biological evidence for the involvement of zyxin in leukaemogenesis.

The ALOGPS 2.1 was developed to predict 1-octanol/water partition coefficients, logP, and aqueous solubility of neutral compounds. An exclusive feature of this program is its ability to incorporate new user-provided data by means of self-learning properties of Associative Neural Networks. Using this feature, it calculated a similar performance, RMSE = 0.7 and mean average error 0.5, for 2569 neutral logP, and 8122 pH-dependent logD(7.4), distribution coefficients from the AstraZeneca "in-house" database. The high performance of the program for the logD(7.4) prediction looks surprising, because this property also depends on ionization constants pKa. Therefore, logD(7.4) is considered to be more difficult to predict than its neutral analog. We explain and illustrate this result and, moreover, discuss a possible application of the approach to calculate other pharmacokinetic and biological activities of chemicals important for drug development., ((c) 2004 Wiley-Liss, Inc. and the American Pharmacists Association)

The Maximal Margin (MAMA) linear programming classification algorithm has recently been proposed and tested for cancer classification based on expression data. It demonstrated sound performance on publicly available expression datasets. We developed a web interface to allow potential users easy access to the MAMA classification tool. Basic and advanced options provide flexibility in exploitation. The input data format is the same as that used in most publicly available datasets. This makes the web resource particularly convenient for non-expert machine learning users working in the field of expression data analysis.

Evaluation of the ALOGPS, ACD Labs LogD, and PALLAS PrologD suites to calculate the log D distribution coefficient resulted in high root-mean-squared error (RMSE) of 1.0-1.5 log for two in-house Pfizer's log D data sets of 17,861 and 640 compounds. Inaccuracy in log P prediction was the limiting factor for the overall log D estimation by these algorithms. The self-learning feature of the ALOGPS (LIBRARY mode) remarkably improved the accuracy in log D prediction, and an rmse of 0.64-0.65 was calculated for both data sets.

Networks of GABAergic interneurons are of utmost importance in generating and promoting synchronous activity and are involved in producing coherent oscillations. These neurons are characterized by their fast-spiking rate and by the expression of the Ca(2+)-binding protein parvalbumin (PV). Alteration of their inhibitory activity has been proposed as a major mechanism leading to epileptic seizures and thus the role of PV in maintaining the stability of neuronal networks was assessed in knockout (PV-/-) mice. Pentylenetetrazole induced generalized tonic-clonic seizures in all genotypes, but the severity of seizures was significantly greater in PV-/- than in PV+/+ animals. Extracellular single-unit activity recorded from over 1000 neurons in vivo in the temporal cortex revealed an increase of units firing regularly and a decrease of cells firing in bursts. In the hippocampus, PV deficiency facilitated the GABA(A)ergic current reversal induced by high-frequency stimulation, a mechanism implied in the generation of epileptic activity. We postulate that PV plays a key role in the regulation of local inhibitory effects exerted by GABAergic interneurons on pyramidal neurons. Through an increase in inhibition, the absence of PV facilitates synchronous activity in the cortex and facilitates hypersynchrony through the depolarizing action of GABA in the hippocampus.

Motivation: Microarray data appear particularly useful to investigate mechanisms in cancer biology and represent one of the most powerful tools to uncover the genetic mechanisms causing loss of cell cycle control. Recently, several different methods to employ microarray data as a diagnostic tool in cancer classification have been proposed. These procedures take changes in the expression of particular genes into account but do not consider disruptions in certain gene interactions caused by the tumor. It is probable that some genes participating in tumor development do not change their expression level dramatically. Thus, they cannot be detected by simple classification approaches used previously. For these reasons, a classification procedure exploiting information related to changes in gene interactions is needed., Results: We propose a MAximal MArgin Linear Programming (MAMA) method for the classification of tumor samples based on microarray data. This procedure detects groups of genes and constructs models (features) that strongly correlate with particular tumor types. The detected features include genes whose functional relations are changed for particular cancer types. The proposed method was tested on two publicly available datasets and demonstrated a prediction ability superior to previously employed classification schemes., Availability: The MAMA system was developed using the linear programming system LINDO http://www.lindo.com. A Perl script that specifies the optimization problem for this software is available upon request from the authors.

The dopamine modulation of microiontophoretic application of antagonists of glutamate metabotropic synaptic transmission was studied in the sensory motor cortex of awake cats during instrumental conditioned reflex. The substances depressed the background and intensity of evoked impulse activity of pyramidal neurons of the sensory-motor cortex and provoked significant increases in the latency of impulse reaction and corresponding conditioned movements of animals. Simultaneous application of the antagonists and dopamine (DA) eliminated their depression of the background and evoked activity of neurons and decreased the latency of the impulse reactions and movements. Similar qualitative effects were observed in experiments with simultaneous application of biccuculine and (RS)-alpha-methyl-4-carboxyphenylglycine. It is supposed that the DA modulation in the brain consists in stabilizing the background and evoked activity of cortical neurons during reduced intensity of metabotropic glutamatergic synaptic transmission. Such modulation can be important when considering some pathological disorders of glutamatergic transmission.

This article analyses molecular property calculation resources available on the Internet. The first section summarizes the on-line database resources that could be useful to search molecular and biological properties of chemicals, and indicates some principal databases with physicochemical, thermochemical, toxicity, cancer and HIV data. The second section overviews popular standalone programs for calculation of molecular descriptors. Some of these programs can be downloaded for free and used as standalone applications for calculation of molecular descriptors. The third section describes on-line tools for the prediction of molecular properties, activities and calculation of molecular descriptors. Analysis of emerging tools that can be useful to developing new on-line servers for the prediction of molecular parameters and properties is also given.

Changes in conditioned reflex spike activity of neurons in the sensorimotor cortex were studied during microiontophoretic application of agonists and antagonists of glutamate and GABAergic transmission. The results of these experiments showed that the glutamate ionotropic receptors (AMPA and NMDA) of neurons in the sensorimotor cortex were intensely activated by the arrival of a conditioned signal in the cortex. This response included not only large pyramidal neurons of the deep cortical layers, but also the surrounding inhibitory interneurons. The existence of constant tonic inhibitory regulation of the activity of large pyramidal neurons by the surrounding inhibitory cells was demonstrated, along with the active involvement of this inhibition in organizing the excitatory responses of neurons in the sensorimotor cortex during a conditioned reflex.

The present study introduces an approach to automatic classification of extracellularly recorded action potentials of neurons. The classification of spike waveform is considered a pattern recognition problem of special segments of signal that correspond to the appearance of spikes. The spikes generated by one neuron should be recognized as members of the same class. The spike waveforms are described by the nonlinear oscillating model as an ordinary differential equation with perturbation, thus characterizing the signal distortions in both amplitude and phase. It is shown that the use of local variables reduces the problem of spike recognition to the separation of a mixture of normal distributions in the transformed feature space. We have developed an unsupervised iteration-learning algorithm that estimates the number of classes and their centers according to the distance between spike trajectories in phase space. This algorithm scans the learning set to evaluate spike trajectories with maximal probability density in their neighborhood. Following the learning, the procedure of minimal distance is used to perform spike recognition. Estimation of trajectories in phase space requires calculation of the first- and second-order derivatives, and integral operators with piecewise polynomial kernels were used. This provided the computational efficiency of the developed approach for real-time application as required by recordings in behaving animals and in human neurosurgical operations. The new method of spike sorting was tested on simulated and real data and performed better than other approaches currently used in neurophysiology.

The effects of iontophoretic application of dopamine and selective D1 or D2 dopamine receptor agonists and antagonists on impulse activity of neurons of the deep layers of the sensorimotor cortex of cat were investigated during performance of a conditioned paw movement task. The application of dopamine, Quinpirole (selective D2 receptor agonist) or SKF 38393 (selective D1 receptor agonist) increased both background (P<0.001) and evoked impulse activity (P<0.05 for selective agonists). Selective D2 and D1 receptor antagonists (Sulpiride and SKF 83566, respectively) both increased the latency of neural responses and significantly increased the latency of the conditioned paw movements (P<0.01). These data suggest that during natural physiological functions subcortical dopamine neurons provide facilitation of activity pyramidal neurons of sensorimotor cortex.

This article provides a systematic study of several important parameters of the Associative Neural Network (ASNN), such as the number of networks in the ensemble, distance measures, neighbor functions, selection of smoothing parameters, and strategies for the user-training feature of the algorithm. The performance of the different methods is assessed with several training/test sets used to predict lipophilicity of chemical compounds. The Spearman rank-order correlation coefficient and Parzen-window regression methods provide the best performance of the algorithm. If additional user data is available, an improved prediction of lipophilicity of chemicals up to 2-5 times can be calculated when the appropriate smoothing parameters for the neural network are selected. The detected best combinations of parameters and strategies are implemented in the ALOGPS 2.1 program that is publicly available at http://www.vcclab.org/lab/alogps.

Changes in conditioned impulse reactions of neurons in sensorimotor cortex were studied during microiontophoretic application of glutamatergic and GABA ergic agonistic and antagonistic drugs. It was shown that ionotropic glutamate receptors (AMPA and NMDA) are activated by a conditioned stimulus. Not only large pyramidal neurons of deep cortical layers but surrounding short-axon inhibitory interneurons are involved in the reaction. It was shown that the activity of pyramidal neurons is under a constant inhibitory control from surrounding interneurons. This inhibition is involved in organization of excitatory cortical responses during conditioning.

Associative neural network (ASNN) represents a combination of an ensemble of feed-forward neural networks and the k-nearest neighbor technique. This method uses the correlation between ensemble responses as a measure of distance amid the analyzed cases for the nearest neighbor technique. This provides an improved prediction by the bias correction of the neural network ensemble. An associative neural network has a memory that can coincide with the training set. If new data becomes available, the network further improves its predictive ability and provides a reasonable approximation of the unknown function without a need to retrain the neural network ensemble. This feature of the method dramatically improves its predictive ability over traditional neural networks and k-nearest neighbor techniques, as demonstrated using several artificial data sets and a program to predict lipophilicity of chemical compounds. Another important feature of ASNN is the possibility to interpret neural network results by analysis of correlations between data cases in the space of models. It is shown that analysis of such correlations makes it possible to provide "property-targeted" clustering of data. The possible applications and importance of ASNN in drug design and medicinal and combinatorial chemistry are discussed. The method is available on-line at http://www.vcclab.org/lab/asnn.

The molecular weight and electrotopological E-state indices were used to estimate by Artificial Neural Networks aqueous solubility for a diverse set of 1291 organic compounds. The neural network with 33-4-1 neurons provided highly predictive results with r(2) = 0.91 and RMS = 0.62. The used parameters included several combinations of E-state indices with similar properties. The calculated results were similar to those published for these data by Huuskonen (2000). However, in the current study only E-state indices were used without need of additional indices (the molecular connectivity, shape, flexibility and indicator indices) also considered in the previous study. In addition, the present neural network contained three times less hidden neurons. Smaller neural networks and use of one homogeneous set of parameters provides a more robust model for prediction of aqueous solubility of chemical compounds. Limitations of the developed method for prediction of large compounds are discussed. The developed approach is available online at http://www.lnh.unil.ch/~itetko/logp.

A new method, ALOGPS v 2.0 (http://www.lnh.unil.ch/~itetko/logp/), for the assessment of n-octanol/water partition coefficient, log P, was developed on the basis of neural network ensemble analysis of 12 908 organic compounds available from PHYSPROP database of Syracuse Research Corporation. The atom and bond-type E-state indices as well as the number of hydrogen and non-hydrogen atoms were used to represent the molecular structures. A preliminary selection of indices was performed by multiple linear regression analysis, and 75 input parameters were chosen. Some of the parameters combined several atom-type or bond-type indices with similar physicochemical properties. The neural network ensemble training was performed by efficient partition algorithm developed by the authors. The ensemble contained 50 neural networks, and each neural network had 10 neurons in one hidden layer. The prediction ability of the developed approach was estimated using both leave-one-out (LOO) technique and training/test protocol. In case of interseries predictions, i.e., when molecules in the test and in the training subsets were selected by chance from the same set of compounds, both approaches provided similar results. ALOGPS performance was significantly better than the results obtained by other tested methods. For a subset of 12 777 molecules the LOO results, namely correlation coefficient r(2)= 0.95, root mean squared error, RMSE = 0.39, and an absolute mean error, MAE = 0.29, were calculated. For two cross-series predictions, i.e., when molecules in the training and in the test sets belong to different series of compounds, all analyzed methods performed less efficiently. The decrease in the performance could be explained by a different diversity of molecules in the training and in the test sets. However, even for such difficult cases the ALOGPS method provided better prediction ability than the other tested methods. We have shown that the diversity of the training sets rather than the design of the methods is the main factor determining their prediction ability for new data. A comparative performance of the methods as well as a dependence on the number of non-hydrogen atoms in a molecule is also presented.

The current study introduces a new method, the volume learning algorithm (VLA), for the investigation of three-dimensional quantitative structure-activity relationships (QSAR) of chemical compounds. This method incorporates the advantages of comparative molecular field analysis (CoMFA) and artificial neural network approaches. VLA is a combination of supervised and unsupervised neural networks applied to solve the same problem. The supervised algorithm is a feed-forward neural network trained with a back-propagation algorithm while the unsupervised network is a self-organizing map of Kohonen. The use of both of these algorithms makes it possible to cluster the input CoMFA field variables and to use only a small number of the most relevant parameters to correlate spatial properties of the molecules with their activity. The statistical coefficients calculated by the proposed algorithm for cannabimimetic aminoalkyl indoles were comparable to, or improved, in comparison to the original study using the partial least squares algorithm. The results of the algorithm can be visualized and easily interpreted. Thus, VLA is a new convenient tool for three-dimensional QSAR studies.

A volume learning algorithm for artificial neural networks was developed to quantitatively describe the three-dimensional structure-activity relationships using as an example N-benzylpiperidine derivatives. The new algorithm combines two types of neural networks, the Kohonen and the feed-forward artificial neural networks, which are used to analyze the input grid data generated by the comparative molecular field approach. Selection of the most informative parameters using the algorithm helped to reveal the most important spatial properties of the molecules, which affect their biological activities. Cluster regions determined using the new algorithm adequately predicted the activity of molecules from a control data set.

It has been established that 5-HT(1A) receptors are expressed both presynaptically as autoreceptors by 5-HT containing neurones, and postsynaptically by a variety of other neurones. Activation of either somatodendritic 5-HT(1A) autoreceptors or postsynaptic 5-HT(1A) receptors induces hyperpolarisation and inhibition of action potential discharge of the neurones, but it is unclear whether 5-HT(1A) receptors are under a general tonic influence by 5-HT. In the present study, using single unit recordings from both anesthetized and non-anesthetized rats, we show that the activity of neurones in the medial prefrontal cortex is not altered by systemic administration of the selective 5-HT(1A) receptor antagonist, WAY 100635. In contrast, WAY 100635 increased the firing rate of 5-HT neurones in the dorsal raphe nucleus. Our findings indicate a tonic activation of presynaptic somatodendritic but not postsynaptic cortical 5-HT(1A) receptors.

In this paper we describe an Internet Java-based technology that allows scientists to make their analytical software available worldwide. The implementation of this technology is exemplified by programs for the calculation of the lipophilicity and water solubility of chemical compounds available at http://www.lnh.unil.ch/~itetko/logp. Both these molecular properties are key parameters in quantitative structure-activity relationship studies and are used to provide invaluable information for the overall understanding of the uptake distribution, biotransformation, and elimination of a wide variety of chemicals. The compounds can be analyzed in batch or single-compound mode. The single-compound analysis offers the possibility to compare our results with several popular lipophilicity calculation methods, including CLOGP, KOWWIN, and XLOGP. The chemical compounds are analyzed according to SMILES line notation that can be prepared with the JME molecular editor of Peter Ertl. Conversion to SMILES from 56 formats is also available using the molecular structure information interchange hub developed by Pat Walters and Matt Stahl.

This study demonstrates the practical application of the pattern grouping algorithm (PGA), presented in the companion paper (Tetko IV, Villa AEP. A pattern grouping algorithm for analysis of spatiotemporal patterns in neuronal spike trains. 1. Detection of repeated patterns. J. Neurosci. Methods 2000; accompanying article), to data sets including up to 30 simultaneously recorded spike trains. The analysis of a large network of simulated neurons shows that the incidence of patterns cannot be simply related to an increase in firing rates obtained after Hebbian learning. Patterns that disappeared and reappeared in the thalamus of anesthetized rats when the cerebral cortex was reversibly inactivated suggest that widespread cell assemblies contribute to the generation and propagation of precisely timed activity. In an another experiment multiple spike trains were recorded from the temporal cortex of freely moving rats performing a complex two-choice discrimination task. The presence or absence of particular patterns in the period preceding the cue was associated with changes in reaction time. In conclusion, neuronal network interactions may generate spatiotemporal firing patterns detectable by PGA. We provide evidence of such patterned activity associated with specific animal's behavior, thus suggesting the existence of complex temporal coding schemes in the higher nervous centers of the brain.

The existence of precise temporal relations in sequences of spike intervals, referred to as 'spatiotemporal patterns', is suggested by brain theories that emphasize the role of temporal coding. Specific analytical methods able to assess the significance of such patterned activity are extremely important to establish its function for information processing in the brain. This study proposes a new method called 'pattern grouping algorithm' (PGA), designed to identify and evaluate the statistical significance of patterns which differ from each other by a defined and small jitter in spike timing of the order of few ms. The algorithm performs a pre-selection of template patterns with a fast computational approach, optimizes the jitter for each spike in the template and evaluates the statistical significance of the pattern group using three complementary statistical approaches. Simulated data sets characterized by various types of known non stationarities are used for validation of PGA and for comparison of its performance to other methods. Applications of PGA to experimental data sets of simultaneously recorded spike trains are described in a companion paper (Tetko IV, Villa AEP. A pattern grouping algorithm for analysis of spatiotemporal patterns in neuronal spike trains. 2. Application to simultaneous single unit recordings. J Neurosci Method 2000; accompanying article).

In the adult rat most of basal forebrain cholinergic neurons (BFCN) express the low-affinity p75 nerve growth factor recceptor (NGFr). The immunotoxin 192 IgG-saporin (SAP) provokes a selective loss of NGFr-positive BFCN, somewhat similar to the loss of integrity of BFCN associated with human senile dementia of Alzheimer's type, whereas NGF exerts a trophic action on BFCN. Cortico-cortical interactions are modulated by cholinergic projections of BFCN and it is proposed that alterations of these projections by SAP and by NGF produce opposite effects. This hypothesis was tested by recording multiple local field potentials (LFPs) in the rat temporal cortex and applying bispectral analysis to measure phase-coupled frequencies, somewhat analogous to frequencies of resonance. Choline acetyltransferase (ChAT) activity was measured in the septal area in order to assess the effects of the treatments. NGF-treatment increased ChAT activity by 45% and frequencies of non-linear coupling were shifted towards frequencies higher than 70 Hz, thus suggesting the presence of increased functional interactions in the short range. By contrast, SAP provoked a decrease of nearly 40% in ChAT activity and an increase of phase-coupling in the low frequencies (< 50 Hz), being interpreted as a decreased functional cortico-cortical interaction. Bispectral analysis revealed features of the effect of BFCN on cortical activity that could not be observed by other means and offers as a valuable tool of study that could be extended to the EEG of Alzheimer's patients.

A method for predicting log P values for a diverse set of 1870 organic molecules has been developed based on atom-type electrotopological-state (E-state) indices and neural network modeling. An extended set of E-state indices, which included specific indices with a more detailed description of amino, carbonyl, and hydroxy groups, was used in the current study. For the training set of 1754 molecules the squared correlation coefficient and root-mean-squared error were r2 = 0.90 and RMS(LOO) = 0.46, respectively. Structural parameters which included molecular weight and 38 atom-type E-state indices were used as the inputs in 39-5-1 artificial neural networks. The results from multilinear regression analysis were r2 = 0.87 and RMS(LOO) = 0.55, respectively. For a test set of 35 nucleosides, 12 nucleoside bases, 19 drug compounds, and 50 general organic compounds (n = 116) not included in the training set, a predictive r2 = 0.94 and RMS = 0.41 were calculated by artificial neural networks. The results for the same set by multilinear regression were r2 = 0.86 and RMS = 0.72. The improved prediction ability of artificial neural networks can be attributed to the nonlinear properties of this method that allowed the detection of high-order relationships between E-state indices and the n-octanol/water partition coefficient. The present approach was found to be an accurate and fast method that can be used for the reliable estimation of log P values for even the most complex structures.

This article presents a self-organising multilayered iterative algorithm that provides linear and non-linear polynomial regression models thus allowing the user to control the number and the power of the terms in the models. The accuracy of the algorithm is compared to the partial least squares (PLS) algorithm using fourteen data sets in quantitative-structure activity relationship studies. The calculated data show that the proposed method is able to select simple models characterized by a high prediction ability and thus provides a considerable interest in quantitative-structure activity relationship studies. The software is developed using client-server protocol (Java and C++ languages) and is available for world-wide users on the Web site of the authors.

The current study introduces an approach for pattern recognition of drug manufacturers according to their HPLC trace impurity data. This method considers signals in phase space and accounts for two different types of noise: additive and perturbative. The pharmaceutical fingerprints are estimated as mean trajectories of HPLC trace impurity data and are used as reference models for recognition of new data by the minimal length classifier. The chromatographic trace organic impurity patterns collected from six different manufacturers of L-tryptophan are analyzed as an example. The prediction ability of the new method tested using three different cross-validation procedures remains about 95% even if the number of available data in the training sets decreases by 5 times. The accuracy of prediction in phase space is superior compared to results calculated using a Window Preprocessing method and artificial neural networks. The difference in performance between new and previous methods becomes more significant under particular conditions that are more adequate for practical application of the method. In addition, the current approach enables simple and comprehensive interpretation of the calculated results.

The present study proposes a general method for constructing pharmaceutical fingerprints in the analysis of HPLC trace organic impurity patterns. The approach considers signals in phase space and accounts for two different types of noise: additive and perturbative. The first type, additive noise, contributes to distortion of the absolute values of signal peaks. The second type, perturbative noise, contributes to variations of the retention times of signal peaks and distorts the time scale of the trace organic impurity patterns. The ability of the proposed approach to consider both types of noise significantly distinguishes it from existing methods of data analysis that are usually designed to treat only the additive noise. Analysis of the HPLC signals in phase space eliminates the problem of perturbation noise and enables detection and comparison of similar signal segments recorded at different retention times. The current study analyzes the chromatographic trace organic impurity patterns collected from six different manufacturers of L-tryptophan using three HPLC columns. For five manufacturers the variability of data recorded with the same column are in perfect agreement with the proposed model. A significant variance of parameters is detected for one manufacturer, thus indicating a possible change in its product consistency. The analysis in phase space is also used to explain the previously detected variability of HPLC signals across columns. The accompanying paper reports an application of the proposed approach for the pattern recognition of HPLC data.

Precise and repeated spike-train timings within and across neurons define spatiotemporal patterns of activity. Although the existence of these patterns in the brain is well established in several species, there has been no direct evidence of their influence on behavioral output. To address this question, up to 15 neurons were recorded simultaneously in the auditory cortex of freely moving rats while animals waited for acoustic cues in a Go/NoGo task. A total of 235 significant patterns were detected during this interval from an analysis of 13 hr of recording involving over 1 million spikes. Of particular interest were 129 (55%) patterns that were significantly associated with the type of response the animal made later, independent of whether the response was that prompted by the cue because the response occurred later and the cue was chosen randomly. Of these behavior-predicting patterns, half (59/129) were associated with an enhanced tendency to go in response to the stimulus, and for 11 patterns of this subset, trials including the pattern were followed by significantly faster reaction time than those lacking the pattern. The remaining behavior-predicting patterns were associated with an enhanced NoGo tendency. Overall mean discharge rates did not vary across trials. Hence, these data demonstrate that particular spatiotemporal patterns predict future behavioral responses. Such presignal activity could form templates for extracting specific sensory information, motor programs prespecifying preference for a particular act, and/or some intermediate, associative brain process.

The aim of this study was to determine the efficacy of atom-type electrotopological state indices for estimation of the octanol-water partition coefficient (log P) values in a set of 345 drug compounds or related complex chemical structures. Multilinear regression analysis and artificial neural networks were used to construct models based on molecular weights and atom-type electrotopological state indices. Both multilinear regression and artificial neural networks provide reliable log P estimations. For the same set of parameters, application of neural networks provided better prediction ability for training and test sets. The present study indicates that atom-type electrotopological state indices offer valuable parameters for fast evaluation of octanol-water partition coefficients that can be applied to screen large databases of chemical compounds, such as combinatorial libraries.

Microelectrode recordings were simultaneously performed at multiple sites in the medial geniculate body (MGB) of anesthetized cats, rats and guinea pigs. We studied the effect of cortical deactivation on the association of neural activity within the thalamus during spontaneous activity. The corticofugal influence was suppressed by temporary cooling of the auditory cortex. Pairs of spike trains recorded from the same electrode were distinguished from cases where units were in MGB but recorded with different electrodes. Time domain analyses included crosscorrelations and search for precise repetition of complex spatiotemporal firing patterns of reverberating thalamic circuits. As a complementary approach we performed bispectral analyses of simultaneously recorded local field potentials in order to uncover the frequency components of their power spectra which are non linearly coupled. All results suggest that new functional neuronal circuits might appear at the thalamic level in the absence of input from the cortex. The newly active intrathalamic connections would provide the necessary input to sustain the reverberating activity of thalamic cell assemblies and generate low frequency non-linear interactions. The dynamic control exerted by the cortex over the functional segregation of information processing carried out in the thalamus conforms with theoretical neural network studies and with the functional selectivity-adaptive filtering theory of thalamic neuronal assemblies. Although this general conclusion remains valid across species, specific differences are discussed in the frame of known differences of the microcircuitry elements.

We have developed a distributed software package which allows neurophysiologists to have at hand a set of network and graphic tools in order to perform the analysis of data collected from the electrical activity of neurons, studied in the experimental laboratory. This system is built around three major components. First, the experimental data are processed in order to select specific data files from the networked computer and display the spike trains as raster dots. The second part selects one type of analysis (e.g., cross-correlation, out of a set of possible choices) after configuration of some specific parameters. Finally, the third component deals with the results of data analyses, which are selected and displayed on a multi-parameter graphic "ring binder." We have used the Java programming language to implement this application. This recent object-oriented language offers platform-independent software tools including graphic package and standard network protocols., (Copyright 1998 Academic Press.)

Simultaneous single unit spike trains were recorded in the auditory cortex of freely moving rats performing a complex cognitive task. The experimental paradigm is based on a two-choice task (Go/Nogo) with a two-component (pitch and location) auditory stimulus lasting 500 ms. We report evidence that firstly functional interactions, measured by cross-correlation analysis, between single units in the auditory cortex are dynamically modified in the period preceding the onset of the auditory stimulation, referred to as the 'waiting period'. We secondly observed that spatio-temporal firing patterns both within, and across cell spike trains also tended to appear in the waiting period, several seconds before the actual stimulus delivery. These patterns indicate a very precise repetition of spike discharges separated by long intervals (up to several hundreds of milliseconds). No consistent changes in mean rate were observed. These results suggest that network activity in the auditory cortex is selectively modified in rate independent ways before the actual sensory stimulation. These modifications may reflect participation of recurrent neuronal networks in processes anticipating the expected sensory input.

The present study investigates an application of artificial neural networks (ANNs) for use in pharmaceutical fingerprinting. Several pruning algorithms were applied to decrease the dimension of the input parameter data set. A localized fingerprint region was identified within the original input parameter space from which a subset of input parameters was extracted leading to enhanced ANN performance. The present results confirm that ANNs can provide a fast, accurate, and consistent methodology applicable to pharmaceutical fingerprinting.

This study investigates the emerging possibilities of combining unsupervised and supervised learning in neural network ensembles. Such strategy is used to get an efficient partition of a noisy input data set in order to focus the training of neural networks on the most complex and informative domains of the data set and accelerate the learning phase. The proposed algorithm provides a good prediction accuracy using fewer cases from non-informative domains according to a correlative measure of dependency between cases of the training set. This measure takes into account internal relationships amid analyzed data and can be used to cluster neighbor cases in a multidimensional space and to filter out the outliers. The possible relation of the proposed algorithm to brain processing occurring in the thalamo-cortical pathway is discussed.