On the relationship between genomic regulatory element organization and gene regulatory dynamics.

In this project we study the relationship between genomic regulatory element organization and gene regulatory dynamics. This paper illustrates an approach to investigating this relationship based on the application of classical nonlinear system analysis techniques to a transcription level, statistical thermodynamical model like that used in Shea & Ackers (1985). Preliminary ideas presented at the ICMCM conference (Wolf & Eeckman, 1998) are developed in this manuscript. We show that, for prokaryotic gene circuits dominated by local promoter control, dynamical system behavior descriptors like the number and stability of equilibrium point steady states and their bifurcation potential can be largely determined from genomic organization (e.g. the number, type, and placement of regulatory protein binding sites). Concepts are illustrated on hypothetical gene regulation systems with one or two genes and varying numbers of regulatory protein binding sites (operators). Gene regulatory systems with a single gene and an arbitrary number of operator sites are shown to be globally stable, with the potential for having multiple equilibrium points and capable of bifurcating. A monomer-controlled gene regulation system with n operator sites is proven to have a maximum of 1+n/2 stable equilibria for even n, and (n+1)/2 for odd n, while a multimer-controlled, n operator site system is shown to have a maximum of 2+n/2 stable equilibria for even n, and (n+3)/2 for odd n. These results are applied to the design of a two-state switch using a gene regulation system with two operator sites. The question "what is the simplest possible gene regulation system capable of acting like a switch?" is answered. The paper ends with an analysis of a two-gene regulation system, the results of which point to the existence of a "soft-switching" mechanism that may account for the "on-off" hypothesized behavior of some gene networks.

Improved splice site detection in Genie.

We present an improved splice site predictor for the genefinding program Genie. Genie is based on a generalized Hidden Markov Model (GHMM) that describes the grammar of a legal parse of a multi-exon gene in a DNA sequence. In Genie, probabilities are estimated for gene features by using dynamic programming to combine information from multiple content and signal sensors, including sensors that integrate matches to homologous sequences from a database. One of the hardest problems in genefinding is to determine the complete gene structure correctly. The splice site sensors are the key signal sensors that address this problem. We replaced the existing splice site sensors in Genie with two novel neural networks based on dinucleotide frequencies. Using these novel sensors, Genie shows significant improvements in the sensitivity and specificity of gene structure identification. Experimental results in tests using a standard set of annotated genes showed that Genie identified 86% of coding nucleotides correctly with a specificity of 85%, versus 80% and 84% in the older system. In further splice site experiments, we also looked at correlations between splice site scores and intron and exon lengths, as well as at the effect of distance to the nearest splice site on false positive rates.

Integrating database homology in a probabilistic gene structure model.

We present an improved stochastic model of genes in DNA, and describe a method for integrating database homology into the probabilistic framework. A generalized hidden Markov model (GHMM) describes the grammar of a legal parse of a DNA sequence. Probabilities are estimated for gene features by using dynamic programming to combine information from multiple sensors. We show how matches to homologous sequences from a database can be integrated into the probability estimation by interpreting the likelihood of a sequence in terms of the bit-cost to encode a sequence given a homology match. We also demonstrate how homology matches in protein databases can be exploited to help identify splice sites. Our experiments show significant improvements in the sensitivity and specificity of gene structure identification when these new features are added to our gene-finding system, Genie. Experimental results in tests using a standard set of annotated genes showed that Genie identified 95% of coding nucleotides correctly with a specificity of 91%, and 77% of exons were identified exactly.

MTT: a software tool for quality control in sequence assembly.

A large-scale sequencing project requires a tool to control the quality of the input data because a sizable number of trace data may be of low quality. If these data are allowed to enter the sequence assembly pipeline, harm will be done. Hence, it is important to detect such data as soon as possible. MTT (Move-Track-Trim) is a software package analyzing the quality of the lanes. It subjects each lane to a series of tests, and if a lane does not pass all tests, it is flagged as a "bad" lane. The use has a chance to examine both the "good" and the "bad" lanes and reclassify a "bad" lane as "good," or vice versa. Alternatively, the user may decide to retrack the gel or get rid of some lanes altogether. As a by-product of the analysis, MTT performs other useful functions. It trims the lanes and compresses the lane files and moves them to the directories where assembly is carried out. It also generates some useful statistics describing the quality of the gel.

Principal component analysis and large-scale correlations in non-coding sequences of human DNA.

We have calculated a full set of second-order correlation functions of nucleotides in noncoding DNA. They are found to be independently invariant in regard to permutations of A and T, and also C and G. Considering correlation functions as a 4 x 4 matrix with a symmetrical basis, we have found the principal components-objects with zero cross-correlations. These three principal components are present the base compositions: (A + T - C - G), (A - T), (C - G). The long-range behavior of these principal components yields power-law dependencies with different critical exponents.

A generalized hidden Markov model for the recognition of human genes in DNA.

We present a statistical model of genes in DNA. A Generalized Hidden Markov Model (GHMM) provides the framework for describing the grammar of a legal parse of a DNA sequence (Stormo & Haussler 1994). Probabilities are assigned to transitions between states in the GHMM and to the generation of each nucleotide base given a particular state. Machine learning techniques are applied to optimize these probabilities using a standardized training set. Given a new candidate sequence, the best parse is deduced from the model using a dynamic programming algorithm to identify the path through the model with maximum probability. The GHMM is flexible and modular, so new sensors and additional states can be inserted easily. In addition, it provides simple solutions for integrating cardinality constraints, reading frame constraints, "indels", and homology searching. The description and results of an implementation of such a gene-finding model, called Genie, is presented. The exon sensor is a codon frequency model conditioned on windowed nucleotide frequency and the preceding codon. Two neural networks are used, as in (Brunak, Engelbrecht, & Knudsen 1991), for splice site prediction. We show that this simple model performs quite well. For a cross-validated standard test set of 304 genes [ftp:@www-hgc.lbl.gov/pub/genesets] in human DNA, our gene-finding system identified up to 85% of protein-coding bases correctly with a specificity of 80%. 58% of exons were exactly identified with a specificity of 51%. Genie is shown to perform favorably compared with several other gene-finding systems.

Asymmetric sigmoid non-linearity in the rat olfactory system.

The statistical relationship between multi-unit spike activity and simultaneously recorded local dendritic field potentials in the olfactory system of the waking rat was studied with chronically placed electrodes. The relationship had the form of a sigmoid increase in axonal firing probability conditional on the amplitude of dendritic potentials. These data were fitted with an asymmetric sigmoid curve previously derived from the Hodgkin-Huxley equations. The curve was fitted using non-linear regression to optimize its parameter: the maximal firing rate. The maximal rate also gave the steepness of the slope of the sigmoid. Pulse trains were recorded from excitatory and inhibitory neurons in the olfactory cortex (including the anterior olfactory nucleus, the prepyriform cortex and the lateral entorhinal area) as identified by the phase relations of the pulse probability and the dendritic potentials, and from the excitatory neurons in the bulb (the inhibitory granule cells do not give extracellularly detectable action potentials). All these neurons are known to interact in disynaptic negative feedback loops giving rise to oscillations. The same sigmoid function fit the data from both types of neurons in all locations. The curves for neurons in all parts of the olfactory cortex had a 3-fold higher slope and maximal value than the curves from bulbar neurons. The significances of this difference and of the asymmetric sigmoid are discussed in terms of models for olfactory oscillatory dynamics and pattern recognition.

Correlations between unit firing and EEG in the rat olfactory system.

The olfactory EEG of awake animals displays oscillatory bursts of activity in the gamma- (30-100 Hz) range. The bursts are correlated with inflow of air over the receptor layer in the nose. None of the inputs to the cortices that display these oscillations carries periodic signals in the gamma-range. Thus these bursts are generated locally, either by neuronal feedback interactions or by coupling of oscillatory neurons. In the first case if the oscillations are generated by negative feedback, then two classes of cells must exist: excitatory neurons and inhibitory neurons with the same frequency of oscillation but with a quarter cycle phase lag by the inhibitory cells from the excitatory cells. On the other hand, if the EEG's result from coupling of cells that are intrinsically oscillatory, there should be a broad but monomodal distribution of phase values. In order to determine the origin of these bursts, we performed simultaneous recordings of EEG and multi-unit spikes in the 4 parts of the olfactory system (olfactory bulb, anterior olfactory nucleus, prepyriform cortex and lateral entorhinal area) of awake and motivated rats. For each sample, the EEG and the multi-unit spikes were recorded from the same local neighborhood. The multi-unit electrode recorded pulses from the principal output neurons of the respective cortical areas. In all locations tested, the oscillations in pulse probabilities of firing were found to have the same frequency as the dominant EEG frequency. In all 4 structures two sets of cells were found. One set displayed pulses in phase with the EEG and the other set displayed pulses that led or lagged the EEG by approximately 1/4 cycle. These data confirm the negative feedback interaction model rather than the coupled oscillator model for the generation of the bursts in the olfactory system. The relevance of these findings to other cortical systems, in casu the visual cortex is discussed.