Genome-wide discovery of transcriptional modules from DNA sequence and gene expression

In this paper, we describe an approach for understanding transcriptional regulation from both gene expression and promoter sequence data. We aim to identify transcriptional modules--sets of genes that are co-regulated in a set of experiments, through a common motif profile. Using the EM algorithm, our approach refines both the module assignment and the motif profile so as to best explain the expression data as a function of transcriptional motifs. It also dynamically adds and deletes motifs, as required to provide a genome-wide explanation of the expression data. We evaluate the method on two Saccharomyces cerevisiae gene expression data sets, showing that our approach is better than a standard one at recovering known motifs and at generating biologically coherent modules. We also combine our results with binding localization data to obtain regulatory relationships with known transcription factors, and show that many of the inferred relationships have support in the literature.

Discovering molecular pathways from protein interaction and gene expression data

In this paper, we describe an approach for identifying 'pathways' from gene expression and protein interaction data. Our approach is based on the assumption that many pathways exhibit two properties: their genes exhibit a similar gene expression profile, and the protein products of the genes often interact. Our approach is based on a unified probabilistic model, which is learned from the data using the EM algorithm. We present results on two Saccharomyces cerevisiae gene expression data sets, combined with a binary protein interaction data set. Our results show that our approach is much more successful than other approaches at discovering both coherent functional groups and entire protein complexes.

Representation Dependence in Probabilistic Inference

Non-deductive reasoning systems are often representation dependent: representing the same situation in two different ways may cause such a system to return two different answers. Some have viewed this as a significant problem. For example, the principle of maximum entropyhas been subjected to much criticism due to its representation dependence. There has, however, been almost no work investigating representation dependence. In this paper, we formalize this notion and show that it is not a problem specific to maximum entropy. In fact, we show that any representation-independent probabilistic inference procedure that ignores irrelevant information is essentially entailment, in a precise sense. Moreover, we show that representation independence is incompatible with even a weak default assumption of independence. We then show that invariance under a restricted class of representation changes can form a reasonable compromise between representation independence and other desiderata, and provide a construction of a family of inference procedures that provides such restricted representation independence, using relative entropy.

Efficient Solution Algorithms for Factored MDPs

This paper addresses the problem of planning under uncertainty in large Markov Decision Processes (MDPs). Factored MDPs represent a complex state space using state variables and the transition model using a dynamic Bayesian network. This representation often allows an exponential reduction in the representation size of structured MDPs, but the complexity of exact solution algorithms for such MDPs can grow exponentially in the representation size. In this paper, we present two approximate solution algorithms that exploit structure in factored MDPs. Both use an approximate value function represented as a linear combination of basis functions, where each basis function involves only a small subset of the domain variables. A key contribution of this paper is that it shows how the basic operations of both algorithms can be performed efficiently in closed form, by exploiting both additive and context-specific structure in a factored MDP. A central element of our algorithms is a novel linear program decomposition technique, analogous to variable elimination in Bayesian networks, which reduces an exponentially large LP to a provably equivalent, polynomial-sized one. One algorithm uses approximate linear programming, and the second approximate dynamic programming. Our dynamic programming algorithm is novel in that it uses an approximation based on max-norm, a technique that more directly minimizes the terms that appear in error bounds for approximate MDP algorithms. We provide experimental results on problems with over 10^40 states, demonstrating a promising indication of the scalability of our approach, and compare our algorithm to an existing state-of-the-art approach, showing, in some problems, exponential gains in computation time.

Decomposing gene expression into cellular processes

We propose a probabilistic model for cellular processes, and an algorithm for discovering them from gene expression data. A process is associated with a set of genes that participate in it; unlike clustering techniques, our model allows genes to participate in multiple processes. Each process may be active to a different degree in each experiment. The expression measurement for gene g in array a is a sum, over all processes in which g participates, of the activity levels of these processes in array a. We describe an iterative procedure, based on the EM algorithm, for decomposing the expression matrix into a given number of processes. We present results on Yeast gene expression data, which indicate that our approach identifies real biological processes.

Rich probabilistic models for gene expression

Clustering is commonly used for analyzing gene expression data. Despite their successes, clustering methods suffer from a number of limitations. First, these methods reveal similarities that exist over all of the measurements, while obscuring relationships that exist over only a subset of the data. Second, clustering methods cannot readily incorporate additional types of information, such as clinical data or known attributes of genes. To circumvent these shortcomings, we propose the use of a single coherent probabilistic model, that encompasses much of the rich structure in the genomic expression data, while incorporating additional information such as experiment type, putative binding sites, or functional information. We show how this model can be learned from the data, allowing us to discover patterns in the data and dependencies between the gene expression patterns and additional attributes. The learned model reveals context-specific relationships, that exist only over a subset of the experiments in the dataset. We demonstrate the power of our approach on synthetic data and on two real-world gene expression data sets for yeast. For example, we demonstrate a novel functionality that falls naturally out of our framework: predicting the "cluster" of the array resulting from a gene mutation based only on the gene's expression pattern in the context of other mutations.

A high-throughput alphavirus-based expression cloning system for mammalian cells

We have developed a widely applicable functional genomics strategy based on alphavirus expression vectors. The technology allows for rapid identification of genes encoding a functional activity such as binding of a defined ligand. Complementary DNA (cDNA) libraries were expressed in mammalian cells following infection with recombinant Sindbis virus (SIN replicon particles), a member of the Alphavirus genus. Virus-infected cells that specifically bound a ligand of choice were isolated using fluorescence-activated cell sorting (FACS). Replication-competent, infective SIN replicon particles harboring the corresponding cDNA were amplified in a next step. Within one round of selection, viral clones encoding proteins recognized by monoclonal antibodies or Fc-fusion molecules could be isolated and sequenced. Moreover, using the same viral libraries, a plaque-lift assay was established that allowed the identification of secreted, intracellular, and membrane proteins.

Eosinophil cationic protein and eosinophil protein X in the nasal lavage of children during the first 4 weeks of life. SPACE Collaborative Study Team

Eosinophil cationic protein (ECP) and eosinophil protein X (EPX) are well established as markers of eosinophil activation. We analyzed ECP and EPX concentrations in nasal lavage fluids (NALF) of 378 neonates during their first 4 weeks of life. Inclusion criteria were a positive history of parental allergy and a positive skin prick test or specific IgE (RAST class > or = 2) against at least one out of a panel of common aeroallergens in one or both parents. Twenty-four infants with no history of parental allergy were used as controls. A volume of 2 ml of 0.9% saline was instilled into each nostril and immediately recovered by a suction device. ECP and EPX were analyzed by radioimmunoassay. In 65 samples of three consecutive lavages, EPX was detected in nine samples (13.8%) in the control group, whereas it was detected in 197/360 samples (54.7%) in the study population. The corresponding figures for ECP were 17/65 (26.2%) in the control group and 173/365 (47.4%) in the study group. Both proteins showed a skewed distribution (median/5-95th percentiles for ECP: 0 microg/l [0-69.4] and EPX: 6.6 microg/l [0-73.2]). The differences between the control group and the study group were statistically significant, regardless of the allergic disease of the parents. In children of allergic parents, activation proteins of the eosinophil granulocyte are released on the nasal mucosal surface in about 50% of the studied population at the age of 4 weeks. This early onset of eosinophil activation in the nasal respiratory epithelium may reflect a genetic predisposition to allergy or early exposure to allergens.

Off-label use of antipsychotic drugs

Despite the fact that most antipsychotics have only been formally evaluated for the treatment of schizophreniform disorder, schizophrenia, mania, and schizoaffective disorder (defined as "classical indications"), antipsychotics are widely used for the treatment of a broad range of symptoms and disorders. In this study, 173 patients who were having their prescriptions for antipsychotics filled at local pharmacies were interviewed. In 115 patients (66.5%), an antipsychotic was prescribed for off-label indications. Patients most often stated that they took antipsychotics as a tranquilizer or an anxiolytic. Neither gender, education, duration of treatment, nor efficacy of treatment showed an influence on the prescription practices for antipsychotics. In contrast, family status and side effects showed a significant influence. A classical indication was more often found in married and widowed patients than in unmarried or divorced ones. Patients in whom antipsychotics were prescribed for the treatment of schizophrenia, schizophreniform disorder, mania, or schizoaffective disorder experienced side effects more often than others. Age was also important for the indication of antipsychotics. Classical indications of antipsychotics were most often found in patients aged 30 to 49 years. In older patients (49-70 years), antipsychotics were almost exclusively used for off-label indications. In classical indications, clozapine was used more frequently (50%) than other antipsychotics. Melperone was primarily prescribed for off-label use.

A temperature-regulated replicon-based DNA expression system

We present a temperature-regulated, alphavirus replicon-based DNA expression system. The system is regulated by a viral temperature-sensitive RNA-dependent RNA replicase, creating a temperature-dependent RNA amplification loop. Because of this positive feedback, the system exhibits both low background and high inducibility. We observed 700-fold induction in transiently transfected cells, and over 104-fold induction in stably transfected cells. The high stringency of inducibility allowed the generation of stable cell lines expressing a highly toxic protein upon temperature shift. These data suggest that the present expression system could simplify bioprocess engineering strategies, especially in situations where the cloned protein has detrimental effects on host cell metabolism.

Efficient gene delivery into adult cardiomyocytes by recombinant Sindbis virus

Somatic gene therapy as a potential strategy for the treatment of myocardial diseases relies on an efficient gene transfer into cardiac muscle cells. The difficulty of delivering genes into adult cardiomyocytes exists not only in vivo but also in primary culture systems. Therefore, possibilities for ex vivo gene transfer and the in vitro study of physiological processes by reverse genetics are limited. We investigated the potential of an alphavirus-based vector system to transduce adult rat cardiomyocytes (ARC) in culture using a replication-deficient Sindbis virus (SIN) encoding beta-galactosidase (SIN-LacZ). Transduction efficiency depended on the virus concentration used, with expression of the reporter gene being detectable in up to 80% of cultured ARC as early as 24 h after infection. We observed a remarkably lower cytotoxicity of this viral vector in ARC than in other cells such as fibroblasts and neonatal cardiomyocytes. Additionally, no perceptible changes in the morphology of the nuclei or cytoskeleton were found in ARC 48 h after infection with SIN-LacZ. We conclude that SIN vectors are useful for gene delivery into adult cardiomyocytes and believe that improved versions of this viral system may be useful for cardiovascular gene therapy in the future.

Heterogeneity in hereditary pancreatitis

Hereditary pancreatitis (HP) is the most common form of chronic relapsing pancreatitis in childhood, and may account for approximately 25% of adult cases with chronic idiopathic pancreatitis. Recently, an arginine-histidine (R117H) mutation within the cationic trypsinogen gene was found in 5/5 families studied with HP. In this study we report on the results of linkage and direct mutational analysis for the common R117H mutation examined in 8 nonrelated families with hereditary pancreatitis. Two-point linkage analysis with the 7q35 marker D7S676, done initially in 4 families, yielded lod scores that were positive in 2, negative in one, and weakly positive in one. Direct mutational analysis of exon 3 of the cationic trypsinogen gene in 6 families showed that all symptomatic individuals tested were heterozygous for the R117H mutation. Also, several asymptomatic but at-risk relatives were found to be heterozygous for this mutation. Affected individuals in the remaining 2 families did not have the mutation. Radiation hybrid mapping using the Genebridge 4 panel assigned the trypsinogen gene to chromosome region 7q35, 2.9 cR distal to ETS WI-9353 and 3.8 cR proximal the dinucleotide repeat marker D7S676. The negative linkage and absence of the trypsinogen mutation in 2/8 families suggest locus heterogeneity in HP. Analysis of the R117H mutation is useful in identifying presymptomatic "at-risk" relatives and in genetic counseling. Also, it can be useful in identifying children and adults with isolated chronic idiopathic pancreatitis.

Familial multiple-system tauopathy with presenile dementia is localized to chromosome 17

An autosomal dominant presenile dementia affecting 39 individuals in a seven-generation, 383-member pedigree has been studied at Indiana University. In the affected members of this family, clinical symptoms occurred early in life, with an average age at onset of 48.8 years. The presenting clinical features include disequilibrium, neck stiffness, dysphagia, and memory loss. As the disease progresses, further cognitive decline, superior-gaze palsy, and dystaxia also are observed. The average duration from onset of symptoms to death is approximately 10 years. Neuropathologic studies of nine affected individuals showed neuronal loss in several areas of the CNS, as well as argentophilic tau-immunopositive inclusions in neurons and in oligodendroglia. A limited genomic screen by use of DNA samples from 28 family members localized the gene for this disorder to a 3-cM region on chromosome 17, between the markers THRA1 and D17S791. The gene for tau also was analyzed, through samples from the family.

Spheroid body myopathy revisited

Having reported spheroid body myopathy from Indiana (IN) inherited in an autosomal-dominant fashion several years ago, we now describe additional findings from the Oregon branch--briefly recorded earlier--and confirm earlier studies in another clinically affected IN member of this kinship demonstrating identical spheroid bodies within the myopathic muscle specimens. The spheroid bodies also contained increased amounts of desmin, alpha-B crystallin, and ubiquitin within muscle fibers. Our studies now have established that spheroid body myopathy is a member of the growing family of desminopathic neuromuscular conditions.

Mapping of the primary binding site of measles virus to its receptor CD46

The measles virus (MV) hemagglutinin binds to the complement control protein (CCP) CD46 primarily through the two external modules, CCP-I and -II. To define the residues involved in binding, 40 amino acids predicted to be solvent-exposed on the CCP-I-II module surface were changed to either alanine or serine. Altered proteins were expressed on the cell surface, and their abilities to bind purified MV particles, a soluble form of hemagglutinin (sH) and nine CD46-specific antibodies competing to different levels with sH attachment, were measured. All proteins retained, at least in part, MV and sH binding, but some completely lost binding to certain antibodies. Amino acids essential for binding of antibodies weakly or moderately competing with sH attachment are situated in the membrane-distal tip of CCP-I, whereas residues involved in binding of strongly sH competing antibodies cluster in the center of CCP-I (Arg-25, Asp-27) or in CCP-II (Arg-69, Asp-70). Both clusters face the same side of CCP-I-II and map close to amino acid exchanges impairing sH binding (E11A, R29A, P39A, and D70A) or MV binding (D70A and E84A) and to a six-amino acid loop, previously shown to be necessary for sH binding.

A 3D model for the measles virus receptor CD46 based on homology modeling, Monte Carlo simulations, and hemagglutinin binding studies

The two terminal complement control protein (CCP) modules of the CD46 glycoprotein mediate measles virus binding. Three-dimensional models for these two domains were derived based on the NMR structures of two CCP modules of factor H. Both CD46 modules are about 35 A long, and form a five-stranded antiparallel beta-barrel structure. Monte Carlo simulations, sampling the backbone torsion angles of the linker peptide and selecting possible orientations on the basis of minimal solvent-exposed hydrophobic area, were used to predict the orientation of CCP-I relative to CCP-II. We tested this procedure successfully for factor H. For CD46, three clusters of structures differing in the tilt angle of the two domains were obtained. To test these models, we mutagenized the CCP modules. Four proteins, two without an oligosaccharide chain and two with mutated short amino acid segments, reached the cell surface efficiently. Only the protein without the CCP-I oligosaccharide chain maintained binding to the viral attachment protein hemagglutinin. These results are consistent with one of our models and suggest that the viral hemagglutinin does not bind at the membrane-distal tip of CD46, but near the concave CCP-II interface region.

Random Worlds and Maximum Entropy

Given a knowledge base KB containing first-order and statistical facts, we consider a principled method, called the random-worlds method, for computing a degree of belief that some formula Phi holds given KB. If we are reasoning about a world or system consisting of N individuals, then we can consider all possible worlds, or first-order models, withdomain {1,...,N} that satisfy KB, and compute thefraction of them in which Phi is true. We define the degree of belief to be the asymptotic value of this fraction as N grows large. We show that when the vocabulary underlying Phi andKB uses constants and unary predicates only, we can naturally associate an entropy with each world. As N grows larger,there are many more worlds with higher entropy. Therefore, we can usea maximum-entropy computation to compute the degree of belief. This result is in a similar spirit to previous work in physics and artificial intelligence, but is far more general. Of equal interest to the result itself are the limitations on its scope. Most importantly, the restriction to unary predicates seems necessary. Although the random-worlds method makes sense in general, the connection to maximum entropy seems to disappear in the non-unary case. These observations suggest unexpected limitations to the applicability of maximum-entropy methods.