Carsten Peterson

Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks.

The purpose of this study was to develop a method of classifying cancers to specific diagnostic categories based on their gene expression signatures using artificial neural networks (ANNs). We trained the ANNs using the small, round blue-cell tumors (SRBCTs) as a model. These cancers belong to four distinct diagnostic categories and often present diagnostic dilemmas in clinical practice. The ANNs correctly classified all samples and identified the genes most relevant to the classification. Expression of several of these genes has been reported in SRBCTs, but most have not been associated with these cancers. To test the ability of the trained ANN models to recognize SRBCTs, we analyzed additional blinded samples that were not previously used for the training procedure, and correctly classified them in all cases. This study demonstrates the potential applications of these methods for tumor diagnosis and the identification of candidate targets for therapy.

BioArray Software Environment (BASE): a platform for comprehensive management and analysis of microarray data.

The microarray technique requires the organization and analysis of vast amounts of data. These data include information about the samples hybridized, the hybridization images and their extracted data matrices, and information about the physical array, the features and reporter molecules. We present a web-based customizable bioinformatics solution called BioArray Software Environment (BASE) for the management and analysis of all areas of microarray experimentation. All software necessary to run a local server is freely available.

A NEW METHOD FOR MAPPING OPTIMIZATION PROBLEMS ONTO NEURAL NETWORKS

A novel modified method for obtaining approximate solutions to difficult optimization problems within the neural network paradigm is presented. We consider the graph partition and the travelling salesman problems. The key new ingredient is a reduction of solution space by one dimension by using graded neurons, thereby avoiding the destructive redundancy that has plagued these problems when using straightforward neural network techniques. This approach maps the problems onto Potts glass rather than spin glass theories. A systematic prescription is given for estimating the phase transition temperatures in advance, which facilitates the choice of optimal parameters. This analysis, which is performed for both serial and synchronous updating of the mean field theory equations, makes it possible to consistently avoid chaotic behavior.When exploring this new technique numerically we find the results very encouraging; the quality of the solutions are in parity with those obtained by using optimally tuned simulated annealing heuristics. Our numerical study, which for TSP extends to 200-city problems, exhibits an impressive level of parameter insensitivity. (Less)

Random Boolean network models and the yeast transcriptional network

The recently measured yeast transcriptional network is analyzed in terms of simplified Boolean network models, with the aim of determining feasible rule structures, given the requirement of stable solutions of the generated Boolean networks. We find that, for ensembles of generated models, those with canalyzing Boolean rules are remarkably stable, whereas those with random Boolean rules are only marginally stable. Furthermore, substantial parts of the generated networks are frozen, in the sense that they reach the same state, regardless of initial state. Thus, our ensemble approach suggests that the yeast network shows highly ordered dynamics.

Clustering ECG complexes using Hermite functions and self-organizing maps

An integrated method for clustering of QRS complexes is presented which includes basis function representation and self-organizing neural networks (NNs). Each QRS complex is decomposed into Hermite basis functions and the resulting coefficients and width parameter are used to represent the complex. By means of this representation, unsupervised self-organizing NNs are employed to cluster the data into 25 groups. Using the MIT-BIH arrhythmia database, the resulting clusters are found to exhibit a very low degree of misclassification (1.5%). The integrated method outperforms, on the MIT-BIH database, both a published supervised learning method as well as a conventional template cross-correlation clustering method.

The Intrinsic Charm of the Proton

Recent data give unexpectedly large cross-sections for charmed particle production at high xF in hadron collisions. This may imply that the proton has a non-negligible uudcc Fock component. The interesting consequences of such a hypothesis are explored.

Genetic networks with canalyzing Boolean rules are always stable

We determine stability and attractor properties of random Boolean genetic network models with canalyzing rules for a variety of architectures. For all power law, exponential, and flat in-degree distributions, we find that the networks are dynamically stable. Furthermore, for architectures with few inputs per node, the dynamics of the networks is close to critical. In addition, the fraction of genes that are active decreases with the number of inputs per node. These results are based upon investigating ensembles of networks using analytical methods. Also, for different in-degree distributions, the numbers of fixed points and cycles are calculated, with results intuitively consistent with stability analysis; fewer inputs per node implies more cycles, and vice versa. There are hints that genetic networks acquire broader degree distributions with evolution, and hence our results indicate that for single cells, the dynamics should become more stable with evolution. However, such an effect is very likely compensated for by multicellular dynamics, because one expects less stability when interactions among cells are included. We verify this by simulations of a simple model for interactions among cells.

Stem Cell States, Fates, and the Rules of Attraction

Understanding cell-fate decisions in stem cell populations is a major goal of modern biology. Stem and progenitor cell populations are often heterogeneous, which may reflect stem cell subsets that express subtly different properties, including different propensities for lineage selection upon differentiation, yet remain able to interconvert. We discuss these properties with examples both from the hematopoietic and embryonic stem cell (ESC) systems. The nature of the stem cell substates and their relationship to commitment to differentiate and lineage selection can be elucidated in terms of a landscape picture in which stable states can be defined mathematically as attractors.

Transcriptional dynamics of the embryonic stem cell switch

Recent ChIP experiments of human and mouse embryonic stem cells have elucidated the architecture of the transcriptional regulatory circuitry responsible for cell determination, which involves the transcription factors OCT4, SOX2, and NANOG. In addition to regulating each other through feedback loops, these genes also regulate downstream target genes involved in the maintenance and differentiation of embryonic stem cells. A search for the OCT4–SOX2–NANOG network motif in other species reveals that it is unique to mammals. With a kinetic modeling approach, we ascribe function to the observed OCT4–SOX2–NANOG network by making plausible assumptions about the interactions between the transcription factors at the gene promoter binding sites and RNA polymerase (RNAP), at each of the three genes as well as at the target genes. We identify a bistable switch in the network, which arises due to several positive feedback loops, and is switched on/off by input environmental signals. The switch stabilizes the expression levels of the three genes, and through their regulatory roles on the downstream target genes, leads to a binary decision: when OCT4, SOX2, and NANOG are expressed and the switch is on, the self-renewal genes are on and the differentiation genes are off. The opposite holds when the switch is off. The model is extremely robust to parameter changes. In addition to providing a self-consistent picture of the transcriptional circuit, the model generates several predictions. Increasing the binding strength of NANOG to OCT4 and SOX2, or increasing its basal transcriptional rate, leads to an irreversible bistable switch: the switch remains on even when the activating signal is removed. Hence, the stem cell can be manipulated to be self-renewing without the requirement of input signals. We also suggest tests that could discriminate between a variety of feedforward regulation architectures of the target genes by OCT4, SOX2, and NANOG.

Estrogen receptor beta expression is associated with tamoxifen response in ER alpha-negative breast carcinoma

Purpose: Endocrine therapies, such as tamoxifen, are commonly given to most patients with estrogen receptor (ERα)–positive breast carcinoma but are not indicated for persons with ERα-negative cancer. The factors responsible for response to tamoxifen in 5% to 10% of patients with ERα-negative tumors are not clear. The aim of the present study was to elucidate the biology and prognostic role of the second ER, ERβ, in patients treated with adjuvant tamoxifen. Experimental Design: We investigated ERβ by immunohistochemistry in 353 stage II primary breast tumors from patients treated with 2 years adjuvant tamoxifen, and generated gene expression profiles for a representative subset of 88 tumors. Results: ERβ was associated with increased survival (distant disease-free survival, P = 0.01; overall survival, P = 0.22), and in particular within ERα-negative patients (P = 0.003; P = 0.04), but not in the ERα-positive subgroup (P = 0.49; P = 0.88). Lack of ERβ conferred early relapse (hazard ratio, 14; 95% confidence interval, 1.8-106; P = 0.01) within the ERα-negative subgroup even after adjustment for other markers. ERα was an independent marker only within the ERβ-negative tumors (hazard ratio, 0.44; 95% confidence interval, 0.21-0.89; P = 0.02). An ERβ gene expression profile was identified and was markedly different from the ERα signature. Conclusion: Expression of ERβ is an independent marker for favorable prognosis after adjuvant tamoxifen treatment in ERα-negative breast cancer patients and involves a gene expression program distinct from ERα. These results may be highly clinically significant, because in the United States alone, ∼10,000 women are diagnosed annually with ERα-negative/ERβ-positive breast carcinoma and may benefit from adjuvant tamoxifen.