Publications

Introduction Most analyses of high throughput cancer data represent tumors by ?atomistic? single-gene properties. Pathifier, a recently introduced method, characterizes a tumor in terms of ?coarse grained? pathway-based variables. Methods We applied Pathifier to study a very large dataset of 2000 breast cancer samples and 144 normal tissues. Pathifier uses known gene assignments to pathways and biological processes to calculate for each pathway and tumor a Pathway Deregulation Score (PDS). Individual samples are represented in terms of their PDSs calculated for several hundred pathways, and the samples of the data set are analyzed and stratified on the basis of their profiles over these ?coarse grained?, biologically meaningful variables. Results We identified nine tumor subtypes; a new subclass (comprising about 7% of the samples) exhibits high deregulation in 38 PKA pathways, induced by overexpression of the gene PRKACB. Another interesting finding is that basal tumors break into two subclasses, with low and high deregulation of a cluster of immune system pathways. High deregulation corresponds to higher concentrations of Tumor Infiltrating Lymphocytes, and the patients of this basal subtype have better prognosis. The analysis used 1000 ?discovery set? tumors; our results were highly reproducible on 1000 independent ?validation? samples. Conclusions The coarse-grained variables that represent pathway deregulation provide a basis for relevant, novel and robust findings for breast cancer. Our analysis indicates that in breast cancer reliable prognostic signatures are most likely to be obtained by treating separately different subgroups of the patients.

The number of Genome Wide Association Studies (GWAS) of schizophrenia is rapidly growing. However, the small effect of individual genes limits the number of reliably implicated genes, which are too few and too diverse to perform reliable pathway analysis; hence the biological roles of the genes implicated in schizophrenia are unclear. To overcome these limitations we combine GWAS with genome-wide expression data from human post-mortem brain samples of schizophrenia patients and controls, taking these steps: 1) Identify 36 GWAS-based genes which are expressed in our dataset. 2) Find a cluster of 19 genes with highly correlated expression. We show that this correlation pattern is robust and statistically significant. 3) GO-enrichment analysis of these 19 genes reveals significant enrichment of ion channels and calcium-related processes. This finding (based on analyzing a small number of coherently expressed genes) is validated and enhanced in two ways: First, the emergence of calcium channels and calcium signaling is corroborated by identifying proteins that interact with those encoded by the cluster of 19. Second, extend the 19 cluster genes into 1028 genes, whose expression is highly correlated with the cluster's average profile. When GO-enrichment analysis is performed on this extended set, many schizophrenia related pathways appear, with calcium-related processes enriched with high statistical significance. Our results give further, expression-based validation to GWAS results, support a central role of calcium-signaling in the pathogenesis of schizophrenia, and point to additional pathways potentially related to the disease.

Accurate prognosis and prediction of response to therapy are essential for personalized treatment of cancer. Even though many prognostic gene lists and predictors have been proposed, especially for breast cancer, highthroughput "omic" methods have so far not revolutionized clinical practice, and their clinical utility has not been satisfactorily established. Different prognostic gene lists have very few shared genes, the biological meaning of most signatures is unclear, and the published success rates are considered to be overoptimistic. This review examines critically the manner in which prognostic classifiers are derived using machine-learning methods and suggests reasons for the shortcomings and problems listed above. Two approaches that may hold hope for obtaining improved prognosis are presented. Both are based on using existing prior knowledge; one proposes combining molecular "omic" predictors with established clinical ones, and the second infers biologically relevant pathway deregulation scores for each tumor from expression data, and uses this representation to study and stratify individual tumors. Approaches such as the second one are referred to in the physics literature as "phenomenology"; they will, hopefully, play a significant role in future studies of cancer.

Systemic lupus erythematosus (SLE) is an autoimmune disease that can attack many different body organs; the triggering event is unknown. SLE has been associated with more than 100 different autoantibody reactivities - anti-dsDNA is prominent. Nevertheless, autoantibodies to dsDNA occur in only two-thirds of SLE patients. We previously reported the use of an antigen microarray to characterize SLE serology. We now report the results of an expanded study of serology in SLE patients and scleroderma (SSc) patients compared with healthy controls. The analysis validated and extended previous findings: two-thirds of SLE patients reacted to a large spectrum of self-molecules that overlapped with their reactivity to dsDNA; moreover, some SLE patients manifested a deficiency of natural IgM autoantibodies. Most significant was the finding that many SLE patients who were negative for autoantibodies to dsDNA manifested abnormal antibody responses to Epstein-Barr virus (EBV): these subjects made IgG antibodies to EBV antigens to which healthy subjects did not respond or they failed to make antibodies to EBV antigens to which healthy subjects did respond. This observation suggests that SLE may be associated with a defective immune response to EBV. The SSc patients shared many of these serological abnormalities with SLE patients, but differed from them in increased IgG autoantibodies to topoisomerase and centromere B; 84% of SLE patients and 58% of SSc patients could be detected by their abnormal antibodies to EBV. Hence an aberrant immune response to a ubiquitous viral infection such as EBV might set the stage for an autoimmune disease.

Cells cope with replication-blocking lesions via translesion DNA synthesis (TLS). TLS is carried out by low-fidelity DNA polymerases that replicate across lesions, thereby preventing genome instability at the cost of increased point mutations. Here we perform a two-stage siRNA-based functional screen for mammalian TLS genes and identify 17 validated TLS genes. One of the genes, NPM1, is frequently mutated in acute myeloid leukaemia (AML). We show that NPM1 (nucleophosmin) regulates TLS via interaction with the catalytic core of DNA polymerase-n (poln), and that NPM1 deficiency causes a TLS defect due to proteasomal degradation of poln. Moreover, the prevalent NPM1c+ mutation that causes NPM1 mislocalization in ∼30% of AML patients results in excessive degradation of poln. These results establish the role of NPM1 as a key TLS regulator, and suggest a mechanism for the better prognosis of AML patients carrying mutations in NPM1.

We propose a method, Temperature Integration, which allows an efficient calculation of free energy differences between two systems of interest, with the same degrees of freedom, which may have rough energy landscapes. The method is based on calculating, for each single system, the difference between the values of lnZ at two temperatures, using a Parallel Tempering procedure. If our two systems of interest have the same phase space volume, they have the same values of lnZ at high-T, and we can obtain the free energy difference between them, using the two single-system calculations described above. If the phase space volume of a system is known, our method can be used to calculate its absolute (versus relative) free energy as well. We apply our method and demonstrate its efficiency on a "toy model" of hard rods on a 1-dimensional ring.

The transmembrane neural cell adhesion receptor L1 is a Wnt/b-catenin target gene expressed in many tumor types. In human colorectal cancer, L1 localizes preferentially to the invasive front of tumors and when overexpressed in colorectal cancer cells, it facilitates their metastasis to the liver. In this study, we investigated genes that are regulated in human colorectal cancer and by the L1-NF-kB pathway that has been implicated in liver metastasis. c-Kit was the most highly suppressed gene in both colorectal cancer tissue and the L1-NF-kB pathway. c-Kit suppression that resulted from L1-mediated signaling relied upon NF-kB, which directly inhibited the transcription of SP1, a major activator of the c-Kit gene promoter. Reconstituting c-Kit expression in L1- transfected cells blocked the biological effects conferred by L1 overexpression in driving motility and liver metastasis. We found that c-Kit expression in colorectal cancer cells is associated with a more pronounced epithelial morphology, along with increased expression of E-cadherin and decreased expression of Slug. Although c-Kit overexpression inhibited the motility and metastasis of L1-expressing colorectal cancer cells, it enhanced colorectal cancer cell proliferation and tumorigenesis, arguing that separate pathways mediate tumorigenicity and metastasis by c-Kit. Our findings provide insights into how colorectal cancer metastasizes to the liver, the most common site of dissemination in this cancer.

Motivation: Real time quantitative polymerase chain reaction (qPCR) is an important tool in quantitative studies of DNA and RNA molecules; especially in transcriptome studies, where different primer combinations allow identification of specific transcripts such as splice variants or precursor messenger RNA. Several softwares that implement various rules for optimal primer design are available. Nevertheless, as designing qPCR primers needs to be done manually, the repeated task is tedious, time consuming and prone to errors.Results: We used a set of rules to automatically design all possible exon-exon and intron-exon junctions in the human and mouse transcriptomes. The resulting database is included as a track in the UCSC genome browser, making it widely accessible and easy to use.

Deregulated proliferation is a hallmark of cancer cells. Here, we show that microRNA-10b* is a master regulator of breast cancer cell proliferation and is downregulated in tumoural samples versus matched peritumoural counterparts. Two canonical CpG islands (5kb) upstream from the precursor sequence are hypermethylated in the analysed breast cancer tissues. Ectopic delivery of synthetic microRNA-10b* in breast cancer cell lines or into xenograft mouse breast tumours inhibits cell proliferation and impairs tumour growth in vivo, respectively. We identified and validated in vitro and in vivo three novel target mRNAs of miR-10b* (BUB1, PLK1 and CCNA2), which play a remarkable role in cell cycle regulation and whose high expression in breast cancer patients is associated with reduced disease-free survival, relapse-free survival and metastasis-free survival when compared to patients with low expression. This also suggests that restoration of microRNA-10b* expression might have therapeutic promise.

Embryonic stem cells (ESCs) maintain high genomic plasticity, which is essential for their capacity to enter diverse differentiation pathways. Posttranscriptional modifications of chromatin histones play a pivotal role in maintaining this plasticity. We now report that one such modification, monoubiquitylation of histone H2B on lysine 120 (H2Bub1), catalyzed by the E3 ligase RNF20, increases during ESC differentiation and is required for efficient execution of this process. This increase is particularly important for the transcriptional induction of relatively long genes during ESC differentiation. Furthermore, we identify the deubiquitinase USP44 as a negative regulator of H2B ubiquitylation, whose downregulation during ESC differentiation contributes to the increase in H2Bub1. Our findings suggest that optimal ESC differentiation requires dynamic changes in H2B ubiquitylation patterns, which must occur in a timely and well-coordinated manner.

Aberrant activation of kinases has emerged to be a key event along with tumor progression, maintenance of tumor phenotype and response to anticancer treatments. This study documents the existence of an oncogenic autoregulatory feedback loop that includes the polo-like kinase-2 (Snk/Plk2) and mutant p53 proteins. Plk2 protein binds to and phosphorylates mutant p53, thereby potentiating its oncogenic activities. Phosphorylated mutant p53 binds more efficiently to p300, consequently strengthening its own transcriptional activity. Plk2 gene is regulated at a transcriptional level by both wt- and mutant p53 proteins. This leads to growth suppression or enhanced cell proliferation and chemoresistance, respectively. In turn, the siRNA-mediated knockdown of either mutant p53 or Plk2 proteins significantly curtails the growth properties of tumor cells and their chemo-resistance to anticancer treatments. Therefore, this paper identifies a novel tumor network including Plk2 and mutant p53 proteins whose triggering in response to DNA damage might disclose important implications for the treatment of human cancers.

Gene expression-based prediction of genomic copy number aberrations in the chromosomal region 12q13 to 12q15 that is flanked by MDM2 and CDK4 identified Wnt inhibitory factor 1 (WIF1) as a candidate tumor suppressor gene in glioblastoma. WIF1 encodes a secreted Wnt antagonist and was strongly downregulated in most glioblastomas as compared with normal brain, implying deregulation of Wnt signaling, which is associated with cancer. WIF1 silencing was mediated by deletion (7/69, 10%) or epigenetic silencing by promoter hypermethylation (29/110, 26%). Co-amplification of MDM2 and CDK4 that is present in 10% of glioblastomas was associated in most cases with deletion of the whole genomic region enclosed, including the WIF1 locus. This interesting pathogenetic constellation targets the RB and p53 tumor suppressor pathways in tandem, while simultaneously activating oncogenic Wnt signaling. Ectopic expression of WIF1 in glioblastoma cell lines revealed a dose-dependent decrease of Wnt pathway activity. Furthermore, WIF1 expression inhibited cell proliferation in vitro, reduced anchorage-independent growth in soft agar, and completely abolished tumorigenicity in vivo. Interestingly, WIF1 overexpression in glioblastoma cells induced a senescence-like phenotype that was dose dependent. These results provide evidence that WIF1 has tumor suppressing properties. Downregulation of WIF1 in 75% of glioblastomas indicates frequent involvement of aberrant Wnt signaling and, hence, may render glioblastomas sensitive to inhibitors of Wnt signaling, potentially by diverting the tumor cells into a senescence-like state.

Normal cells require continuous exposure to growth factors in order to cross a restriction point and commit to cell-cycle progression. This can be replaced by two short, appropriately spaced pulses of growth factors, where the first pulse primes a process, which is completed by the second pulse, and enables restriction point crossing. Through integration of comprehensive proteomic and transcriptomic analyses of each pulse, we identified three processes that regulate restriction point crossing: (1) The first pulse induces essential metabolic enzymes and activates p53-dependent restraining processes. (2) The second pulse eliminates, via the PI3K/AKT pathway, the suppressive action of p53, as well as (3) sets an ERK-EGR1 threshold mechanism, which digitizes graded external signals into an all-or-none decision obligatory for S phase entry. Together, our findings uncover two gating mechanisms, which ensure that cells ignore fortuitous growth factors and undergo proliferation only in response to consistent mitogenic signals.

Background: Alzheimer's disease and Schizophrenia are two common diseases of the brain with significant differences in neuropathology, etiology and symptoms. This dissimilarity in the two diseases makes a comparison of the two ideal for detecting molecular substrates that are common to brain disorders in general. Methods: In this study, we compared gene expression profiles across multiple brain areas, taken postmortem from patients with well-characterized Alzheimer's disease and Schizophrenia, and from cognitively normal control group with no neuro- or psychopathology. Results: Although the totality of gene expression changes in the two diseases is dissimilar, a subset of genes appears to play a role in both diseases in specific brain regions. We find at Brodmann area 22, the superior temporal gyrus, a statistically significant number of genes with apparently disregulated expression in both diseases. Furthermore, we found genes that differentiate the two diseases from the control across multiple brain regions, and note that these genes were usually down-regulated. Brodmann area 8, part of the superior frontal cortex, is relatively abundant with them. Conclusion: We show overwhelming statistical evidence for Alzheimer's and Schizophrenia sharing a specific molecular background at the superior temporal gyrus. We suggest that impairment of the regulation of autophagy pathway is shared, in BA 22, by the two diseases.

A mutation within one allele of the p53 tumor suppressor gene can inactivate the remaining wild-type allele in a dominant-negative manner and in some cases can exert an additional oncogenic activity, known as mutant p53 gain of function (GOF). To study the role of p53 mutations in prostate cancer and to discriminate between the dominant-negative effect and the GOF activity of mutant p53, we measured, using microarrays, the expression profiles of three immortalized prostate epithelial cultures expressing wild-type, inactivated p53 or mutated p53. Analysis of these gene expression profiles showed that both inactivated p53 and p53 R175H mutant expression resulted in the upregulation of cell cycle progression genes. A second group, which was upregulated exclusively by mutant p53 R175H, was predominantly enriched in developmental genes. This group of genes included the Twist1, a regulator of metastasis and epithelial-mesenchymal transition (EMT). Twist1 levels were also elevated in metastatic prostate cancer-derived cell line DU145, in immortalized lung fibroblasts and in a subset of lung cancer samples, all in a mutant p53-dependent manner. p53 R175H mutant bearing immortalized epithelial cells showed typical features of EMT, such as higher expression of mesenchymal markers, lower expression of epithelial markers and enhanced invasive properties in vitro. The mechanism by which p53 R175H mutant induces Twist1 expression involves alleviation of the epigenetic repression. Our data suggest that Twist1 expression might be upregulated following p53 mutation in cancer cells.

The recent decade has witnessed a surge of physicists to biology. Some of the activities of the participating groups focus on bona fide physics questions, posed on biological systems (such as the physics of molecular motors, for example). Another kind of research in which physicists take part alongside computer scientists and applied mathematicians, deals with questions that are of direct interest to biologists; they come under the umbrella of computational and systems biology. The topic of these lectures lies at the most biological end of this spectrum, addressing problems of clinical relevance which were posed and initiated by biologists. The objective of these lectures is to help the curious physicist to learn and to understand more about this emerging, highly interdisciplinary field of research, by providing brief introductions to molecular biology and cancer research. This is followed by a cursory review of some recent research done by the "Domany group" and its collaborations with biological and clinical labs. Furthermore, we mention (mainly in footnotes) a small subset of studies in which physicists have contributed to this field during the past years. A more detailed review of recent contributions by physicists is beyond the scope of this introductory text. The introductory nature of these lecture notes naturally induces a strong bias regarding publications cited; consequently, these lecture notes do not provide a fair, historically correct and updated review of relevant literature.

Highly regenerative tissues such as blood must possess effective DNA damage responses (DDR) that balance long-term regeneration with protection from leukemogenesis. Hematopoietic stem cells (HSCs) sustain life-long blood production, yet their response to DNA damage remains largely unexplored. We report that human HSCs exhibit delayed DNA double-strand break rejoining, persistent γH2AX foci, and enhanced p53- and ASPP1-dependent apoptosis after γ-radiation compared to progenitors. p53 inactivation or Bcl-2 overexpression reduced radiation-induced apoptosis and preserved in vivo repopulating HSC function. Despite similar protection from irradiation-induced apoptosis, only Bcl-2-overexpressing HSCs showed higher self-renewal capacity, establishing that intact p53 positively regulates self-renewal independently from apoptosis. The reduced self-renewal of HSCs with inactivated p53 was associated with increased spontaneous γH2AX foci in secondary transplants of HSCs. Our data reveal distinct physiological roles of p53 that together ensure optimal HSC function: apoptosis regulation and prevention of γH2AX foci accumulation upon HSC self-renewal.

Obliterative bronchiolitis (OB) continues to be the major limitation to long-term survival after lung transplantation. The specific aetiology and pathogenesis of OB are not well understood. To explore the role of autoreactivity in OB, we spotted 751 different self molecules onto glass slides, and used these antigen microarrays to profile 48 human serum samples for immunoglobulin G (IgG) and IgM autoantibodies; 27 patients showed no or mild bronchiolitis obliterans syndrome (BOS; a clinical correlate of OB) and 15 patients showed medium to severe BOS. We now report that these BOS grades could be differentiated by a profile of autoantibodies binding to 28 proteins or their peptides. The informative autoantibody profile included down-regulation as well as up-regulation of both IgM and IgG specific reactivities. This profile was evaluated for robustness using a panel of six independent test patients. Analysis of the functions of the 28 informative self antigens showed that eight of them are connected in an interaction network involved in apoptosis and protein metabolism. Thus, a profile of autoantibodies may reflect pathological processes in the lung allograft, suggesting a role for autoimmunity in chronic rejection leading to OB.

The genetic profiling of B-cell malignancies is rapidly expanding, providing important information on the tumorigenic potential, response to treatment, and clinical outcome of these diseases. However, the relative contributions of inherent gene expression versus microenvironmental effects are poorly understood. The regulation of gene expression programs by means of adhesive interactions was studied here in ARH-77 human malignant B-cell variants, derived from the same cell line by selective adhesion to a fibronectin matrix. The populations included cells that adhere to fibronectin and are highly tumorigenic (designated "type A" cells) and cells that fail to adhere to fibronectin and fail to develop tumors in vivo ("type F" cells). To identify genes directly affected by cell adhesion to fibronectin, type A cells deprived of an adhesive substrate (designated "AF cells") were also examined. Bioinformatic analyses revealed a remarkable correlation between cell adhesion and both B-cell differentiation state and the expression of multiple myeloma (MM)-associated genes. The highly adherent type A cells expressed higher levels of NFκB-regulated genes, many of them associated with MM. Moreover, we found that the transcription of several MM-related proto-oncogenes is stimulated by adhesion to fibronectin. In contrast, type F cells, which display poor adhesive and tumorigenic properties, expressed genes associated with higher levels of B-cell differentiation. Our findings indicate that B-cell differentiation, as manifested by gene expression profiles, is attenuated by cell adhesion to fibronectin, leading to upregulation of specific genes known to be associated with the pathogenesis of MM.

ID4 (inhibitor of DNA binding 4) is a member of a family of proteins that function as dominant-negative regulators of basic helix-loop-helix transcription factors. Growing evidence links ID proteins to cell proliferation, differentiation and tumorigenesis. Here we identify ID4 as a transcriptional target of gain-of-function p53 mutants R175H, R273H and R280K. Depletion of mutant p53 protein severely impairs ID4 expression in proliferating tumor cells. The protein complex mutant p53-E2F1 assembles on specific regions of the ID4 promoter and positively controls ID4 expression. The ID4 protein binds to and stabilizes mRNAs encoding pro-angiogenic factors IL8 and GRO-α. This results in the increase of the angiogenic potential of cancer cells expressing mutant p53. These findings highlight the transcriptional axis mutant p53, E2F1 and ID4 as a still undefined molecular mechanism contributing to tumor neo-angiogenesis.

Prostate cancer (PC) is a heterogeneous disease whose aggressive phenotype is the second leading cause of cancer-related death in men. The identification of key molecules and pathways that play a pivotal role in PC progression towards an aggressive form is crucial. A major effort towards this end has been taken by global analyses of gene expression profiles. However, the large body of data did not provide a definitive idea about the genes which are associated with the aggressive growth of PC. In order to identify such genes, we performed an interspecies comparison between several human data sets and high quality microarray data that we generated from the transgenic adenocarcinoma of mouse prostate (TRAMP) strain. TheTRAMP PC mimics the histological and pathological appearance as well as the aggressive phenotype of human PC (huPC). Analysis of the microarray data, derived from microdissected TRAMP specimens removed at different stages of the disease yielded genetic signatures delineating the TRAMP PC development and progression. Comparison of the TRAMP data with a set of genes representing the core expression signature of huPC yielded a limited set genes. Some of these genes are known predictors of poor prognosis in huPC. Interestingly, the modulation of genes responsible for the invasive phenotype of huPC occurs in TRAMP already during the transition to prostate intraepithelial neoplasia (PIN) and onwards to localized tumors. We therefore suggest that critical oncogenic events leading to an aggressive phenotype of huPC can be studied in the PIN stage of TRAMP.

Aging is often associated with a decline in hippocampus-dependent spatial memory. Here, we show that functional cell-mediated immunity is required for the maintenance of hippocampus-dependent spatial memory. Sudden imposition of immune compromise in young mice caused spatial memory impairment, whereas immune reconstitution reversed memory deficit in immune-deficient mice. Analysis of hippocampal gene expression suggested that immune-dependent spatial memory performance was associated with the expression of insulin-like growth factor (Igf1) and of genes encoding proteins related to presynaptic activity (Syt10, Cplx2). We further showed that memory loss in aged mice could be attributed to age-related attenuation of the immune response and could be reversed by immune system activation. Homeostatic-driven proliferation of lymphocytes, which expands the existing T cell repertoire, restored spatial memory deficits in aged mice. Thus, our results identify a novel function of the immune system in the maintenance of spatial memory and suggest an original approach for arresting or reversing age-associated memory loss.

Purpose: Glioblastomas are notorious for resistance to therapy, which has been attributed to DNA-repair proficiency, a multitude of deregulated molecular pathways, and, more recently, to the particular biologic behavior of tumor stem-like cells. Here, we aimed to identify molecular profiles specific for treatment resistance to the current standard of care of concomitant chemoradiotherapy with the alkylating agent temozolomide. Patients and Methods: Gene expression profiles of 80 glioblastomas were interrogated for associations with resistance to therapy. Patients were treated within clinical trials testing the addition of concomitant and adjuvant temozolomide to radiotherapy. Results: An expression signature dominated by HOX genes, which comprises Prominin-1 (CD133), emerged as a predictor for poor survival in patients treated with concomitant chemoradiotherapy (n = 42; hazard ratio = 2.69; 95% CI, 1.38 to 5.26; P = .004). This association could be validated in an independent data set. Provocatively, the HOX cluster was reminiscent of a "self-renewal" signature (P = .008; Gene Set Enrichment Analysis) recently characterized in a mouse leukemia model. The HOX signature and EGFR expression were independent prognostic factors in multivariate analysis, adjusted for the O-6-methylguanine-DNA methyltransferase (MGMT) methylation status, a known predictive factor for benefit from temozolomide, and age. Better outcome was associated with gene clusters characterizing features of tumor-host interaction including tumor vascularization and cell adhesion, and innate immune response. Conclusion: This study provides first clinical evidence for the implication of a "glioma stem cell" or "self-renewal" phenotype in treatment resistance of glioblastoma. Biologic mechanisms identified here to be relevant for resistance will guide future targeted therapies and respective marker development for individualized treatment and patient selection.

Keywords: Oncology

L1-CAM, a neuronal cell adhesion receptor, is also expressed in a variety of cancer cells. Recent studies identified L1-CAM as a target gene of β-catenin-T-cell factor (TCF) signaling expressed at the invasive front of human colon cancer tissue. We found that L1-CAM expression in colon cancer cells lacking L1-CAM confers metastatic capacity, and mice injected in their spleen with such cells form liver metastases. We identified ADAM10, a metalloproteinase that cleaves the L1-CAM extracellular domain, as a novel target gene of β-catenin-TCF signaling. ADAM10 overexpression in colon cancer cells displaying endogenous L1-CAM enhanced L1-CAM cleavage and induced liver metastasis, and ADAM10 also enhanced metastasis in colon cancer cells stably transfected with L1-CAM. DNA microarray analysis of genes induced by L1-CAM in colon cancer cells identified a cluster of genes also elevated in a large set of human colon carcinoma tissue samples. Expression of these genes in normal colon epithelium was low. These results indicate that there is a gene program induced by L1-CAM in colon cancer cells that is also present in colorectal cancer tissue and suggest that L1-CAM can serve as target for colon cancer therapy.

Background. Transcription factors (TF) regulate expression by binding to specific DNA sequences. A binding event is functional when it affects gene expression. Functionality of a binding site is reflected in conservation of the binding sequence during evolution and in over represented binding in gene groups with coherent biological functions. Functionality is governed by several parameters such as the TF-DNA binding strength, distance of the binding site from the transcription start site (TSS), DNA packing, and more. Understanding how these parameters control functionality of different TFs in different biological contexts is a must for identifying functional TF binding sites and for understanding regulation of transcription. Methodology/Principal Findings. We introduce a novel method to screen the promoters of a set of genes with shared biological function (obtained from the functional Gene Ontology (GO) classification) against a precompiled library of motifs, and find those motifs which are statistically over-represented in the gene set. More than 8000 human (and 23,000 mouse) genes, were assigned to one of 134 GO sets. Their promoters were searched (from 200 bp downstream to 1000 bp upstream the TSS) for 414 known DNA motifs. We optimized the sequence similarity score threshold, independently for every location window, taking into account nucleotide heterogeneity along the promoters of the target genes. The method, combined with binding sequence and location conservation between human and mouse, identifies with high probability functional binding sites for groups of functionally-related genes. We found many location-sensitive functional binding events,and showed that they clustered close to the TSS. Our method and findings were tested experimentally. Conclusions/Significance. We identified reliably functional TF binding sites. This is an essential step towards constructing regulatory networks. The promoter region proximal to the TSS is of central importance for regulation of transcription in human and mouse, just as it is in bacteria and yeast.

Primary immune response to pathogens involves the maturation of antigen-presenting dendritic cells ( DC). Bacterial lipopolysacharride (LPS) is a potent inducer of DC maturation, whereas the transforming growth factor beta (TGF beta) attenuates much of this process. Here, we analyzed the global gene expression pattern in LPS-treated bone marrow derived DC during inhibition of their maturation process by TGF beta. Exposure of DC to LPS induces a pronounced cell response, manifested in altered expression of a large number of genes. Interestingly, TGFb did not affect most of the LPS responding genes. Nevertheless, analysis identified a subset of genes that did respond to TGF beta, among them the two inflammatory cytokines interleukin (IL)-12 and IL-18. Expression of IL-12, the major proinflammatory cytokine secreted by mature DC, was downregulated by TGF beta, whereas the expression level of the proinflammatory cytokine IL-18, known to potentiate the IL-12 effect, was upregulated. Expression of the peroxisome proliferator-activated receptor gamma (PPAR gamma) increased in response to TGF beta, concomitantly with reduced expression of chemokine receptor 7 (CCR7). This finding supports the possibility that TGF beta-dependent inhibition of CCR7 expression in DC is mediated by PPAR gamma.

Dietary antioxidants and selenium compounds were shown to have a therapeutic effect against hepatocellular carcinoma in several mouse models. We tested the effects of tannic acid and selenomethionine on hepatocellular carcinoma development in Mdr2 knockout (Mdr2-KO) mice. Mdr2-KO and age-matched Mdr2 heterozygous control mice were fed with tannic acid or selenomethionine during the first 3 months of life. Then, several mice from each group were sacrificed, and liver tissue samples were removed for analysis. The remaining mice were fed a regular diet until the age of 16 months, at which time the number and size of liver tumors were determined. Liver tissue samples of 3-month-old mice were subjected to gene expression profiling analysis using cDNA macroarrays containing probes for 240 genes that regulate responses to oxidative stress and inflammation or lipid metabolism. Both tannic acid and selenomethionine had partial chemopreventive effect on development of hepatocellular carcinoma in Mdr2-KO mice: they reduced the incidence of large tumor nodules (diameter > 1 cm) at age 16 months. Both agents inhibited gene expression and reversed up-regulation of many genes that control inflammation or response to oxidative stress in Mdr2-KO livers at age 3 months. This inhibitory effect on gene expression correlated with the ability of agents to reduce incidence of large tumors: selenomethionine was more active than tannic acid in both aspects. Understanding the molecular mechanism of chemoprevention effect could improve our therapeutic modalities while using these agents.

A bioinformatics-based analysis of endochondral bone formation model detected several genes upregulated in this process. Among these genes the dickkopf homolog 3 (Dkk3) was upregulated and further studies showed that its expression affects in vitro and in vivo osteogenesis. This study indicates a possible role of Dkk3 in regulating bone formation. Introduction: Endochondral bone formation is a complex biological process involving numerous chondrogenic, osteogenic, and angiogenic proteins, only some of which have been well studied. Additional key genes may have important roles as well. We hypothesized that to identify key genes and signaling pathways crucial for bone formation, a comprehensive gene discovery strategy should be applied to an established in vivo model of osteogenesis. Materials and Methods: We used in vivo implanted C3H10T1/2 cells that had been genetically engineered to express human bone morphogenetic protein-2 (BMP2) in a tetracycline-regulated system that controls osteogenic differentiation. Oligonucleotide microarray data from the implants (n = 4 repeats) was analyzed using coupled two-way clustering (CTWC) and statistical methods. For studying the effects of dickkopf homolog 3 (Dkk3) in chondrogenesis and osteogenesis. C3H10T1/2 mesenchymal progenitors were used. Results: The CTWC revealed temporal expression of Dkk3 with other chondrogenesis-, osteogenesis-, and Wnt-related genes. Quantitative RT-PCR confirmed the expression of Dkk3 in the implants. C3H10T1/2 cells that expressed Dkk3 in the presence of BMP2 displayed lower levels of alkaline phosphatase and collagen 1 mRNA expression than control C3H10T1/2 cells that did not express Dkk3. Interestingly, the levels of collagen II mRNA expression, Alcian blue staining, and glucose aminoglycans (GAGs) production were not influenced by Dkk3 expression. In vivo μCT and bioluminescence imaging revealed that co-expression of Dkk3 and BMP2 by implanted C3H10T1/2 cells induced the formation of significantly lower quantities of bone than cells expressing only BMP2. Conclusions: A bioinformatics analysis enabled the identification of Dkk3 as a pivotal gene with a novel function in endochondral bone formation. Our results showed that Dkk3 might have inhibitory effects on osteogenesis, but no effect on chondrogenesis, indicating that Dkk3 plays a regulatory role in endochondral bone formation. Further mechanistic studies are required to reveal the mechanism of action of Dkk3 in endochondral bone formation.

Background: The human genome contains over one million Alu repeat elements whose distribution is not uniform. While metabolism-related genes were shown to be enriched with Alu, in structural genes Alu elements are under-represented. Such observations led researchers to suggest that Alu elements were involved in gene regulation and were selected to be present in some genes and absent from others. This hypothesis is gaining strength due to findings that indicate involvement of Alu elements in a variety of functions; for example, Alu sequences were found to contain several functional transcription factor (TF) binding sites (BSs). We performed a search for new putative BSs on Alu elements, using a database of Position Specific Score Matrices (PSSMs). We searched consensus Alu sequences as well as specific Alu elements that appear on the 5 Kbp regions upstream to the transcription start site (TSS) of about 14000 genes. Results: We found that the upstream regions of the TSS are enriched with Alu elements, and the Alu consensus sequences contain dozens of putative BSs for TFs. Hence several TFs have Alu-associated BSs upstream of the TSS of many genes. For several TFs most of the putative BSs reside on Alu; a few of these were previously found and their association with Alu was also reported. In four cases the fact that the identified BSs resided on Alu went unnoticed, and we report this association for the first time. We found dozens of new putative BSs. Interestingly, many of the corresponding TFs are associated with early markers of development, even though the upstream regions of development-related genes are Alu-poor, compared with translational and protein biosynthesis related genes, which are Alu-rich. Finally, we found a correlation between the mouse B1 and human Alu densities within the corresponding upstream regions of orthologous genes. Conclusion: We propose that evolution used transposable elements to insert TF binding motifs into promoter regions. We observed enrichment of biosynthesis genes with Alu-associated BSs of developmental TFs. Since development and cell proliferation (of which biosynthesis is an essential component) were proposed to be opposing processes, these TFs possibly play inhibitory roles, suppressing proliferation during differentiation.

The transcription factor Nanog is uniquely expressed in embryonic stem (ES) cells and in germ cell tumors and is important for self-renewal. To understand the relation between this and cell transformation, we expressed Nanog in NIH3T3 cells, and these cells showed an increased growth rate and a transformed phenotype as demonstrated by foci formation and colony growth in soft agar. This suggests that Nanog possesses an oncogenic potential that may be related to the role it plays in germ cell tumors and to its function in self renewal of ES cells. We studied the transcription targets of Nanog using microarrays to identify Nanog regulated genes. The list of genes regulated by Nanog was unique for each cell type and more than 10% of the Nanog regulated genes, including transcription factors, are primary Nanog targets since their promoters bind Nanog in ES cells. Some of these target genes can explain the transformation of NIH3T3.

Keywords: Oncology; Radiology, Nuclear Medicine & Medical Imaging

Traditionally, immunologic diagnosis has been based on an attempt to correlate each disease with a specific immune reactivity, such as an antibody or a T-cell response to a single antigen specific for the disease entity. The state of the body, however, appears to be encoded by the immune system in collectives of reactivities and not by single reactivities. Here we describe our use of microarray technology and informatics to develop an antigen chip capable of detecting global patterns of antibodies binding to hundreds of antigens simultaneously. The patterns fashion diagnostic signatures.

Specific HPV DNA sequences are associated with more than 90% of invasive carcinomas of the uterine cervix. Viral E6 and E7 oncogenes are key mediators in cell transformation by disrupting TP53 and RB pathways. To investigate molecular mechanisms involved in the progression of invasive cervical carcinoma, we performed a gene expression study on cases selected according to viral and clinical parameters. Using Coupled Two-Way Clustering and Sorting Points Into Neighbourhoods methods, we identified a 'cervical cancer proliferation cluster' composed of 163 highly correlated transcripts. Most of these transcripts corresponded to E2F pathway genes controlling cell division or proliferation, whereas none was known as TP53 primary target. The average expression level of the genes of this cluster was higher in tumours with an early relapse than in tumours with a favourable course (P=0.026). Moreover, we found that E6/E7 mRNA expression level was positively correlated with the expression level of the cluster genes and with viral DNA load. These findings suggest that HPV E6/E7 expression level plays a key role in the progression of invasive carcinoma of the uterine cervix via the deregulation of cellular genes controlling tumour cell proliferation. HPV expression level may thus provide a biological marker useful for prognosis assessment and specific therapy of the disease.

On the basis of epidemiological studies, infection was suggested to play a role in the etiology of human cancer. While for some cancers such a role was indeed demonstrated, there is no direct biological support for the role of viral pathogens in the pathogenesis of childhood leukemia. Using a novel bioinformatic tool that alternates between clustering and standard statistical methods of analysis, we performed a 'double-blind' search of published gene expression data of subjects with different childhood acute lymphoblastic leukemia (ALL) subtypes, looking for unanticipated partitions of patients, induced by unexpected groups of genes with correlated expression. We discovered a group of about 30 genes, related to the interferon response pathway, whose expression levels divide the ALL samples into two subgroups; high in 50, low in 285 patients. Leukemic subclasses prevalent in early childhood (the age most susceptible to infection) are over-represented in the high-expression subgroup. Similar partitions, induced by the same genes, were found also in breast and ovarian cancer but not in lung cancer, prostate cancer and lymphoma. About 40% of breast cancer samples expressed the 'interferon-related' signature. It is of interest that several studies demonstrated mouse mammary tumor virus-like sequences in about 40% of breast cancer samples. Our discovery of an unanticipated strong signature of an interferon-induced pathway provides molecular support for a role for either inflammation or viral infection in the pathogenesis of childhood leukemia as well as breast and ovarian cancer.

The difficulty to dissect a complex phenotype of established malignant cells to several critical transcriptional programs greatly impends our understanding of the malignant transformation. The genetic elements required to transform some primary human cells to a tumorigenic state were described in several recent studies. We took the advantage of the global genomic profiling approach and tried to go one step further in the dissection of the transformation network. We sought to identify the genetic signatures and key target genes, which underlie the genetic alterations in p53, Ras, INK4A locus, and telotnerase, introduced in a stepwise manner into primary human fibroblasts. Here, we show that these are the minimally required genetic alterations for sarcomagenesis in vivo. A genome-wide expression profiling identified distinct genetic signatures corresponding to the genetic alterations listed above. Most importantly, unique transformation hallmarks, such as differentiation block, aberrant mitotic progression, increased angiogenesis, and invasiveness, were identified and coupled with genetic signatures assigned for the genetic alterations in the p53, INK4A locus, and H-Ras, respectively. Furthermore, a transcriptional program that defines the cellular response to p53 inactivation was an excellent predictor of metastasis development and bad prognosis in breast cancer patients. Deciphering these transformation fingerprints, which are affected by the most common oncogenic mutations, provides considerable insight into regulatory circuits controlling malignant transformation and will hopefully open new avenues for rational therapeutic decisions.

Motivation: Genes are often characterized dichotomously as either housekeeping or single-tissue specific. We conjectured that crucial functional information resides in genes with midrange profiles of expression. Results: To obtain such novel information genome-wide, we have determined the mRNA expression levels for one of the largest hitherto analyzed set of 62839 probesets in 12 representative normal human tissues. Indeed, when using a newly defined graded tissue specificity index τ, valued between 0 for housekeeping genes and 1 for tissue-specific genes, genes with midrange profiles having 0.15 50% of all expression patterns. We developed a binary classification, indicating for every gene the l_B tissues in which it is overly expressed, and the 12 - l_B tissues in which it shows low expression. The 85 dominant midrange patterns with l_B = 2-11 were found to be bimodally distributed, and to contribute most significantly to the definition of tissue specification dendrograms. Our analyses provide a novel route to infer expression profiles for presumed ancestral nodes in the tissue dendrogram. Such definition has uncovered an unsuspected correlation, whereby de novo enhancement and diminution of gene expression go hand in hand. These findings highlight the importance of gene suppression events, with implications to the course of tissue specification in ontogeny and phylogeny.

Human embryonic stem cells (ESC) are undifferentiated and are endowed with the capacities of self-renewal and pluripotential differentiation. Adult stem cells renew their own tissue, but whether they can transdifferentiate to other tissues is still controversial. To understand the genetic program that underlies the pluripotency of stem cells, we compared the transcription profile of ESC with that of progenitor/stem cells of human hematopoietic and keratinocytic origins, along with their mature cells to be viewed as snapshots along tissue differentiation. ESC gene profiles show higher complexity with significantly more highly expressed genes than adult cells. We hypothesize that ESC use a strategy of expressing genes that represent various differentiation pathways and selection of only a few for continuous expression upon differentiation to a particular target. Such a strategy may be necessary for the pluripotency of ESC. The progenitors of either hematopoietic or keratinocytic cells also follow the same design principle. Using advanced clustering, we show that many of the ESC expressed genes are turned off in the progenitors/stem cells followed by a further down-regulation in adult tissues. Concomitantly, genes specific to the target tissue are up-regulated toward mature cells of skin or blood.

We calculate the Shannon entropy rate of a binary Hidden Markov Process (HMP), of given transition rate and noise ε (emission), as a series expansion in ε. The first two orders are calculated exactly. We then evaluate, for finite histories, simple upper-bounds of Cover and Thomas. Surprisingly, we find that for a fixed order k and history of n steps, the bounds become independent of n for large enough n. This observation is the basis of a conjecture, that the upper-bound obtained for n ≥ (k + 3)/2 gives the exact entropy rate for any desired order k of ε.

Motivation: Predicting the metastatic potential of primary malignant tissues has direct bearing on the choice of therapy. Several microarray studies yielded gene sets whose expression profiles successfully predicted survival. Nevertheless, the overlap between these gene sets is almost zero. Such small overlaps were observed also in other complex diseases, and the variables that could account for the differences had evoked a wide interest. One of the main open questions in this context is whether the disparity can be attributed only to trivial reasons such as different technologies, different patients and different types of analyses. Results: To answer this question, we concentrated on a single breast cancer dataset, and analyzed it by a single method, the one which was used by van't Veer et al. to produce a set of outcome-predictive genes. We showed that, in fact, the resulting set of genes is not unique; it is strongly influenced by the subset of patients used for gene selection. Many equally predictive lists could have been produced from the same analysis. Three main properties of the data explain this sensitivity: (1) many genes are correlated with survival; (2) the differences between these correlations are small; (3) the correlations fluctuate strongly when measured over different subsets of patients. A possible biological explanation for these properties is discussed.

One's present repertoire of antibodies encodes the history of one's past immunological experience. Can the present autoantibody repertoire be consulted to predict resistance or susceptibility to the future development of an autoimmune disease? Here, we developed an antigen microarray chip and used bioinformatic analysis to study a model of type 1 diabetes developing in nonobese diabetic male mice in which the disease was accelerated and synchronized by exposing the mice to cyclophosphamide at 4 weeks of age. We obtained sera from 19 individual mice, treated the mice to induce cyclophosphamide-accelerated diabetes (CAD), and found, as expected, that 9 mice became severely diabetic, whereas 10 mice permanently resisted diabetes. We again obtained serum from each mouse after CAD induction. We then analyzed, by using rank-order and superparamagnetic clustering, the patterns of antibodies in individual mice to 266 different antigens spotted on the chip. A selected panel of 27 different antigens (10% of the array) revealed a pattern of IgG antibody reactivity in the pre-CAD sera that discriminated between the mice resistant or susceptible to CAD with 100% sensitivity and 82% specificity (P = 0.017). Surprisingly, the set of IgG antibodies that was informative before CAD induction did not separate the resistant and susceptible groups after the onset of CAD; new antigens became critical for post-CAD repertoire discrimination. Thus, at least for a model disease, present antibody repertoires can predict future disease, predictive and diagnostic repertoires can differ, and decisive information about immune system behavior can be mined by bioinformatic technology. Repertoires matter.

Informatic methodologies are being applied successfully to analyze the complexity of the genome. But beyond the genome, the immune system reflects the state of the body in health and disease. Traditionally, immunologists have reduced the immune system, where possible, to one-to-one relationships between particular antigens and particular antibodies or T-cell clones. Autoimmune diseases, caused by an immune attack against a body component, are usually investigated by following the response to single self-antigens. In this study, we apply informatics to analyze patterns of autoantibodies rather than single species of autoantibodies. This study was designed not to replace traditional approaches to immune diagnosis, but to test whether meaningful patterns of autoantibodies might exist. Using an unbiased solid-phase ELISA antibody test, we detected serum IgG and IgM antibodies in the sera of 20 healthy persons and 20 persons with type 1 diabetes mellitus binding to an array of 87 different antigens, mostly self-antigens. The healthy subjects manifested autoantibodies to a variety of self-antigens, many known to be associated with autoimmune diseases. We investigated the patterns of these autoantibodies using a coupled two-way clustering algorithm developed for analyzing data from gene arrays. We now report that the reactivity patterns of autoantibodies to particular subsets of self-antigens exhibited non-trivial structure, which significantly discriminated between healthy persons and persons with type 1 diabetes. The results show that despite the wide prevalence of autoantibodies, the patterns of reactivity to defined subsets of self-antigens can provide information about the state of the body.

The ALL-1 gene is directly involved in 5-10% of acute lymphoblastic leukemias (ALLs) and acute myeloid leukemias (AMLs) by fusion to other genes or through internal rearrangements. DNA microarrays were used to determine expression profiles of ALLs and AMLs with ALL-1 rearrangements. These profiles distinguish those tumors from other ALLs and AMLs. The expression patterns of ALL-1-associated tumors, in particular ALLs, involve oncogenes, tumor suppressors, antiapoptotic genes, drug-resistance genes, etc., and correlate with the aggressive nature of the tumors. The genes whose expression differentiates between ALLs with and without ALL-1 rearrangement were further divided into several groups, enabling separation of ALL-1-associated ALLs into two subclasses. One of the groups included 43 genes that exhibited expression profiles closely linked to ALLs with ALL-1 rearrangements. Further, there were evident differences between the expression profiles of AMLs in which ALL-1 had undergone fusion to other genes and AMLs with partial duplication of ALL-1. The extensive analysis described here pinpointed genes that might have a direct role in pathogenesis.

To gain insight into the transformation of epidermal cells into squamous carcinoma cells (SCC), we compared the response to ultraviolet B radiation (UVB) of normal human epidermal keratinocytes (NHEK) versus their transformed counterpart, SCC, using biological and molecular profiling. DNA microarray analyses (Affymetrix®, ∼ 12000 genes) indicated that the major group of upregulated genes in keratinocytes fall into three categories: (i) antiapoptotic and cell survival factors, including chemokines of the CXC/CC subfamilies (e.g. IL-8, GRO-1, -2, -3, SCYA20), growth factors (e.g. HB-EGF, CTGF, INSL-4), and proinflammatory mediators (e.g. COX-2, S100A9), (ii) DNA repair-related genes (e.g. GADD45, ERCC, BTG-1, Histories), and (iii) ECM proteases (MMP-1, -10). The major downregulated genes are ΔNp63 and PUMILIO, two potential markers for the maintenance of keratinocyte stem cells. NHEK were found to be more resistant than SCC to UVB-induced apoptosis and this resistance was mainly because of the protection from cell death by secreted survival factors, since it can be transferred from NHEK to SCC cultures by the conditioned medium. Whereas the response of keratinocytes to UVB involved regulation of key checkpoint genes (p53, MDM2, p21^Cip1, ΔNp63), as well as antiapoptotic and DNA repair-related genes - no or little regulation of these genes was observed in SCC. The effect of UVB on NHEK and SCC resulted in upregulation of 251 and 127 genes, respectively, and downregulation of 322 genes in NHEK and 117 genes in SCC. To further analyse these changes, we used a novel unsupervised coupled two-way clustering method that allowed the identification of groups of genes that clearly partitioned keratinocytes from SCC, including a group of genes whose constitutive expression levels were similar before UVB. This allowed the identification of discriminating genes not otherwise revealed by simple static comparison in the absence of UVB irradiation. The implication of the changes in gene profile in keratinocytes for epithelial cancer is discussed.

Clustering gene expression data by exploiting phase transitions in granular ferromagnets requires transforming the data to a granular substrate. We present a method using the recently introduced homogeneity order parameter [Formula presented] [H. Agrawal, Phys. Rev. Lett. 89, 268702 (2002)] for optimizing the parameter controlling the \u201cgranularity\u201d and thus the stability of partitions. The model substrates obtained for gene expression data have a highly granular structure. We explore properties of phase transition in high [Formula presented] ferromagnetic Potts models on these substrates and show that the maximum of the width of superparamagnetic domain, corresponding to maximally stable partitions, coincides with the minimum of [Formula presented].

We present an automated procedure to assign CATH and SCOP classifications to proteins whose FSSP score is available. CATH classification is assigned down to the topology level, and SCOP classification is assigned to the fold level. Because the FSSP database is updated weekly, this method makes it possible to update also CATH and SCOP with the same frequency. Our predictions have a nearly perfect success rate when ambiguous cases are discarded. These ambiguous cases are intrinsic in any protein structure classification that relies on structural information alone. Hence, we introduce the "twilight zone for structure classification." We further suggest that to resolve these ambiguous cases, other criteria of classification, based also on information about sequence and function, must be used.

We introduce a method for validation of results obtained by clustering analysis of data. The method is based on resampling the available data. A figure of merit that measures the stability of clustering solutions against resampling is introduced. Clusters that are stable against resampling give rise to local maxima of this figure of merit. This is presented first for a one-dimensional data set, for which an analytic approximation for the figure of merit is derived and compared with numerical measurements. Next, the applicability of the method is demonstrated for higher-dimensional data, including gene microarray expression data.

The unbiased samples of ground states (GS) of the ±J short-range Ising spin glass possess a characteristics of the mean-field (MF) description, such as nontrivial and a hierarchical but nonultrametric, treelike structure. Clustering the GS showed that this structure was induced by correlated spin domains, which served as the cores of macroscopic zero energy excitations, taking the system from one state-cluster to another.

Two methods were proposed recently to derive energy parameters from known native protein conformations and corresponding sets of decoys. One is based on finding, by means of a perceptron learning scheme, energy parameters such that the native conformations have lower energies than the decoys. The second method maximizes the difference between the native energy and the average energy of the decoys, measured in terms of the width of the decoys' energy distribution (Z-score). Whereas the perceptron method is sensitive mainly to 'outlier' (i.e., extremal) decoys, the Z-score optimization is governed by the high density regions in decoy-space. We compare the two methods by deriving contact energies for two very different sets of decoys: the first obtained for model lattice proteins and the second by threading. We find that the potentials derived by the two methods are of similar quality and fairly closely related. This finding indicates that standard, naturally occurring sets of decoys are distributed in a way that yields robust energy parameters (that are quite insensitive to the particular method used to derive them). The main practical implication of this finding is that it is not necessary to fine-tune the potential search method to the particular set of decoys used. (C) 2000 Wiley-Liss, Inc.

We analyzed several energy functions for predicting the native state of proteins from an energy minimization procedure. We derived the parameters of a given energy function by imposing the basic requirement that the energy of the native conformation of a protein is lower than that of any conformation chosen from a set of decoys. Our work is motivated by a recent result which proved that the simple pairwise contact approximation of the energy is insufficient to satisfy simultaneously such a basic requirement for all the proteins in a database. Here, we investigate the reasons of such negative results and show how to improve the predictive power of methods based on energy minimization. We generated decoys by gapless threading, and we derive energy parameters by perceptron learning. We first considered hydrophobic contributions to the energy, defined in several ways, and showed that the additional hydrophobic terms enlarge slightly the number of proteins that can be stabilized together. Next, we performed various modifications of the pairwise energy term. We introduced (1) a distinction between inter-residue contacts on the surface and in the core of a protein and (2) a simple distance-dependent pairwise interaction in which a two-tier definition of contact replaces the original (single-tier) one. Our results suggest that a detailed treatment of the pairwise potential is likely to be more relevant than the consideration of other forces. (C) 2000 Wiley-Liss, Inc.

A simple model for flowing sand on an inclined plane is introduced. The model is related to recent experiments by Douady and Daerr and reproduces some of the experimentally observed features. Avalanches of intermediate size appear to be compact, placing the critical behavior of the model into the universality class of compact directed percolation. On very large scales, however, the avalanches break up into several branches, leading to a crossover from compact to ordinary directed percolation. Thus, systems of flowing granular matter on an inclined plane could serve as a first physical realization of directed percolation.

The aim of clustering is to partition data according to natural classes present in it. We proposed recently a method that makes no explicit assumption about the structure of the data and under very general and natural assumptions solves the clustering problem by evaluating thermal properties of a disordered (granular) magnet. The method was tested successfully on a variety of artificial and real-life problems; here we emphasize its application to analyze results obtained by a novel method of computer vision. The combination of these two techniques provides a powerful tool that succeeded to cluster properly 90 images of 6 objects on the basis of their pairwise dissimilarities. These dissimilarities, which constitute a highly non-metric set of pairwise distances between the images, form the input for clustering. A hierarchical organization of the images that agrees with human intuition, was obtained without assigning to the images coordinates in some abstract space.

Changing a few contacts in a contact map corresponds to a large scale move in confrontation space; hence, one gains a lot by using the contact map representation for protein folding. We developed an efficient search procedure in the space of physical contact maps, which could identify the native fold as of the lowest free energy, provided on had a free energy function whose ground state is the native map. We prove rigorously that the widely used pairwise contact approximation to the free energy cannot stabilize even a single proteins native map. Testing the native map against a set of decoys obtained by gapless threading, one may be misled to the opposite conclusion.

Damage spreading for Ising cluster dynamics is investigated numerically by using random numbers in a way that conforms with the notion of submitting the two evolving replicas to the same thermal noise. Two damage spreading transitions are found; damage does not spread either at low or high temperatures. We determine some critical exponents at the high-temperature transition point, which seem consistent with directed percolation.

We demonstrate that pairwise contact potentials alone cannot be used to predict the native fold of a protein. Ideally, one would hope that a universal energy function exists, for which the native folds of all proteins are the respective ground states. Here we pose a much more restricted question: Is it possible to find a set of contact parameters for which the energy of the native contact map of a single protein (crambin) is lower than that of all possible physically realizable decoy maps? The set of maps we used was derived by energy minimization (not by threading). We seek such a set of parameters by perceptron learning, a procedure which is guaranteed to find such a set if it exists. We found that it is impossible to fine-tune contact parameters that will assign all alternative conformations higher energy than that of the native map. This finding proves that there is no pairwise contact potential that can be used to fold any given protein. Inclusion of additional energy terms, such as hydrophobic (solvation), hydrogen bond, or multibody interactions may help to attain foldability within specific structural families.

We evaluate by Monte Carlo simulations various singular thermodynamic quantities [Formula Presented] for ensembles of quenched random Ising and Ashkin-Teller models. The measurements are taken at [Formula Presented] and we study how the distributions [Formula Presented] (and, in particular, their relative squared width, [Formula Presented] over the ensemble depend on the system size [Formula Presented] The Ashkin-Teller model was studied in the regime where bond randomness is irrelevant and we found weak self-averaging; [Formula Presented] where [Formula Presented] and [Formula Presented] are the exponents (of the pure model fixed point) governing the transition. For the site-dilute Ising model on a cubic lattice, known to be governed by a random fixed point, we find that [Formula Presented] tends to a constant, as predicted by Aharony and Harris. We tested whether this constant is universal. However, this constant is different for canonical and grand canonical disorder. We identify the pseudocritical temperature of each sample [Formula Presented] [Formula Presented] as the temperature at which the susceptibility reaches its maximal value [Formula Presented] The distribution of these sample dependent [Formula Presented] was investigated; we found that its variance scales as [Formula Presented] Our previously proposed finite size scaling ansatz for disordered systems was tested and found to hold. We did observe deviations from a single function, which imply that sample dependent scaling functions are needed. These deviations are, however, relatively small and hence to obtain a fixed statistical error it may be more computationally efficient to measure [Formula Presented] than the commonly used [Formula Presented]

We discuss the results of our attempt to predict the folded state of proteins by using the contact map representation. In order to use contact maps, one has to solve a few methodological problems. First, we propose an efficient way to explore the space of contact maps to generate creditable candidates for the native state. Second, we introduce a procedure to ensure that the generated maps are physical. Third, and most important, we present a method to derive contact energy parameters based on perceptron learning. The energy function must be able to discriminate between the native state and a set of decoys, simultaneously for all the proteins in a given database. We show that such an energy function exists if the candidates are produced by gapless threading, for a database of 153 proteins. If, however, we use as decoys maps of low energy that are generated by our dynamical scheme, not even one single protein can be stabilized. We conclude by discussing the perspectives of our approach.

We present a general definition of damage spreading in a pair of models. Using this general framework, one can define damage spreading in an objective manner that does not depend on the particular dynamic procedure that is being used. The formalism can be used for any spin-model or cellular automaton, with sequential or parallel update rules. At this point we present its application to the Domany-Kinzel cellular automaton in one dimension, this being the simplest model in which damage spreading has been found and studied extensively. We show that the active phase of this model consists of three subphases characterized by different damage-spreading properties.

We present two interesting results regarding damage spreading in ferromagnetic Ising models. First, we show that a damage spreading transition can occur in an Ising chain that evolves in contact with a thermal reservoir. Damage heals at low temperature and spreads at high T. The dynamic rules for the system's evolution for which such a transition is observed are as legitimate as the conventional rules (Glauber, Metropolis, heat bath). Our second result is that such transitions are not always in the directed percolation universality class.

We present a new approach for clustering, based on the physical properties of an inhomogeneous ferromagnetic model. We do not assume any structure of the underlying distribution of the data. A Potts spin is assigned to each data point and short range interactions between neighboring points are introduced. Spin-spin correlations, measured (by Monte Carlo procedure) in a superparamagnetic regime in which aligned domains appear, serve to partition the data points into clusters. Our method outperforms other algorithms for toy problems as well as for real data.

The dynamic behavior of cluster algorithms is analyzed in the classical mean-field limit. Rigorous analytical results below [Formula Presented] establish that the dynamic exponent has the value [Formula Presented]=1 for the Swendsen-Wang algorithm and [Formula Presented]=0 for the Wolff algorithm. An efficient Monte Carlo implementation is introduced, adapted for using these algorithms for fully connected graphs. Extensive simulations both above and below [Formula Presented] demonstrate scaling and evaluate the finite-size scaling function by means of a rather impressive collapse of the data.

A new approach for clustering is proposed. This method is based on an analogy to a physical model; the ferromagnetic Potts model at thermal equilibrium is used as an analog computer for this hard optimization problem. We do not assume any structure of the underlying distribution of the data. Phase space of the Potts model is divided into three regions; ferromagnetic, super-paramagnetic and paramagnetic phases. The region of interest is that corresponding to the super-paramagnetic one, where domains of aligned spins appear. The range of temperatures where these structures are stable is indicated by a non-vanishing magnetic susceptibility. We use a very efficient Monte Carlo algorithm to measure the susceptibility and the spin spin correlation function. The values of the spin spin correlation function, at the super-paramagnetic phase, serve to identify the partition of the data points into clusters.

The critical behavior of a bond-disordered Ashkin-Teller model on a square lattice is investigated by intensive Monte Carlo simulations. A duality transformation is used to locate a critical plane of the disordered model. This critical plane corresponds to the line of critical points of the pure model, along which critical exponents vary continuously. Along this line the scaling exponent corresponding to randomness φ=(α/ν) varies continuously and is positive so that the randomness is relevant, and different critical behavior is expected for the disordered model. We use a cluster algorithm for the Monte Carlo simulations based on the Wolff embedding idea, and perform a finite size scaling study of several critical models, extrapolating between the critical bond-disordered Ising and bond-disordered four-state Potts models. The critical behavior of the disordered model is compared with the critical behavior of an anisotropic Ashkin-Teller model, which is used as a reference pure model. We find no essential change in the order parameters' critical exponents with respect to those of the pure model. The divergence of the specific heat C is changed dramatically. Our results favor a logarithmic type divergence at Tc, C∼lnL for the random-bond Ashkin-Teller and four-state Potts models and C∼ln lnL for the random-bond Ising model.

A classification problem in high energy physics has been solved on simulated data using a simple multilayer perceptron comprising binary units which was trained with the CHIR algorithm. The unstable training of such a network on a nonseparable set has been overcome by selecting those weight vectors with good performance while providing a flexible choice of the two types of classification errors. Specific features of the problem have been exploited in order to simplify and optimize the solution which has been compared to the popular backpropagation algorithm and found to perform on a similar level. Additional aspects of this work are the use of the CHIR algorithm on continuous input and incorporating the classic idea of a phi-machine in a multilayer perceptron.

A simple model describing nucleation and growth following quench into a liquid-solid coexistence region is presented. The model is compared with small-angle X-ray scattering results on di-block copolymers.

The authors report calculations of the critical capacity of perceptrons that are subject to pattern-dependent noise. They also calculate the capacity of a perceptron whose weights take values on a shifted sphere, and show how this system resembles 'noisy' behaviour in some limits.

We present a novel cluster algorithm for Monte Carlo simulations of the fully frustrated Ising model on the square lattice. The new method does not suffer from problems of metastability, and is extremely efficient even at T=0. Our algorithm is a special case of a more general Monte Carlo simulation scheme. The general scheme unifies many cluster algorithms that were developed recently in order to accelerate Monte Carlo simulations.

A model layered feed-forward neural network is studied and solved exactly in the thermodynamic limit. Layer-to-layer recursion relations are found and analyzed as a function of the relevant external parameters. Stochasticity is introduced by a temperature variable. A region of good recall is found, separated from a region of no recall by a first-order line terminating at a critical point. The exact time evolution of mixtures of patterns is given as well.

We have developed a novel simulation method that combines a multigrid technique with a stochastic blocking procedure. Our algorithm eliminates critical slowing down completely, as demonstrated by simulations of the two-dimensional Ising model at criticality.

The problem of recognizing noisy patterns is trivially solved by identifying the key pattern which has the largest overlap with the input. We show that a procedure that does this can be implemented in a simple layered neural network. Recognition is free of spurious states, the network never reaches a state of total confusion. Processing time scales as log M, where M is the number of stored patterns.

A renormalization-group analysis of critical behavior near a Lifshitz tricritical point (LTP) is presented, with emphasis on the role played by new, momentum-dependent quartic terms. These result in new stable fixed points which determine the critical behavior. For some values of n (the number of order-parameter components) and m (the dimensionality of the softsubspace, characterized by quartic momentum-dependent inverse correlation functions), the renormalization-group recursion relations have two stable and accessible fixed points. However, one of these can never be reached in practice, due to a thermodynamic instability which results in a first-order phase transition. For m=d-1, one of the fixed points describes the critical dynamics of the usual n-vector spin model in (d-1) dimensions. This dynamic fixed point also characterizes LTP behavior for large n and n=1. In all other cases, the LTP has new exponents, which are not related to the dynamic model. Our results may be relevant to Lifshitz tricritical behavior in RbCaF3 and in some liquid-crystal systems.

We have recently found a phase transition in the dynamics of a particle that diffuses in the presence of a hierarchical array of barriers. The transition occurs as R, a parameter that controls the relative size of the barriers, is varied. For R>Rc, the system is diffusive; as R Rc+ the diffusion constant vanishes. For R

We study diffusion in the presence of a hierarchical set of barriers. We find a phase transition in the dynamics from ordinary to anomalous diffusion as a parameter controlling the relative barrier heights is varied. Similar behavior is found in suitably defined random systems. Possible connections to glassy dynamics are discussed.

Time development of cellular automata in d dimensions is mapped onto equilibrium statistical mechanics of Ising models in d+1 dimensions. Directed percolation is equivalent to a cellular automaton, and thus to an Ising model. For a particular case of directed percolation we find 1/2=2, 1/2=1, =0.

The appropriate equations for the use of elastic waves to probe an inhomogeneous state of residual stress caused by several types of misfitted inclusions are summarized. An elastic material is considered that initially is in an undeformed state and subsequently placed in a state of residual stress. Since after the deformation the material still behaves elastically, and since the residual stress state is one of mechanical equilibrium, the probing linear elastic wave can couple (i. e. , interact anharmonically) with the residual stress state only through nonlinear (acoustoelastic) effects. These interaction terms are viewed as local variations in the elastic stiffness tensor (magnitude and symmetry changes), produced by residual deformation gradients, that cause scattering of the probing elastic wave propagating through the residual stress field. By examining this scattering, the residual stress state is characterized. The results of applying the Born approximation are applied to two additional misfit problems: a 'screened' spherical misfit and a cylindrical misfit. In the first case results suggestive of broadened time-domain signals are obtained; in the second case, they are not.

An x-y model with nearest-neighbor interaction V()=2J{1-[cos2(2)]p2} is studied by Monte Carlo simulation. The model has a transition from a paramagnetic to an algebraic (massless) phase. An increase of the parameter p, which changes the shape of the potential, brings about a change in the nature of the transition from continuous to first order.

We present a multiple scattering formalism that was developed in order to calculate elastic wave scattering by complex defects. The formalism is based on an expansion of the T matrix in terms of single scatterer T matrices; we include all single and double scattering processes. The method has been numerically implemented for the case of two spherical cavities; the interaction terms were found to be small, which indicates that other problems of interest, such as a subsurface spherical cavity or crack can also be treated.

Results are presented for elastic wave scattering from irregular voids embedded in Ti alloy by the diffusion bonding process. The defects examined are: two overlapping spherical voids of unequal radii, two overlapping voids consisting of a sphere and a prolate spheroid, and a spherical void with an encircling crack. Representative plots are given for the raw waveforms, magnitude of the deconvolved Fourier transform, and in some cases the time impulse response function. The data are compared to and analyzed in terms of two current theoretical approaches. While good quantitative agreement was observed over certain ranges, the comparisons point to definite (in some cases not unexpected) limitations in either the pertaining theory or experiment or both. The results are discussed with an eye toward applications to nondestructive evaluation.

It is suggested that phase transitions induced by a magnetic field H//(111) in certain type I FCC antiferromagnets provide a physical realisation of the q=3- and q=4-state Potts models. The (H, T) phase diagram associated with these antiferromagnets for a field lying in the (110) plane is studied using mean field approximation and renormalisation group calculations in d=- epsilon dimensions. The phase diagram exhibits a critical and two first-order surfaces connected by lines to tricritical and polycritical points.

The q=3-state Potts model undergoes a first-order transition in three dimensions, and a continuous one in two dimensions. For a system of finite thickness L, the transition will be either first order (for L>L^x) or continuous (for L

We study directed percolation in two dimensions. Critical properties, such as the threshold pc and the correlation length exponent , depend on the direction · at which percolation is observed. We observe crossover from (·=0) to (·0); similar crossover is found when the probabilities assigned to vertical and horizontal bonds, pV and pH, become unequal. For the case pH=1 we solved the problem exactly, obtaining pc (·)=[1+cot(14- ·)]-1,=2.

The critical properties of Potts models with random bonds are considered in two dimensions. A position-space renormalization-group procedure, based on the Migdal-Kadanoff method, is developed. While all previous position-space calculations satisfied the Harris criterion and the resulting scaling relation only approximately, we found conditions under which these relations are exactly satisfied, and constructed our renormalization-group procedure accordingly. Numerical results for phase diagrams and thermodynamic functions for various random-bond Potts models are presented. In addition, some exact results obtained using a duality transformation, as well as an heuristic derivation of scaling properties that correspond to the percolation problem are given.

A position-space renormalization-group method, suitable for studying the localization properties of electrons in a disordered system, was developed. Two different approximations to a well-defined exact procedure were used. The first method is a perturbative treatment to lowest order in the intercell couplings. This yields a localization edge in three dimensions, with a fixed point at the band center (E=0) at a critical disorder σc7.0. In the neighborhood of the fixed point the localization length L is predicted to diverge as L∼(σ-σc+βE2)-ν. In two dimensions no fixed point is found, indicating localization even for small randomness, in agreement with Abrahams, Anderson, Licciardello, and Ramakrishnan. The second method is an application of the finite-lattice approximation, in which the intercell hopping between two (or more) cells is treated to infinite order in perturbation theory. To our knowledge, this method has not been previously used for quantum systems. Calculations based on this approximation were carried out in two dimensions only, yielding results that are in agreement with those of the lowest-order approximation.