Shin Ishii

On-line EM Algorithm for the Normalized Gaussian Network

A normalized gaussian network (NGnet) (Moody & Darken, 1989) is a network of local linear regression units. The model softly partitions the input space by normalized gaussian functions, and each local unit linearly approximates the output within the partition. In this article, we propose a new on-line EM algorithm for the NGnet, which is derived from the batch EM algorithm (Xu, Jordan, & Hinton 1995), by introducing a discount factor. We show that the on-line EM algorithm is equivalent to the batch EM algorithm if a specific scheduling of the discount factor is employed. In addition, we show that the on-line EM algorithm can be considered as a stochastic approximation method to find the maximum likelihood estimator. A new regularization method is proposed in order to deal with a singular input distribution. In order to manage dynamic environments, where the input-output distribution of data changes over time, unit manipulation mechanisms such as unit production, unit deletion, and unit division are also introduced based on probabilistic interpretation. Experimental results show that our approach is suitable for function approximation problems in dynamic environments. We also apply our on-line EM algorithm to robot dynamics problems and compare our algorithm with the mixtures-of-experts family.

Resolution of Uncertainty in Prefrontal Cortex

Making optimal decisions in the face of uncertain or incomplete information arises as a common problem in everyday behavior, but the neural processes underlying this ability remain poorly understood. A typical case is navigation, in which a subject has to search for a known goal from an unknown location. Navigating under uncertain conditions requires making decisions on the basis of the current belief about location and updating that belief based on incoming information. Here, we use functional magnetic resonance imaging during a maze navigation task to study neural activity relating to the resolution of uncertainty as subjects make sequential decisions to reach a goal. We show that distinct regions of prefrontal cortex are engaged in specific computational functions that are well described by a Bayesian model of decision making. This permits efficient goal-oriented navigation and provides new insights into decision making by humans.

A critical time window for dopamine actions on the structural plasticity of dendritic spines

Animal behavior follows rewards Animal behavior is learned and reinforced by rewards. On a molecular level, the reward comes in the form of the neurotransmitter, dopamine, which modulates synapses. The exact timing and mechanism of this process remain unknown. Using optical stimulation, Yagishita et al. found that dopaminergic modulation involved dendritic spine enlargement only during an extremely narrow time window. Known as reinforcement plasticity, this cellular basis for learning could provide insight into psychiatric disorders involving dopaminergic regulation, such as depression, drug addiction, and schizophrenia. Science, this issue p. 1616 Dopamine promotes spine structural plasticity during a narrow time window in mouse neuron distal dendrites. Animal behaviors are reinforced by subsequent rewards following within a narrow time window. Such reward signals are primarily coded by dopamine, which modulates the synaptic connections of medium spiny neurons in the striatum. The mechanisms of the narrow timing detection, however, remain unknown. Here, we optically stimulated dopaminergic and glutamatergic inputs separately and found that dopamine promoted spine enlargement only during a narrow time window (0.3 to 2 seconds) after the glutamatergic inputs. The temporal contingency was detected by rapid regulation of adenosine 3′,5′-cyclic monophosphate in thin distal dendrites, in which protein-kinase A was activated only within the time window because of a high phosphodiesterase activity. Thus, we describe a molecular basis of reinforcement plasticity at the level of single dendritic spines.

A network of chaotic elements for information processing

Abstract A globally coupled map (GCM) model is a network of chaotic elements that are globally coupled with each other. In this paper, first, a modified GCM model called the “globally coupled map using the symmetric map (S-GCM)” is proposed. The S-GCM is designed for information-processing applications. The S-GCM has attractors called “cluster frozen attractors”, each of which is taken to represent information. This paper also describes the following characteristics of the S-GCM which are important to information-processing applications: (a) the S-GCM falls into one of the cluster frozen attractors over a wide range of parameters. This means that the information representation is stable over parameters; (b) represented information can be preserved or broken by controlling parameters; (c) the cluster partitioning is restricted, i.e. the representation of information has a limitation. Finally, our techniques for applying the S-GCM to information processing are shown, considering these characteristics. Two associative memory systems are proposed and their performance is compared with that of the Hopfield network.

Novel risk stratification of patients with neuroblastoma by genomic signature, which is independent of molecular signature

Human neuroblastoma remains enigmatic because it often shows spontaneous regression and aggressive growth. The prognosis of advanced stage of sporadic neuroblastomas is still poor. Here, we investigated whether genomic and molecular signatures could categorize new therapeutic risk groups in primary neuroblastomas. We conducted microarray-based comparative genomic hybridization (array-CGH) with a DNA chip carrying 2464 BAC clones to examine genomic aberrations of 236 neuroblastomas and used in-house cDNA microarrays for gene-expression profiling. Array-CGH demonstrated three major genomic groups of chromosomal aberrations: silent (GGS), partial gains and/or losses (GGP) and whole gains and/or losses (GGW), which well corresponded with the patterns of chromosome 17 abnormalities. They were further classified into subgroups with different outcomes. In 112 sporadic neuroblastomas, MYCN amplification was frequent in GGS (22%) and GGP (53%) and caused serious outcomes in patients. Sporadic tumors with a single copy of MYCN showed the 5-year cumulative survival rates of 89% in GGS, 53% in GGP and 85% in GGW. Molecular signatures also segregated patients into the favorable and unfavorable prognosis groups (P=0.001). Both univariate and multivariate analyses revealed that genomic and molecular signatures were mutually independent, powerful prognostic indicators. Thus, combined genomic and molecular signatures may categorize novel risk groups and confer new clues for allowing tailored or even individualized medicine to patients with neuroblastoma.

Identification of expressed genes linked to malignancy of human colorectal carcinoma by parametric clustering of quantitative expression data

BackgroundIndividual human carcinomas have distinct biological and clinical properties: gene-expression profiling is expected to unveil the underlying molecular features. Particular interest has been focused on potential diagnostic and therapeutic applications. Solid tumors, such as colorectal carcinoma, present additional obstacles for experimental and data analysis.ResultsWe analyzed the expression levels of 1,536 genes in 100 colorectal cancer and 11 normal tissues using adaptor-tagged competitive PCR, a high-throughput reverse transcription-PCR technique. A parametric clustering method using the Gaussian mixture model and the Bayes inference revealed three groups of expressed genes. Two contained large numbers of genes. One of these groups correlated well with both the differences between tumor and normal tissues and the presence or absence of distant metastasis, whereas the other correlated only with the tumor/normal difference. The third group comprised a small number of genes. Approximately half showed an identical expression pattern, and cancer tissues were classified into two groups by their expression levels. The high-expression group had strong correlation with distant metastasis, and a poorer survival rate than the low-expression group, indicating possible clinical applications of these genes. In addition to c-yes, a homolog of a viral oncogene, prognostic indicators included genes specific to glial cells, which gives a new link between malignancy and ectopic gene expression.ConclusionsThe malignancy of human colorectal carcinoma is correlated with a unique expression pattern of a specific group of genes, allowing the classification of tumor tissues into two clinically distinct groups.

A diffusion-based neurite length-sensing mechanism involved in neuronal symmetry breaking

Although there has been significant progress in understanding the molecular signals that change cell morphology, mechanisms that cells use to monitor their size and length to regulate their morphology remain elusive. Previous studies suggest that polarizing cultured hippocampal neurons can sense neurite length, identify the longest neurite, and induce its subsequent outgrowth for axonogenesis. We observed that shootin1, a key regulator of axon outgrowth and neuronal polarization, accumulates in neurite tips in a neurite length‐dependent manner; here, the property of cell length is translated into shootin1 signals. Quantitative live cell imaging combined with modeling analyses revealed that intraneuritic anterograde transport and retrograde diffusion of shootin1 account for its neurite length‐dependent accumulation. Our quantitative model further explains that the length‐dependent shootin1 accumulation, together with shootin1‐dependent neurite outgrowth, constitutes a positive feedback loop that amplifies stochastic fluctuations of shootin1 signals, thereby generating an asymmetric signal for axon specification and neuronal symmetry breaking.

Molecular Prediction of Response to 5-Fluorouracil and Interferon-α Combination Chemotherapy in Advanced Hepatocellular Carcinoma

Purpose: The prognosis of hepatocellular carcinoma (HCC) is very poor, particularly in patients with tumors that have invaded the major branches of the portal vein. Combination chemotherapy with intra-arterial 5-fluorouracil and subcutaneous interferon-α has shown promising results for such advanced HCC, but it is important to develop the ability to accurately predict chemotherapeutic responses. Experimental Design: We analyzed the expression of 3,080 genes using a polymerase chain reaction-based array in 20 HCC patients who were treated with combination chemotherapy after reduction surgery. After unsupervised analyses, a supervised classification method for predicting chemotherapeutic responses was constructed. To minimize the number of predictive genes, we used a random permutation test to select only significant (P < 0.01) genes. A leave-one-out cross-validation confirmed the gene selection. We also prepared an additional 11 cases for validation of predictive performance. Results: Hierarchical clustering analysis and principal component analysis with all 3,080 genes revealed distinct gene expression patterns in responders (those with complete response or partial response) and nonresponders (those with stable disease or progressive disease) to the combination chemotherapy. Using a weighted-voting classification method with either all genes or only significant genes as assessed by permutation testing, the objective responses to treatment were correctly predicted in 17 of 20 cases (accuracy, 85%; positive predictive value, 100%; negative predictive value, 80%). Moreover, patients in the validation dataset could be classified into two distinct prognostic groups using 63 predictive genes. Conclusions: Molecular analysis of 63 genes can predict the response of patients with advanced HCC and major portal vein tumor thrombi to combination chemotherapy with 5-fluorouracil and interferon-α.

Spiking network simulation code for petascale computers.

Brain-scale networks exhibit a breathtaking heterogeneity in the dynamical properties and parameters of their constituents. At cellular resolution, the entities of theory are neurons and synapses and over the past decade researchers have learned to manage the heterogeneity of neurons and synapses with efficient data structures. Already early parallel simulation codes stored synapses in a distributed fashion such that a synapse solely consumes memory on the compute node harboring the target neuron. As petaflop computers with some 100,000 nodes become increasingly available for neuroscience, new challenges arise for neuronal network simulation software: Each neuron contacts on the order of 10,000 other neurons and thus has targets only on a fraction of all compute nodes; furthermore, for any given source neuron, at most a single synapse is typically created on any compute node. From the viewpoint of an individual compute node, the heterogeneity in the synaptic target lists thus collapses along two dimensions: the dimension of the types of synapses and the dimension of the number of synapses of a given type. Here we present a data structure taking advantage of this double collapse using metaprogramming techniques. After introducing the relevant scaling scenario for brain-scale simulations, we quantitatively discuss the performance on two supercomputers. We show that the novel architecture scales to the largest petascale supercomputers available today.

Semaphorin 3A induces CaV2.3 channel-dependent conversion of axons to dendrites.

Polarized neurites (axons and dendrites) form the functional circuitry of the nervous system. Secreted guidance cues often control the polarity of neuron migration and neurite outgrowth by regulating ion channels. Here, we show that secreted semaphorin 3A (Sema3A) induces the neurite identity of Xenopus spinal commissural interneurons (xSCINs) by activating CaV2.3 channels (CaV2.3). Sema3A treatment converted the identity of axons of cultured xSCINs to that of dendrites by recruiting functional CaV2.3. Inhibition of Sema3A signalling prevented both the expression of CaV2.3 and acquisition of the dendrite identity, and inhibition of CaV2.3 function resulted in multiple axon-like neurites of xSCINs in the spinal cord. Furthermore, Sema3A-triggered cGMP production and PKG activity induced, respectively, the expression of functional CaV2.3 and the dendrite identity. These results reveal a mechanism by which a guidance cue controls the identity of neurites during nervous system development.