Self-organized neural maps of human protein sequences.

We have recently described a method based on artificial neural networks to cluster protein sequences into families. The network was trained with Kohonen''s unsupervised learning algorithm using, as inputs, the matrix patterns derived from the dipeptide composition of the proteins. We present here a large-scale application of that method to classify the 1,758 human protein sequences stored in the SwissProt database (release 19.0), whose lengths are greater than 50 amino acids. In the final 2-dimensional topologically ordered map of 15 x 15 neurons, proteins belonging to known families were associated with the same neuron or with neighboring ones. Also, as an attempt to reduce the time-consuming learning procedure, we compared 2 learning protocols: one of 500 epochs (100 SUN CPU-hours [CPU-h]), and another one of 30 epochs (6.7 CPU-h). A further reduction of learning-computing time, by a factor of about 3.3, with similar protein clustering results, was achieved using a matrix of 11 x 11 components to represent the sequences. Although network training is time consuming, the classification of a new protein in the final ordered map is very fast (14.6 CPU-seconds). We also show a comparison between the artificial neural network approach and conventional methods of biosequence analysis.     

Nerve-independent formation of a topologically complex postsynaptic apparatus

As the mammalian neuromuscular junction matures, its acetylcholine receptor (AChR)-rich postsynaptic apparatus is transformed from an oval plaque into a pretzel-shaped array of branches that precisely mirrors the branching pattern of the motor nerve terminal. Although the nerve has been believed to direct postsynaptic maturation, we report here that myotubes cultured aneurally on matrix-coated substrates form elaborately branched AChR-rich domains remarkably similar to those seen in vivo. These domains share several characteristics with the mature postsynaptic apparatus, including colocalization of multiple postsynaptic markers, clustering of subjacent myonuclei, and dependence on the muscle-specific kinase and rapsyn for their formation. Time-lapse imaging showed that branched structures arise from plaques by formation and fusion of AChR-poor perforations through a series of steps mirroring that seen in vivo. Multiple fluorophore imaging showed that growth occurs by circumferential, asymmetric addition of AChRs. Analysis in vivo revealed similar patterns of AChR addition during normal development. These results reveal the sequence of steps by which a topologically complex domain forms on a cell and suggest an unexpected nerve-independent role for the postsynaptic cell in generating this topological complexity.     

Kohonen's feature maps for fly ash categorization

Fly ash is a common admixture used in concrete and may constitute up to 50% by weight of the total binder material. Incorporation of fly ash in Portland-cement concrete is highly desirable due to technological, economic, and environmental benefits. This article demonstrates the use of artificial intelligence neural networks for the classification of fly ashes in to different groups. Kohonen's Self Organizing Feature Maps is used for the purpose. As chemical composition of fly ash is crucial in the performance of concrete, eight chemical attributes of fly ashes have been considered. The application of simple Kohonen's one-dimensional feature maps permitted to differentiate three main groups of fly ashes. Three one-dimensional feature maps of topology 8-16, 8-24 and 8-32 were explored. The overall classification result of 8-16 topology was found to be significant and encouraging. The data pertaining to 80 fly ash samples were collected from standard published works. The categorization was found to be excellent and compares well with Canadian Standard Association's [CSA A 3000] classification scheme.     

Merging regions across feature maps in texture segmentation

This paper describes segmentation phenomena of superimposed textures and the Linking phenomena. These phenomena provide us with information about the mechanism of a late stage of segmenting textures. The late stage takes place after the segmentation process forms regions in feature maps such that parameter values in one region are substantially different from those in the neighboring regions. At the stage, the segmentation process merges the regions across the feature maps to determine output regions by integrating information about how the regions occupy two-dimensional space. The segmentation process gets such information both from local areas and from areas far away from the local areas where it determines the output regions and boundaries.     

Pattern formation and cortical maps

The response selectivities of neurons in adult primary sensory cortices depend on intricate patterns of synaptic connections; these selectivities are arranged over cortex in equally rich fashion. Characterising these patterns, and particularly the activity-dependence (and independence) of their developmental trajectories, has been a major task for experimental and theoretical neuroscience. Here, we describe and analyse a paradigmatic algorithm for activity-dependent development of the refinement and generation of neuronal selectivities, and relate it to some of the wealth of suggestions in the literature.     

Modeling epigenome folding: formation and dynamics of topologically associated chromatin domains

Genomes of eukaryotes are partitioned into domains of functionally distinct chromatin states. These domains are stably inherited across many cell generations and can be remodeled in response to developmental and external cues, hence contributing to the robustness and plasticity of expression patterns and cell phenotypes. Remarkably, recent studies indicate that these 1D epigenomic domains tend to fold into 3D topologically associated domains forming specialized nuclear chromatin compartments. However, the general mechanisms behind such compartmentalization including the contribution of epigenetic regulation remain unclear. Here, we address the question of the coupling between chromatin folding and epigenome. Using polymer physics, we analyze the properties of a block copolymer model that accounts for local epigenomic information. Considering copolymers build from the epigenomic landscape of Drosophila, we observe a very good agreement with the folding patterns observed in chromosome conformation capture experiments. Moreover, this model provides a physical basis for the existence of multistability in epigenome folding at sub-chromosomal scale. We show how experiments are fully consistent with multistable conformations where topologically associated domains of the same epigenomic state interact dynamically with each other. Our approach provides a general framework to improve our understanding of chromatin folding during cell cycle and differentiation and its relation to epigenetics.     

Self-organized lane formation and optimized traffic flow in army ants

We show how the movement rules of individual ants on trails can lead to a collective choice of direction and the formation of distinct traffic lanes that minimize congestion. We develop and evaluate the results of a new model with a quantitative study of the behaviour of the army ant Eciton burchelli. Colonies of this species have up to 200 000 foragers and transport more than 3000 prey items per hour over raiding columns that exceed 100 m. It is an ideal species in which to test the predictions of our model because it forms pheromone trails that are densely populated with very swift ants. The model explores the influences of turning rates and local perception on traffic flow. The behaviour of real army ants is such that they occupy the specific region of parameter space in which lanes form and traffic flow is maximized.     

Self-organized formation of a set of scaling filters and their neighbouring connections

A set of scaling feedforward filters is developed in an unsupervised way via inputting pixel-discretized extended objects into a winner-take-all artificial neural network. The system discretizes the input space by both position and size. Depending on the distribution of input samples and below a certain number of neurons the spatial filters may form groups of similar filter sizes with each group covering the whole input space in a quasi-uniform fashion. Thus a multi-discretizing system may be formed. Interneural connections of scaling filters are also developed with the help of extended objects. It is shown both theoretically and with the help of numerical simulation that competitive Hebbian learning is suitable for defining neighbours for the multi-discretizing system. Taking into account the neighbouring connections between filters of similar sizes only, i.e. within the groups of filters, the system may be considered as a self-organizing multi-grid system.     

Self-organizing maps for visual feature representation based on natural binocular stimuli

We model the stimulus-induced development of the topography of the primary visual cortex. The analysis uses a self-organizing Kohonen model based on high-dimensional coding. It allows us to obtain an arbitrary number of feature maps by defining different operators. Using natural binocular stimuli, we concentrate on discussing the orientation, ocular dominance, and disparity maps. We obtain orientation and ocular dominance maps that agree with essential aspects of biological findings. In contrast to orientation and ocular dominance, not much is known about the cortical representation of disparity. As a result of numerical simulations, we predict substructures of orientation and ocular dominance maps that correspond to disparity maps. In regions of constant orientation, we find a wide range of horizontal disparities to be represented. This points to geometrical relations between orientation, ocular dominance, and disparity maps that might be tested in experiments. Received: 9 July 1998 / Accepted in revised form: 2 June 1999     

Multiple feature affixation in Seenku plural formation

Nominal plural formation in Seenku involves two surface changes: fronting of the final vowel and raising of the final tone. Diachronically, the plural patterns likely derive from a suffix *- Open image in new window , which has been obscured through the loss of falling sonority diphthongs. By comparing plural formation to other morphophonological processes in Seenku, this paper argues that the plural suffix has been restructured as a featural suffix consisting of two features: the vocalic feature [+front], resulting in vowel fronting, and the tonal feature [+raised], resulting in tone raising. Given the atomic nature of the morphosyntactic feature plural, Seenku plural formation represents a strong case of multiple feature affixation, albeit a case that can be accounted for through Max constraints on feature values (Lombardi 2001, etc.) and Realize-Morpheme (van Oostendorp 2005; Trommer 2012) rather than the constraint Max-Flt argued for by Wolf (2007). A level-ordered approach retaining the vocalic suffix is also considered but is shown to suffer from a number of shortcomings, particularly with respect to tone and a challenging class of nasal stems. In short, this paper explores how the synchronic grammar copes with the vestiges of affixal morphology in a language that has undergone heavy reduction.     

Modelling of batch fermentation and estimation of state variables using self-organizing feature maps

Summary A self-organizing feature map was used for modelling of batch yeast cultures. The model was constructed by training the neural network with experimental data of the specific rates. Estimates of state variables were obtained from the neural network model and differential mass balance equations via integration. They were compared with the experimental data. The neural network model showed a good modelling accuracy and interpolation capability.     

Self-organized discrimination of resources

When selecting a resource to exploit, an insect colony must take into account at least two constraints: the resource must be abundant enough to sustain the whole group, but not too large to limit exploitation costs, and risks of conflicts with other colonies. Following recent results on cockroaches and ants, we introduce here a behavioral mechanism that satisfies these two constraints. Individuals simply modulate their probability to switch to another resource as a function of the local density of conspecifics locally detected. As a result, the individuals gather at the smallest resource that can host the whole group, hence reducing competition and exploitation costs while fulfilling the overall group's needs. Our analysis reveals that the group becomes better at discriminating between similar resources as it grows in size. Also, the discrimination mechanism is flexible and the group readily switches to a better suited resource as it appears in the environment. The collective decision emerges through the self-organization of individuals, that is, in absence of any centralized control. It also requires a minimal individual cognitive investment, making the proposed mechanism likely to occur in other social species and suitable for the development of distributed decision making tools.     

Self-organized formation of polarized cortical tissues from ESCs and its active manipulation by extrinsic signals

Here, we demonstrate self-organized formation of apico-basally polarized cortical tissues from ESCs using an efficient three-dimensional aggregation culture (SFEBq culture). The generated cortical neurons are functional, transplantable, and capable of forming proper long-range connections in vivo and in vitro. The regional identity of the generated pallial tissues can be selectively controlled (into olfactory bulb, rostral and caudal cortices, hem, and choroid plexus) by secreted patterning factors such as Fgf, Wnt, and BMP. In addition, the in vivo-mimicking birth order of distinct cortical neurons permits the selective generation of particular layer-specific neurons by timed induction of cell-cycle exit. Importantly, cortical tissues generated from mouse and human ESCs form a self-organized structure that includes four distinct zones (ventricular, early and late cortical-plate, and Cajal-Retzius cell zones) along the apico-basal direction. Thus, spatial and temporal aspects of early corticogenesis are recapitulated and can be manipulated in this ESC culture.     

Self-organized patterns of airway narrowing

The behavior of respiratory diseases such as asthma and COPD may involve complicated interactions among multiple factors. Theoretical and experimental data suggest that interdependence among the airways of the bronchial tree leads to the emergence of self-organized patterns of airway narrowing, ventilation defects, and other phenomena when a tipping point is passed. Additionally, evidence from several studies shows that the behavior of an isolated airway is different from an identical airway embedded in the bronchial tree so that experimental results of isolated elements such as airways, airway smooth muscle, or inflammatory pathways may not explain the whole organ behavior. However, there may be factors in the isolated elements that can dramatically change the complex system's behavior. More effective strategies for prevention or recovery from a disease, such as asthma, will depend on our progress in identifying and understanding the essential parts of the self-organized behavior that is involved.     

Eph and ephrin signaling in the formation of topographic maps

The axonal connections between the retina and its midbrain target, the superior colliculus (SC), is mapped topographically, such that the spatial relationships of cell bodies in the retina are maintained when terminating in the SC. Topographic map development uses a Cartesian mapping system such that each axis of the retina is mapped independently. Along the nasal-temporal mapping axis, EphAs and ephrin-As, are graded molecular cues required for topographic mapping while the dorsal-ventral axis is mapped in part via EphB and ephrin-Bs. Because both Ephs and ephrins are cell surface molecules they can signal in the forward and reverse directions. Eph/ephrin signaling leads to changes in cytoskeletal dynamics that lead to actin depolymerization and endocytosis guiding axons via attraction and repulsion.     

Genomes as geography: using GIS technology to build interactive genome feature maps

Background  

Many commonly used genome browsers display sequence annotations and related attributes as horizontal data tracks that can be toggled on and off according to user preferences. Most genome browsers use only simple keyword searches and limit the display of detailed annotations to one chromosomal region of the genome at a time. We have employed concepts, methodologies, and tools that were developed for the display of geographic data to develop a Genome Spatial Information System (GenoSIS) for displaying genomes spatially, and interacting with genome annotations and related attribute data. In contrast to the paradigm of horizontally stacked data tracks used by most genome browsers, GenoSIS uses the concept of registered spatial layers composed of spatial objects for integrated display of diverse data. In addition to basic keyword searches, GenoSIS supports complex queries, including spatial queries, and dynamically generates genome maps. Our adaptation of the geographic information system (GIS) model in a genome context supports spatial representation of genome features at multiple scales with a versatile and expressive query capability beyond that supported by existing genome browsers.     

Application of Kohonen's self-organizing feature map algorithm to cortical maps of orientation and direction preference

Cortical maps of orientation preference in cats, ferrets and monkeys contain numerous half-rotation point singularities. Experimental data have shown that direction preference also has a smooth representation in these maps, with preferences being for the most part orthogonal to the axis of preferred orientation. As a result, the orientation singularities induce an extensive set of linear fractures in the direction map. These fractures run between and connect nearby point orientation singularities. Their existence appears to pose a puzzle for theories that postulate that cortical maps maximize continuity of representation, because the fractures could be avoided if the orientation map contained full-rotation singularities. Here we show that a dimension-reduction model of cortical map formation, which implements principles of continuity and completeness, produces an arrangement of linear direction fractures connecting point orientation singularities which is similar to that observed experimentally. We analyse the behaviour of this model and suggest reasons why the model maps contain half-rotation rather than full-rotation orientation singularities.     

Self-organization of associative memory and pattern classification: recurrent signal processing on topological feature maps

We extend the neural concepts of topological feature maps towards self-organization of auto-associative memory and hierarchical pattern classification. As is well-known, topological maps for statistical data sets store information on the associated probability densities. To extract that information we introduce a recurrent dynamics of signal processing. We show that the dynamics converts a topological map into an auto-associative memory for real-valued feature vectors which is capable to perform a cluster analysis. The neural network scheme thus developed represents a generalization of non-linear matrix-type associative memories. The results naturally lead to the concept of a feature atlas and an associated scheme of self-organized, hierarchical pattern classification.