Browse
Submit Data
Databases
API
Help

Dataset Information

22 Views

0 Connections

0 Citations

0 Reanalyses

0 Downloads

Omics score: 0

Gated Recurrent Units Viewed Through the Lens of Continuous Time Dynamical Systems.

ABSTRACT: Gated recurrent units (GRUs) are specialized memory elements for building recurrent neural networks. Despite their incredible success on various tasks, including extracting dynamics underlying neural data, little is understood about the specific dynamics representable in a GRU network. As a result, it is both difficult to know a priori how successful a GRU network will perform on a given task, and also their capacity to mimic the underlying behavior of their biological counterparts. Using a continuous time analysis, we gain intuition on the inner workings of GRU networks. We restrict our presentation to low dimensions, allowing for a comprehensive visualization. We found a surprisingly rich repertoire of dynamical features that includes stable limit cycles (nonlinear oscillations), multi-stable dynamics with various topologies, and homoclinic bifurcations. At the same time we were unable to train GRU networks to produce continuous attractors, which are hypothesized to exist in biological neural networks. We contextualize the usefulness of different kinds of observed dynamics and support our claims experimentally.

SUBMITTER: Jordan ID

PROVIDER: S-EPMC8339926 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Json Xml

Similar Datasets

Mitochondria and Peroxisome Remodeling across Cytomegalovirus Infection Time Viewed through the Lens of Inter-ViSTA.

Project description:Nearly all biological processes rely on the finely tuned coordination of protein interactions across cellular space and time. Accordingly, generating protein interactomes has become routine in biological studies, yet interpreting these datasets remains computationally challenging. Here, we introduce Inter-ViSTA (Interaction Visualization in Space and Time Analysis), a web-based platform that quickly builds animated protein interaction networks and automatically synthesizes information on protein abundances, functions, complexes, and subcellular localizations. Using Inter-ViSTA with proteomics and molecular virology, we define virus-host interactions for the human cytomegalovirus (HCMV) anti-apoptotic protein, pUL37x1. We find that spatiotemporal controlled interactions underlie pUL37x1 functions, facilitating the pro-viral remodeling of mitochondria and peroxisomes during infection. Reciprocal isolations, microscopy, and genetic manipulations further characterize these associations, revealing the interplay between pUL37x1 and the MIB complex, which is critical for mitochondrial integrity. At the peroxisome, we show that pUL37x1 activates PEX11β to regulate fission, a key aspect of virus assembly and spread.

| S-EPMC8489195 | biostudies-literature

Time step rescaling recovers continuous-time dynamical properties for discrete-time Langevin integration of nonequilibrium systems.

Project description:When simulating molecular systems using deterministic equations of motion (e.g., Newtonian dynamics), such equations are generally numerically integrated according to a well-developed set of algorithms that share commonly agreed-upon desirable properties. However, for stochastic equations of motion (e.g., Langevin dynamics), there is still broad disagreement over which integration algorithms are most appropriate. While multiple desiderata have been proposed throughout the literature, consensus on which criteria are important is absent, and no published integration scheme satisfies all desiderata simultaneously. Additional nontrivial complications stem from simulating systems driven out of equilibrium using existing stochastic integration schemes in conjunction with recently developed nonequilibrium fluctuation theorems. Here, we examine a family of discrete time integration schemes for Langevin dynamics, assessing how each member satisfies a variety of desiderata that have been enumerated in prior efforts to construct suitable Langevin integrators. We show that the incorporation of a novel time step rescaling in the deterministic updates of position and velocity can correct a number of dynamical defects in these integrators. Finally, we identify a particular splitting (related to the velocity Verlet discretization) that has essentially universally appropriate properties for the simulation of Langevin dynamics for molecular systems in equilibrium, nonequilibrium, and path sampling contexts.

| S-EPMC4065221 | biostudies-literature

Organelle remodeling across infection time: Cytomegalovirus-host protein interactions viewed through the lens of Inter-ViSTA

Project description:Nearly all biological processes rely on the finely-tuned coordination of protein interactions across cellular space and time. Accordingly, generating protein interactome datasets has become routine in biological studies, yet effectively interpreting these datasets remains computationally challenging. Here, we describe the development of Inter-ViSTA, a computational web platform for protein Interaction Visualization in Space and Time Analysis. Inter-ViSTA enables users to quickly build animated interaction networks by automatically synthesizing information on protein abundances, functional annotations, multiprotein complexes, and subcellular localizations. We then leverage Inter-ViSTA, in conjunction with quantitative mass spectrometry and molecular virology, to define the virus-host protein interactions of the human cytomegalovirus (HCMV) anti-apoptotic protein, pUL37x1 (vMIA). We find that spatially and temporally controlled protein interactions underly pUL37x1 functions during infection, facilitating the pro-viral remodeling of both mitochondria and peroxisomes. Reciprocal isolations, microscopy, and CRISPR-based genetic tools further characterize these associations, revealing new mechanisms at the core of the pUL37x1 manipulation of mitochondrial integrity, such as interactions with the MICOS and SAMM50 complexes. Finally, we show that pUL37x1 activates the peroxisomal protein PEX11ß to regulate peroxisome fission during infection, a key metabolic aspect of virus assembly and spread.

2020-08-03 | PXD017163 | Pride

Plasmodium vivax Malaria Viewed through the Lens of an Eradicated European Strain.

Project description:The protozoan Plasmodium vivax is responsible for 42% of all cases of malaria outside Africa. The parasite is currently largely restricted to tropical and subtropical latitudes in Asia, Oceania, and the Americas. Though, it was historically present in most of Europe before being finally eradicated during the second half of the 20th century. The lack of genomic information on the extinct European lineage has prevented a clear understanding of historical population structuring and past migrations of P. vivax. We used medical microscope slides prepared in 1944 from malaria-affected patients from the Ebro Delta in Spain, one of the last footholds of malaria in Europe, to generate a genome of a European P. vivax strain. Population genetics and phylogenetic analyses placed this strain basal to a cluster including samples from the Americas. This genome allowed us to calibrate a genomic mutation rate for P. vivax, and to estimate the mean age of the last common ancestor between European and American strains to the 15th century. This date points to an introduction of the parasite during the European colonization of the Americas. In addition, we found that some known variants for resistance to antimalarial drugs, including Chloroquine and Sulfadoxine, were already present in this European strain, predating their use. Our results shed light on the evolution of an important human pathogen and illustrate the value of antique medical collections as a resource for retrieving genomic information on pathogens from the past.

| S-EPMC7038659 | biostudies-literature

Mediation analysis with intermediate confounding: structural equation modeling viewed through the causal inference lens.

Project description:The study of mediation has a long tradition in the social sciences and a relatively more recent one in epidemiology. The first school is linked to path analysis and structural equation models (SEMs), while the second is related mostly to methods developed within the potential outcomes approach to causal inference. By giving model-free definitions of direct and indirect effects and clear assumptions for their identification, the latter school has formalized notions intuitively developed in the former and has greatly increased the flexibility of the models involved. However, through its predominant focus on nonparametric identification, the causal inference approach to effect decomposition via natural effects is limited to settings that exclude intermediate confounders. Such confounders are naturally dealt with (albeit with the caveats of informality and modeling inflexibility) in the SEM framework. Therefore, it seems pertinent to revisit SEMs with intermediate confounders, armed with the formal definitions and (parametric) identification assumptions from causal inference. Here we investigate: 1) how identification assumptions affect the specification of SEMs, 2) whether the more restrictive SEM assumptions can be relaxed, and 3) whether existing sensitivity analyses can be extended to this setting. Data from the Avon Longitudinal Study of Parents and Children (1990-2005) are used for illustration.

| S-EPMC4383385 | biostudies-literature

Unveiling the complex organization of recurrent patterns in spiking dynamical systems.

Project description:Complex systems displaying recurrent spike patterns are ubiquitous in nature. Understanding the organization of these patterns is a challenging task. Here we study experimentally the spiking output of a semiconductor laser with feedback. By using symbolic analysis we unveil a nontrivial organization of patterns, revealing serial spike correlations. The probabilities of the patterns display a well-defined, hierarchical and clustered structure that can be understood in terms of a delayed model. Most importantly, we identify a minimal model, a modified circle map, which displays the same symbolic organization. The validity of this minimal model is confirmed by analyzing the output of the forced laser. Since the circle map describes many dynamical systems, including neurons and cardiac cells, our results suggest that similar correlations and hierarchies of patterns can be found in other systems. Our findings also pave the way for optical neurons that could provide a controllable set up to mimic neuronal activity.

| S-EPMC3986700 | biostudies-literature

Reverse engineering time discrete finite dynamical systems: a feasible undertaking?

Project description:With the advent of high-throughput profiling methods, interest in reverse engineering the structure and dynamics of biochemical networks is high. Recently an algorithm for reverse engineering of biochemical networks was developed by Laubenbacher and Stigler. It is a top-down approach using time discrete dynamical systems. One of its key steps includes the choice of a term order, a technicality imposed by the use of Gröbner-bases calculations. The aim of this paper is to identify minimal requirements on data sets to be used with this algorithm and to characterize optimal data sets. We found minimal requirements on a data set based on how many terms the functions to be reverse engineered display. Furthermore, we identified optimal data sets, which we characterized using a geometric property called "general position". Moreover, we developed a constructive method to generate optimal data sets, provided a codimensional condition is fulfilled. In addition, we present a generalization of their algorithm that does not depend on the choice of a term order. For this method we derived a formula for the probability of finding the correct model, provided the data set used is optimal. We analyzed the asymptotic behavior of the probability formula for a growing number of variables n (i.e. interacting chemicals). Unfortunately, this formula converges to zero as fast as , where and . Therefore, even if an optimal data set is used and the restrictions in using term orders are overcome, the reverse engineering problem remains unfeasible, unless prodigious amounts of data are available. Such large data sets are experimentally impossible to generate with today's technologies.

| S-EPMC2654658 | biostudies-literature

Identifying nonlinear dynamical systems via generative recurrent neural networks with applications to fMRI.

Project description:A major tenet in theoretical neuroscience is that cognitive and behavioral processes are ultimately implemented in terms of the neural system dynamics. Accordingly, a major aim for the analysis of neurophysiological measurements should lie in the identification of the computational dynamics underlying task processing. Here we advance a state space model (SSM) based on generative piecewise-linear recurrent neural networks (PLRNN) to assess dynamics from neuroimaging data. In contrast to many other nonlinear time series models which have been proposed for reconstructing latent dynamics, our model is easily interpretable in neural terms, amenable to systematic dynamical systems analysis of the resulting set of equations, and can straightforwardly be transformed into an equivalent continuous-time dynamical system. The major contributions of this paper are the introduction of a new observation model suitable for functional magnetic resonance imaging (fMRI) coupled to the latent PLRNN, an efficient stepwise training procedure that forces the latent model to capture the 'true' underlying dynamics rather than just fitting (or predicting) the observations, and of an empirical measure based on the Kullback-Leibler divergence to evaluate from empirical time series how well this goal of approximating the underlying dynamics has been achieved. We validate and illustrate the power of our approach on simulated 'ground-truth' dynamical systems as well as on experimental fMRI time series, and demonstrate that the learnt dynamics harbors task-related nonlinear structure that a linear dynamical model fails to capture. Given that fMRI is one of the most common techniques for measuring brain activity non-invasively in human subjects, this approach may provide a novel step toward analyzing aberrant (nonlinear) dynamics for clinical assessment or neuroscientific research.

| S-EPMC6719895 | biostudies-literature

Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging.

Project description:There has recently been a concerted effort to derive mechanisms in vision and machine learning systems to offer uncertainty estimates of the predictions they make. Clearly, there are benefits to a system that is not only accurate but also has a sense for when it is not. Existing proposals center around Bayesian interpretations of modern deep architectures - these are effective but can often be computationally demanding. We show how classical ideas in the literature on exponential families on probabilistic networks provide an excellent starting point to derive uncertainty estimates in Gated Recurrent Units (GRU). Our proposal directly quantifies uncertainty deterministically, without the need for costly sampling-based estimation. We show that while uncertainty is quite useful by itself in computer vision and machine learning, we also demonstrate that it can play a key role in enabling statistical analysis with deep networks in neuroimaging studies with normative modeling methods. To our knowledge, this is the first result describing sampling-free uncertainty estimation for powerful sequential models such as GRUs.

| S-EPMC7220220 | biostudies-literature

MDSINE: Microbial Dynamical Systems INference Engine for microbiome time-series analyses.

Project description:Predicting dynamics of host-microbial ecosystems is crucial for the rational design of bacteriotherapies. We present MDSINE, a suite of algorithms for inferring dynamical systems models from microbiome time-series data and predicting temporal behaviors. Using simulated data, we demonstrate that MDSINE significantly outperforms the existing inference method. We then show MDSINE's utility on two new gnotobiotic mice datasets, investigating infection with Clostridium difficile and an immune-modulatory probiotic. Using these datasets, we demonstrate new capabilities, including accurate forecasting of microbial dynamics, prediction of stable sub-communities that inhibit pathogen growth, and identification of bacteria most crucial to community integrity in response to perturbations.

| S-EPMC4893271 | biostudies-literature

OmicsDI is part of the ELIXIR infrastructure

OmicsDI is an Elixir interoperability service. Learn more ›

Tweets

OmicsDI Databases

PRIDE
PeptideAtlas
MassIVE
JPOST Repository
Physiome Model Repository

EGA
EVA
ENA
LINCS
PAXDB
Cell Collective

MetaboLights
Metabolomics Workbench
MetabolomeExpress
GNPS
BioModels
FAIRDOMHub

ArrayExpress
dbGaP
ExpressionAtlas
GEO
NODE

Information

Databases
Help
API
Contact us
Code on GitHub
Terms of use
Submit Data