arXiv subjects

Jesus M. Cortes

Publications and source records attributed to Jesus M. Cortes.

14 recordsLinked to original sources

arXiv2026

A Tensor Network Framework for Interpretable Graph Analysis of Brain Networks

Identifying robust neurobiological signatures of brain disorders requires machine learning approaches that combine predictive performance with interpretable representations of feature interactions. Here we introduce a quantum-inspired framework based on tensor network machine learning that learns distributed representations of gray-matter features encoded in a Matrix Product State representation, a variational ansatz originally developed for quantum many-body systems. The trained model is then used not only as a classifier but to extract quantum connected correlations between features, which encode higher-order feature interactions and which we map onto a weighted graph. This construction allows us to track, within a single representation, both: (i) the global spectral properties of the network (capturing collective learning dynamics), and (ii) node-level centrality measures (providing interpretable signatures of individual brain regions). Using repeated train-test sampling schemes, we analyze two classification tasks on structural MRI data as examples of complex brain disorders: healthy controls versus schizophrenia and versus bipolar disorder. Node-level analysis identifies a stable set of gray-matter features, most prominently Heschl gyrus, insular cortex, and frontal regions, that act as hubs across multiple centrality measures and across resamplings. These centralities display lower variability across resamplings than Shapley values, supporting the interpretive value of the network representation. The bipolar feature set emerges as a subset of the schizophrenia one, consistent with the hierarchically organized neuroanatomical alterations reported in neuroimaging studies and offering a network-based characterization of this hierarchy.

q-bio.NC↗

arXiv2025

Intrinsic Resonance depends on Network Size of Coupled-Delayed Interacting Oscillators

The collective frequency that emerges from synchronized neuronal populations--the network resonance--shows a systematic relationship with brain size: whole-brain's large networks oscillate slowly, whereas finer parcellations of fixed volume exhibit faster rhythms. This resonance-size scaling has been reported in delayed neural mass models and human neuroimaging, yet the physical mechanism remained unresolved. Here we show that size-dependent resonance follows directly from propagation delays in delay-coupled phase oscillators. Starting from a Kuramoto model with heterogeneous delays, we linearize around the near-synchronous solution and obtain a closed-form approximation linking the resonance $Ω$ to the mean delay and the effective coupling field. The analysis predicts a generic scaling law: $Ω\approx (\sum_j c_{ij} τ)^{-1}$, so resonance is delay-limited and therefore depends systematically on geometric size or parcellation density. We evaluate four growth scenarios--expanding geometry, fixed-volume parcellation, constant geometry, and an unphysical reference case--and show that only geometry-consistent scaling satisfies the analytical prediction. Numerical simulations with heterogeneous delays validate the law and quantify its error as a function of delay dispersion. These results identify a minimal physical mechanism for size-dependent cortical resonance and provide an analytical framework that unifies numeric simulation outputs.

q-bio.NC↗

arXiv2024

Disentangling high order effects in the transfer entropy

Transfer Entropy (TE), the primary method for determining directed information flow within a network system, can exhibit bias - either in deficiency or excess - during both pairwise and conditioned calculations, owing to high-order dependencies among the dynamic processes under consideration and the remaining processes in the system used for conditioning. Here, we propose a novel approach. Instead of conditioning TE on all network processes except the driver and target, as in its fully conditioned version, or not conditioning at all, as in the pairwise approach, our method searches for both the multiplets of variables that maximize information flow and those that minimize it. This provides a decomposition of TE into unique, redundant, and synergistic atoms. Our approach enables the quantification of the relative importance of high-order effects compared to pure two-body effects in information transfer between two processes, while also highlighting the processes that contribute to building these high-order effects alongside the driver. We demonstrate the application of our approach in climatology by analyzing data from El Niño and the Southern Oscillation.

physics.data-an↗

arXiv2021

Persistence of hierarchical network organization and emergent topologies in models of functional connectivity

Functional networks provide a topological description of activity patterns in the brain, as they stem from the propagation of neural activity on the underlying anatomical or structural network of synaptic connections. This latter is well known to be organized in hierarchical and modular way. While it is assumed that structural networks shape their functional counterparts, it is also hypothesized that alterations of brain dynamics come with transformations of functional connectivity. In this computational study, we introduce a novel methodology to monitor the persistence and breakdown of hierarchical order in functional networks, generated from computational models of activity spreading on both synthetic and real structural connectomes. We show that hierarchical connectivity appears in functional networks in a persistent way if the dynamics is set to be in the quasi-critical regime associated with optimal processing capabilities and normal brain function, while it breaks down in other (supercritical) dynamical regimes, often associated with pathological conditions. Our results offer important clues for the study of optimal neurocomputing architectures and processes, which are capable of controlling patterns of activity and information flow. We conclude that functional connectivity patterns achieve optimal balance between local specialized processing (i.e. segregation) and global integration by inheriting the hierarchical organization of the underlying structural architecture.

cond-mat.dis-nn↗

arXiv2020

Small variation in dynamic functional connectivity in cerebellar networks

Brain networks can be defined and explored through their connectivity. Here, we analyzed the relationship between structural connectivity (SC) across 2,514 regions that cover the entire brain and brainstem, and their dynamic functional connectivity (DFC). To do so, we focused on a combination of two metrics: the first assesses the degree of SC-DFC similarity and the second is the intrinsic variability of the DFC networks over time. Overall, we found that cerebellar networks have a smaller DFC variability than other networks in the brain. Moreover, the internal structure of the cerebellum could be clearly divided in two distinct posterior and anterior parts, the latter also connected to the brainstem. The mechanism to maintain small variability of the DFC in the posterior part of the cerebellum is consistent with another of our findings, namely, that this structure exhibits the highest SC-DFC similarity relative to the other networks studied. By contrast, the anterior part of the cerebellum also exhibits small DFC variability but it has the lowest SC-DFC similarity, suggesting a different mechanism is at play. Because this structure connects to the brainstem, which regulates sleep cycles, cardiac and respiratory functioning, we suggest that such critical functionality drives the low variability in the DFC. Overall, the low variability detected in DFC expands our current knowledge of cerebellar networks, which are extremely rich and complex, participating in a wide range of cognitive functions, from movement control and coordination to executive function or emotional regulation. Moreover, the association between such low variability and structure suggests that differentiated computational principles can be applied in the cerebellum as opposed to other structures, such as the cerebral cortex.

q-bio.NC↗

arXiv2017

Consensus clustering approach to group brain connectivity matrices

A novel approach rooted on the notion of consensus clustering, a strategy developed for community detection in complex networks, is proposed to cope with the heterogeneity that characterizes connectivity matrices in health and disease. The method can be summarized as follows: (i) define, for each node, a distance matrix for the set of subjects by comparing the connectivity pattern of that node in all pairs of subjects; (ii) cluster the distance matrix for each node; (iii) build the consensus network from the corresponding partitions; (iv) extract groups of subjects by finding the communities of the consensus network thus obtained. Differently from the previous implementations of consensus clustering, we thus propose to use the consensus strategy to combine the information arising from the connectivity patterns of each node. The proposed approach may be seen either as an exploratory technique or as an unsupervised pre-training step to help the subsequent construction of a supervised classifier. Applications on a toy model and two real data sets, show the effectiveness of the proposed methodology, which represents heterogeneity of a set of subjects in terms of a weighted network, the consensus matrix.

q-bio.NC↗

arXiv2015

Identification of redundant and synergetic circuits in triplets of electrophysiological data

Neural systems are comprised of interacting units, and relevant information regarding their function or malfunction can be inferred by analyzing the statistical dependencies between the activity of each unit. Whilst correlations and mutual information are commonly used to characterize these dependencies, our objective here is to extend interactions to triplets of variables to better detect and characterize dynamic information transfer. Our approach relies on the measure of interaction information (II). The sign of II provides information as to the extent to which the interaction of variables in triplets is redundant (R) or synergetic (S). Here, based on this approach, we calculated the R and S status for triplets of electrophysiological data recorded from drug-resistant patients with mesial temporal lobe epilepsy in order to study the spatial organization and dynamics of R and S close to the epileptogenic zone (the area responsible for seizure propagation). In terms of spatial organization, our results show that R matched the epileptogenic zone while S was distributed more in the surrounding area. In relation to dynamics, R made the largest contribution to high frequency bands (14-100Hz), whilst S was expressed more strongly at lower frequencies (1-7Hz). Thus, applying interaction information to such clinical data reveals new aspects of epileptogenic structure in terms of the nature (redundancy vs. synergy) and dynamics (fast vs. slow rhythms) of the interactions. We expect this methodology, robust and simple, can reveal new aspects beyond pair-interactions in networks of interacting units in other setups with multi-recording data sets (and thus, not necessarily in epilepsy, the pathology we have approached here).

q-bio.NC↗

arXiv2015

Information and long term memory in calcium signals

Unicellular organisms are open metabolic systems that need to process information about their external environment in order to survive. In most types of tissues and organisms, cells use calcium signaling to carry information from the extracellular side of the plasma membrane to the different metabolic targets of their internal medium. This information might be encoded in the amplitude, frequency, duration, waveform or timing of the calcium oscillations. Thus, specific information coming from extracellular stimuli can be encoded in the calcium signal and decoded again later in different locations within the cell. Despite its cellular importance, little is known about the quantitative informative properties of the calcium concentration dynamics inside the cell. In order to understand some of these informational properties, we have studied experimental Ca2+ series of Xenopus laevis oocytes under different external pH stimulus. The data has been analyzed by means of information-theoretic approaches such as Conditional Entropy, Information Retention, and other non-linear dynamics tools such as the power spectra, the Largest Lyapunov exponent and the bridge detrended Scaled Window Variance analysis. We have quantified the biomolecular information flows of the experimental data in bits, and essential aspects of the information contained in the experimental calcium fluxes have been exhaustively analyzed. Our main result shows that inside all the studied intracellular Ca2+ flows a highly organized informational structure emerge, which exhibit deterministic chaotic behavior, long term memory and complex oscillations of the uncertainty reduction based on past values. The understanding of the informational properties of calcium signals is one of the key elements to elucidate the physiological functional coupling of the cell and the integrative dynamics of cellular life.

q-bio.QM↗

arXiv2015

Information flow between resting state networks

The resting brain dynamics self-organizes into a finite number of correlated patterns known as resting state networks (RSNs). It is well known that techniques like independent component analysis can separate the brain activity at rest to provide such RSNs, but the specific pattern of interaction between RSNs is not yet fully understood. To this aim, we propose here a novel method to compute the information flow (IF) between different RSNs from resting state magnetic resonance imaging. After haemodynamic response function blind deconvolution of all voxel signals, and under the hypothesis that RSNs define regions of interest, our method first uses principal component analysis to reduce dimensionality in each RSN to next compute IF (estimated here in terms of Transfer Entropy) between the different RSNs by systematically increasing k (the number of principal components used in the calculation). When k = 1, this method is equivalent to computing IF using the average of all voxel activities in each RSN. For k greater than one our method calculates the k-multivariate IF between the different RSNs. We find that the average IF among RSNs is dimension-dependent, increasing from k =1 (i.e., the average voxels activity) up to a maximum occurring at k =5 to finally decay to zero for k greater than 10. This suggests that a small number of components (close to 5) is sufficient to describe the IF pattern between RSNs. Our method - addressing differences in IF between RSNs for any generic data - can be used for group comparison in health or disease. To illustrate this, we have calculated the interRSNs IF in a dataset of Alzheimer's Disease (AD) to find that the most significant differences between AD and controls occurred for k =2, in addition to AD showing increased IF w.r.t. controls.

q-bio.NC↗

arXiv2015

A novel brain partition highlights the modular skeleton shared by structure and function

Elucidating the intricate relationship between brain structure and function, both in healthy and pathological conditions, is a key challenge for modern neuroscience. Recent technical and methodological progress in neuroimaging has helped advance our understanding of this important issue, with diffusion weighted images providing information about structural connectivity (SC) and functional magnetic resonance imaging shedding light on resting state functional connectivity (rsFC). However, comparing these two distinct datasets, each of which can be encoded into a different complex network, is by no means trivial as pairwise link-to-link comparisons represent a relatively restricted perspective and provide only limited information. Thus, we have adopted a more integrative systems approach, exploiting theoretical graph analyses to study both SC and rsFC datasets gathered independently from healthy human subjects. The aim is to find the main architectural traits shared by the structural and functional networks by paying special attention to their common hierarchical modular organization. This approach allows us to identify a common skeleton from which a new, optimal, brain partition can be extracted, with modules sharing both structure and function. We describe these emerging common structure-function modules (SFMs) in detail. In addition, we compare SFMs with the classical Resting State Networks derived from independent component analysis of rs-fMRI functional activity, as well as with anatomical parcellations in the Automated Anatomical Labeling atlas and with the Broadmann partition, highlighting their similitude and differences. The unveiling of SFMs brings to light the strong correspondence between brain structure and resting-state dynamics.

q-bio.NC↗

arXiv2014

Synergy and redundancy in the Granger causal analysis of dynamical networks

We analyze by means of Granger causality the effect of synergy and redundancy in the inference (from time series data) of the information flow between subsystems of a complex network. Whilst we show that fully conditioned Granger causality is not affected by synergy, the pairwise analysis fails to put in evidence synergetic effects. In cases when the number of samples is low, thus making the fully conditioned approach unfeasible, we show that partially conditioned Granger causality is an effective approach if the set of conditioning variables is properly chosen. We consider here two different strategies (based either on informational content for the candidate driver or on selecting the variables with highest pairwise influences) for partially conditioned Granger causality and show that depending on the data structure either one or the other might be valid. On the other hand, we observe that fully conditioned approaches do not work well in presence of redundancy, thus suggesting the strategy of separating the pairwise links in two subsets: those corresponding to indirect connections of the fully conditioned Granger causality (which should thus be excluded) and links that can be ascribed to redundancy effects and, together with the results from the fully connected approach, provide a better description of the causality pattern in presence of redundancy. We finally apply these methods to two different real datasets. First, analyzing electrophysiological data from an epileptic brain, we show that synergetic effects are dominant just before seizure occurrences. Second, our analysis applied to gene expression time series from HeLa culture shows that the underlying regulatory networks are characterized by both redundancy and synergy.

q-bio.QM↗

arXiv2013

Disruption of transfer entropy and inter-hemispheric brain functional connectivity in patients with disorder of consciousness

Severe traumatic brain injury can lead to disorders of consciousness (DOC) characterized by deficit in conscious awareness and cognitive impairment including coma, vegetative state, minimally consciousness, and lock-in syndrome. Of crucial importance is to find objective markers that can account for the large-scale disturbances of brain function to help the diagnosis and prognosis of DOC patients and eventually the prediction of the coma outcome. Following recent studies suggesting that the functional organization of brain networks can be altered in comatose patients, this work analyzes brain functional connectivity (FC) networks obtained from resting-state functional magnetic resonance imaging (rs-fMRI). Two approaches are used to estimate the FC: the Partial Correlation (PC) and the Transfer Entropy (TE). Both the PC and the TE show significant statistical differences between the group of patients and control subjects; in brief, the inter-hemispheric PC and the intra-hemispheric TE account for such differences. Overall, these results suggest two possible rs-fMRI markers useful to design new strategies for the management and neuropsychological rehabilitation of DOC patients.

q-bio.NC↗

arXiv2012

Stochastic Amplification of Fluctuations in Cortical Up-states

Cortical neurons are bistable; as a consequence their local field potentials can fluctuate between quiescent and active states, generating slow 0.5-2 Hz oscillations which are widely known as transitions between Up and Down States. Despite a large number of studies on Up-Down transitions, deciphering its nature, mechanisms and function are still today challenging tasks. In this paper we focus on recent experimental evidence, showing that a class of spontaneous oscillations can emerge within the Up states. In particular, a non-trivial peak around 20 Hz appears in their associated power-spectra, what produces an enhancement of the activity power for higher frequencies (in the 30-90 Hz band). Moreover, this rhythm within Ups seems to be an emergent or collective phenomenon given that individual neurons do not lock to it as they remain mostly unsynchronized. Remarkably, similar oscillations (and the concomitant peak in the spectrum) do not appear in the Down states. Here we shed light on these findings by using different computational models for the dynamics of cortical networks in presence of different levels of physiological complexity. Our conclusion, supported by both theory and simulations, is that the collective phenomenon of "stochastic amplification of fluctuations" -previously described in other contexts such as Ecology and Epidemiology-- explains in an elegant and parsimonious manner, beyond model-dependent details, this extra-rhythm emerging only in the Up states but not in the Downs.

cond-mat.dis-nn↗

arXiv2011

Quantitative Analysis of the Effective Functional Structure in Yeast Glycolysis

Yeast glycolysis is considered the prototype of dissipative biochemical oscillators. In cellular conditions, under sinusoidal source of glucose, the activity of glycolytic enzymes can display either periodic, quasiperiodic or chaotic behavior. In order to quantify the functional connectivity for the glycolytic enzymes in dissipative conditions we have analyzed different catalytic patterns using the non-linear statistical tool of Transfer Entropy. The data were obtained by means of a yeast glycolytic model formed by three delay differential equations where the enzymatic speed functions of the irreversible stages have been explicitly considered. These enzymatic activity functions were previously modeled and tested experimentally by other different groups. In agreement with experimental conditions, the studied time series corresponded to a quasi-periodic route to chaos. The results of the analysis are three-fold: first, in addition to the classical topological structure characterized by the specific location of enzymes, substrates, products and feedback regulatory metabolites, an effective functional structure emerges in the modeled glycolytic system, which is dynamical and characterized by notable variations of the functional interactions. Second, the dynamical structure exhibits a metabolic invariant which constrains the functional attributes of the enzymes. Finally, in accordance with the classical biochemical studies, our numerical analysis reveals in a quantitative manner that the enzyme phosphofructokinase is the key-core of the metabolic system, behaving for all conditions as the main source of the effective causal flows in yeast glycolysis.

q-bio.QM↗