Biology subjects

Radulescu, A.

Publications and source records attributed to Radulescu, A..

8 recordsLinked to original sources

bioRxiv2026

Task-Based Value Generalization Correlates With Positive Overgeneralization and Bipolar Symptoms

Positive overgeneralization -- the tendency to generalize from specific successes to broad expectations of future reward -- has been linked to vulnerability to mania. Because positive overgeneralization has only been assessed using self-report measures, we have limited insight into the underlying cognitive process. Here, we introduce a behavioral paradigm that quantifies how reward generalizes to novel stimuli, and characterize individual generalization profiles by fitting psychometric functions to choice data. In an online transdiagnostic study (N = 163), broader task-based generalization was associated with both higher self-reported positive overgeneralization and subclinical bipolar traits. Response times provided converging evidence: a smaller increase in response time as stimuli became less similar to the rewarded stimulus was associated with higher positive overgeneralization. In computational simulations, we show that self-efficacy, operationalized as the relative weight placed on the best and worst available future actions, shapes generalization, with more weight on the best possible action broadening the spread of expected reward to similar states. Together, these findings show that positive overgeneralization can be measured behaviorally, and identify one computational factor that may determine how broadly reward generalizes.

neuroscience↗

bioRxiv2026

Coordinated human prefrontal dynamics sustain task-state representations during learning

Making decisions in complex, real-world environments is challenging. Biologically plausible strategies like reinforcement learning (RL) require attention toward reward-predictive stimuli to define task states, yet how attention and decision processes coordinate in the human brain remains unclear. We hypothesized this arises through interactions between orbitofrontal (OFC) value-based mechanisms and lateral prefrontal (LPFC) attention filtering. To test this, we combined behavioral modeling with local field potential (LFP) and single-unit recordings in 22 subjects performing a multidimensional RL task. Reward expectations were encoded in OFC and LPFC, as reflected in high-frequency LFP and OFC single-unit spiking, but modulated by attention only in LPFC. Theta LFPs encoded reward expectations and indexed attention-dependent LPFC-OFC coordination, with value-related coupling emerging pre-choice in high-attention subjects and post-choice in low-attention subjects. These findings show that prefrontal circuits dynamically coordinate to encode attention-weighted value signals, shaping state representations and providing a tractable solution to learning in complex environments.

neuroscience↗

bioRxiv2026

Decoding Covert Human Attention in Multidimensional Environments

In complex environments, humans must determine which features are relevant for learning and decision-making. Psychological theories offer competing accounts of this process: associative models suggest that attention emerges gradually through learned changes in feature values, whereas hypothesis-driven accounts propose that learners selectively attend to actively tested rules. Because attentional states are covert, similar behavior can arise from different underlying strategies, making these accounts difficult to distinguish using choice data alone. We inferred latent attention dynamics during learning and decision-making by training recurrent neural networks on synthetic data generated from feature-based reinforcement learning (FRL) and serial hypothesis testing (SHT) models. A network trained on hybrid (FRL+SHT) data outperformed single-model networks, decoding latent human attention with more than 80% accuracy. These results suggest that human attention reflects an interaction between value-based learning and hypothesis testing, in which learned feature value guides the generation and evaluation of candidate rules.

animal behavior and cognition↗

bioRxiv2024

Unveiling the latent dynamics in social cognition with multi-agent inverse reinforcement learning

Social behavior requires individuals to consider not only their own goals but also those of others. Latent value functions that encode such goals can be recovered from behavior using inverse reinforcement learning in single-agent settings. However, extending it to multi-agent interactions is challenging, because value functions are defined over joint state spaces that grow exponentially with the number of agents. Existing approaches often manage this complexity by imposing strong structural assumptions about social interactions, thereby limiting their applicability and interpretability. Here we show that joint value functions governing social interactions can be effectively represented through value decomposition into individual value maps for each agent and low-dimensional interaction terms. We develop a multi-agent inverse reinforcement learning framework (MAIRL) to infer these representations from behavior. In mouse and primate social tasks, MAIRL reveals interpretable value maps that are conditioned on the distinct social roles animals play during group behavior. Together, these results establish MAIRL as an interpretable and scalable framework for identifying latent value representations guiding multi-agent behavior across species.

animal behavior and cognition↗

bioRxiv2022

Modeling human eye movements during immersive visual search

The nature of eye movements during visual search has been widely studied in psychology and neuroscience. Virtual reality (VR) paradigms provide an opportunity to test whether computational models of search can predict naturalistic search behavior. However, existing ideal observer models are constrained by strong assumptions about the structure of the world, rendering them impractical for modeling the complexity of environments that can be studied in VR. To address these limitations, we frame naturalistic visual search as a problem of allocating limited cognitive resources, formalized as a meta-level Markov decision process (meta-MDP) over a representation of the environment encoded by a deep neural network. We train reinforcement learning agents to solve the meta-MDP, showing that the agents optimal policy converges to a classic ideal observer model of search developed for simplified environments. We compare the learned policy with human gaze data from a visual search experiment conducted in VR, finding a qualitative and quantitative correspondence between model predictions and human behavior. Our results suggest that gaze behavior in naturalistic visual search is consistent with rational allocation of limited cognitive resources.

neuroscience↗

bioRxiv2021

Estimating the glutamate transporter surface density in mouse hippocampal astrocytes

Glutamate transporters preserve the spatial specificity of synaptic transmission by limiting glutamate diffusion away from the synaptic cleft, and prevent excitotoxicity by keeping the extracellular concentration of glutamate at low nanomolar levels. Glutamate transporters are abundantly expressed in astrocytes, and previous estimates have been obtained about their surface expression in astrocytes of the rat hippocampus and cerebellum. Analogous estimates for the mouse hippocampus are currently not available. In this work, we derive the surface density of astrocytic glutamate transporters in mice of different ages via quantitative dot blot. We find that the surface density of glial glutamate transporters is similar in 7-8 week old mice and rats. In mice, the levels of glutamate transporters increase until about 6 months of age and then begin to decline slowly. Our data, obtained from a combination of experimental and modeling approaches, show that there are stark differences in the density of expression of glutamate transporters across different sub-cellular compartments, indicating that the extent to which astrocytes limit extrasynaptic glutamate diffusion depends not only on their level of synaptic coverage, but also on the identity of the astrocyte compartment in contact with the synapse. Together, these findings provide information on how heterogeneity in the spatial distribution of glutamate transporters in the plasma membrane of hippocampal astrocytes my alter glutamate receptor activation out of the synaptic cleft. Author summaryIn this work, we use experimental and modeling approaches to estimate the surface density of the glial glutamate transporters GLAST and GLT-1 in mouse hippocampal astrocytes. The results show that the surface density of glutamate transporters in hippocampal astrocytes is similar in 7-8 week old mice and rats. In mice, GLAST and GLT-1 reach their peak expression at ~ 6 months of age. Their expression, however, varies widely across different sub-cellular compartments. These findings indicate that the identity of the sub-cellular compartment of an astrocyte in contact with a glutamatergic synapses, in addition to the extent of its synaptic coverage, is a main factor to limit glutamate spillover and receptor activation at a distance from the release site.

neuroscience↗

bioRxiv2020

MATHEMATICAL MODELING OF RETINAL DEGENERATION: AEROBIC GLYCOLYSIS IN A SINGLE CONE

Cell degeneration, including that resulting in retinal diseases, is linked to metabolic issues. In the retina, photoreceptor degeneration can result from imbalance in lactate production and consumption as well as disturbances to pyruvate and glucose levels. To identify the key mechanisms in metabolism that may be culprits of this degeneration, we use a nonlinear system of differential equations to mathematically model the metabolic pathway of aerobic glycolysis in a healthy cone photoreceptor. This model allows us to analyze the levels of lactate, glucose, and pyruvate within a single cone cell. We perform numerical simulations, use available metabolic data to estimate parameters and fit the model to this data, and conduct a sensitivity analysis using two different methods (LHS/PRCC and eFAST) to identify pathways that have the largest impact on the system. Using bifurcation techniques, we find that the system has a bistable regime, biologically corresponding to a healthy versus a pathological state. The system exhibits a saddle node bifurcation and hysteresis. This work confirms the necessity for the external glucose concentration to sustain the cell even at low initial internal glucose levels. It also validates the role of {beta}-oxidation of fatty acids which fuel oxidative phosphorylation under glucose- and lactate-depleted conditions, by showing that the rate of {beta}-oxidation of ingested outer segment fatty acids in a healthy cone cell must be low. Model simulations reveal the modulating effect of external lactate in bringing the system to steady state; the bigger the difference between external lactate and initial internal lactate concentrations, the longer the system takes to achieve steady state. Parameter estimation for metabolic data demonstrates the importance of rerouting glucose and other intermediate metabolites to produce glycerol 3-phosphate (G3P), thus increasing lipid synthesis (a precursor to fatty acid production) to support their high growth rate. While a number of parameters are found to be significant by one or both of the methods for sensitivity analysis, the rate of {beta}-oxidation of ingested outer segment fatty acids is shown to consistently play an important role in the concentration of glucose, G3P, and pyruvate, whereas the extracellular lactate level is shown to consistently play an important role in the concentration of lactate and acetyl coenzyme A. The ability of these mechanisms to affect key metabolites variability and levels (as revealed in our analyses) signifies the importance of inter-dependent and inter-connected feedback processes modulated by and affecting both the RPEs and cones metabolism.

systems biology↗

bioRxiv2019

Cold and distant: structural features of the nucleoprotein complex of a cold-adapted influenza A virus strain

Two influenza A nucleoprotein variants (wt: G102R; and mutant: G102R and E292G) were studied with regard to macro-molecular interactions in oligomeric form (24-mers). The E292G mutation has been previously shown to provide cold adaptation. Molecular dynamics simulations of these complexes and trajectory analysis showed that the most significant difference between the obtained models was distance differences between nucleoprotein complex strands. Influenza virus nucleoprotein complexes were isolated from strains bearing the corresponding NP amino acid substitutions. The isolated complexes were characterized by transmission electron microscopy and differential scanning fluorimetry (DSF). Presence of the E292G substitution was shown by DSF to affect nucleoprotein complex melting temperature. In the filament interface peptide model, it was shown that the peptide corresponding in primary structure to the wild-type NP (SGYDFEREGYS, wild type peptide) is prone to temperature-dependent self-association, unlike the peptide carrying the substitution corresponding to E292G (SGYDFGREGYS, mutant peptide). It was also shown that the SGYDFEREGYS peptide (wt) is capable of interacting with a recombinant full-size monomeric nucleoprotein (with primary structure corresponding to wild type); this interactions equilibrium dissociation constant is five orders of magnitude lower than for the SGYDFGREGYS peptide. Using small-angle neutron scattering (SANS), the supramolecular structures of isolated complexes of these proteins was studied at temperatures of 15, 32, and 37{degrees}C. SANS data show that the structures of the studied complexes (mutant or normal proteins with RNA) at elevated temperature differ from the rod-like particle model and react differently to temperature changes. The data suggest that the mechanism behind cold adaptation with E292G is associated with a weakening of the interaction between strands of the ribonucleoprotein complex and, as a result, the appearance of inter-chain interface flexibility necessary for complex function at low temperature.

biophysics↗