Biology subjects

Diaconescu, A.

Publications and source records attributed to Diaconescu, A..

4 recordsLinked to original sources

bioRxiv2025

Ketamine and Psilocybin Differentially Impact Sensory Learning During the Mismatch Negativity

Ketamine and psilocybin show potential as therapies for various mental illnesses, including major depressive disorder. However, further investigation into their neural mechanisms is required to understand their effects on the brain. By combining computational modelling with electroencephalography (EEG), we examine the effects of ketamine and psilocybin on hierarchical sensory pwPE learning in the context of the auditory mismatch negativity, an event-related potential consistently shown to be reduced under psychotomimetic interventions. We employed a Bayesian framework and re-analyzed a previously acquired EEG dataset (Schmidt et al., 2012) by modelling single-trial EEG data using the Hierarchical Gaussian Filter. Using a placebo-controlled within-subject crossover design, healthy subjects were administered either S-ketamine or psilocybin during an auditory roving paradigm of pure sinusoidal tones. Our findings elucidate distinct neural impacts of ketamine and psilocybin on sensory learning: ketamine led to a larger reduction in the effect of sensory precision compared to placebo from 207 to 316 ms peaking at 277 ms in the frontal central channels, while psilocybin showed no significant effect. Both drugs reduced the expression of belief precision between 160 to 184 ms, peaking at 172 ms. For higher-level volatility pwPEs, ketamine reduced the expression at 312 ms while psilocybin had a null effect. For perception of elementary imagery, ketamine had a greater effect than psilocybin on sensory and volatility precision, while psilocybin had a greater effect on volatility pwPEs. Our findings suggest hallucinogens have distinct effects on sensory learning that could inform tailored therapies for major depression.

neuroscience↗

bioRxiv2025

Comparative Computational Modeling of Approach-Avoidance Biases in Suicidal Populations via Hierarchical Bayesian Inference

Pavlovian "approach or avoid" impulses are critical behavioral biases that, in excess, are linked to multiple psychiatric conditions. To investigate how such biases contribute to suicidal thoughts and behaviors, we analyzed data from two clinical populations completing an aversive Go/NoGo task. This task disentangles motor action (Go or NoGo) from outcome valence (escape from, or avoidance of, an aversive stimulus), enabling the isolation of Pavlovian biases from instrumental learning processes. We compared multiple computational models that had previously been proposed to explain Pavlovian tendencies, including reinforcement learning, active inference, and drift diffusion-based approaches. We employed a hierarchical Bayesian inference procedure that treats model identity as a random factor at the individual level, allowing an unbiased determination of which mechanisms most accurately captured participants behavior. Across both datasets, models featuring Pavlovian context biases plus a value-decay mechanism best accounted for performance. By contrast, policy-based Pavlovian models and more complex approaches, such as those integrating working memory or active inference, were supported by fewer study participants. These findings suggest that reflexive biases exert a persistent influence on decision-making, and that value decay plays a critical role in shaping behavior over time. Our results demonstrate the importance of systematically comparing and accounting for relevant cognitive processes to explain observed task behaviors. Understanding the factors contributing to task performance may help clarify how Pavlovian tendencies relate to psychopathology, including, in our case, elevated suicide risk. Finally, we illustrate how a complete hierarchical model selection framework can be applied to identify the most plausible mechanisms underlying Pavlovian biases, offering a robust approach for advancing our understanding of task behaviors and establishing clinical utility in future studies. Author summaryAutomatic "approach or avoid" reactions shape behavior, particularly in stressful or negative situations. In this study, we explored how these reflex-like tendencies might contribute to suicidal thoughts and behaviors. Two clinical groups completed a computerized task measuring responses to unpleasant sounds. Participants made either active responses (pressing a button to stop a sound) or passive responses (refraining from pressing to avoid starting a sound), allowing us to examine the interplay of automatic impulses and learning from past experiences. Our analysis showed that behavior was best explained by a model combining stable "approach or avoid" impulses with a forgetting process that reduced reliance on past experiences over time. More complex models involving strategies or memory-based control were less effective. These findings suggest that individuals with suicidal tendencies may rely on persistent reflex-like behaviors and over-index recent outcomes, compromising their ability to learn in uncertain environmental conditions. Understanding these cognitive processes provides insights into why some individuals feel trapped in harmful patterns of thought and behavior. Our work highlights how identifying shared traits in clinical populations using model-based methods can inform targeted mental health interventions and improve our understanding of cognitive functioning across disorders.

neuroscience↗

bioRxiv2022

Deep Learning-Based Parameter Estimation for Neurophysiological Models of Neuroimaging Data

AO_SCPLOWBSTRACTC_SCPLOWConnectome-based neural mass modelling is the emerging computational neuroscience paradigm for simulating large-scale network dynamics observed in whole-brain activity measurements such as fMRI, M/EEG, and related techniques. Estimating physiological parameters by fitting these models to empirical data is challenging however, due to large network sizes, often physiologically detailed fast-timescale system equations, and the need for long (e.g. tens of minutes) simulation runs. Here we introduce a novel approach to connectome-based neural mass model parameter estimation by employing optimization tools developed for deep learning. We cast the system of differential equations representing both neural and haemodynamic activity dynamics as a deep neural network, implemented within a widely used machine learning programming environment (PyTorch). This allows us to use robust industry-standard optimization algorithms, automatic differentiation for computation of gradients, and other useful functionality. The approach is demonstrated using a connectome-based network with nodal dynamics specified by the two-state RWW mean-field neural mass model equations, which we use here as a model of fMRI-measured activity and correlation fluctuations. Additional optimization constraints are explored and prove fruitful, including restricting the model to domains of parameter space near a bifurcation point that yield metastable dynamics. Using these techniques, we first show robust recovery of physiological model parameters in synthetic data and then, as a proof-of-principle, apply the framework to modelling of empirical resting-state fMRI data from the Human Connectome Project database. For resting state activity, the system can be understood as a deep net that receives uncorrelated noise on its input layer, which is transformed into network-wide modelled functional connectivity on its output layer. This is consistent with the prevailing conception in theoretical neuroscience of resting-state functional connectivity patterns as an emergent phenomenon that is driven by (effectively) random activity fluctuations, which are then in turn spatiotemporally filtered by anatomical connectivity and local neural dynamics.

neuroscience↗

bioRxiv2018

Catecholamine Challenge Uncovers Distinct Mechanisms For Direct Versus Indirect, But Not Social Versus Non-Social, Learning

Evidence that social and individual learning are at least partially dissociable sustains the belief that humans possess adaptive specialisations for social learning. However, in most extant paradigms, social information comprises an indirect source that can be used to supplement ones own, direct, experience. Thus, social and individual learning differ both in terms of social nature (social versus non-social) and directness (indirect versus direct). To test whether the dissociation between social and individual learning is best explained in terms of social nature or directness, we used a catecholaminergic challenge known to modulate learning. Two groups completed a decision-making task which required direct learning, from own experience, and indirect learning from an additional source. The groups differed in terms of whether the indirect source was social or non-social. The catecholamine transporter blocker, methylphenidate, affected direct learning by improving adaptation to changes in the volatility of the environment but there was no effect of methylphenidate on learning from the social or non-social indirect source. Thus, we report positive evidence for a dissociable effect of methylphenidate on direct and indirect learning, but no evidence for a distinction between social and non-social. These data fail to support the adaptive specialisation view, instead providing evidence for distinct mechanisms for direct versus indirect learning.

neuroscience↗