For best experience please turn on javascript and use a modern browser!
You are using a browser that is no longer supported by Microsoft. Please upgrade your browser. The site may not present itself correctly if you continue browsing.

Dr. J.F. (Jorge) Mejias

Faculty of Science
Swammerdam Institute for Life Sciences

Visiting address
  • Science Park 904
  • Room number: C4.110
Postal address
  • Postbus 94246
    1090 GE Amsterdam
  • Research: computational neuroscience of perception and memory

    The Mejias lab uses a combination of theoretical techniques, computational modeling and data analysis to scout the neural mechanisms underlying large-scale brain communication, hierarchical brain dynamics, sensory predictions and distributed cognitive functions. You can learn more about our research and the organization of the lab below.

    Research line #1: The role of neural heterogeneity in neural dynamics

    Computational models have traditionally considered that neurons in the brain are fairly homogeneous –for example, many network models ignore the existence of different subtypes of neurons, or they neglect the intrinsic variability of physiological properties even within a given neuron subtype. While washing heterogeneity away is convenient to build useful simplifications of the brain, it is now clear that neural heterogeneity plays a major role in neural computations. Our previous work identified neural heterogeneity as a key factor in rate and temporal coding (Mejias and Longtin 2012; 2014) and also uncovered interactions between specific cell types which give rise to paradoxical neural dynamics (Garcia del Molino et al., 2017). We currently study the effects of neural heterogeneity on perception and cognition.

    Impact of neural heterogeneity on neural computations. (A) Two example networks of excitatory (red) and inhibitory (blue) neurons. Neurons in the second network are more heterogeneous, as their firing threshold is cell-specific (denoted by different color tones). An external signal modulates the network firing rate. Networks with firing threshold variability of ~4 mV are strongly responsive to the signal, while homogeneous networks (variability of ~0 mV) are not. (B) Circuit with four different cell types. The analysis of this circuit reveals the existence of two regimes (disinhibition and response reversal) which explain discrepancies between low and high baseline activity as also observed in experimental recordings in mice V1. Modified from Mejias and Longtin (Physical Review Letters 2012; Front. Comput. Neurosci. 2014) and Garcia del Molino et al., eLife 2017.

    Research line #2: Neural dynamics across multiple scales

    Computational models are a perfect way to link neural phenomena at different scales. In previous work, we have built models of the macaque brain spanning several scales, from microscopic circuits to full-brain networks, by incorporating in the models precise anatomical and electrophysiological data. These models correctly predict and reproduce neural dynamics across multiple spatial and temporal scales (Mejias et al., Science Advances 2016), provide biologically plausible solutions for problems in efficient brain communication (Joglekar et al., Neuron 2018) and explore the emergence of working memory and other cognitive functions in distributed brain networks (Mejias and Wang, 2021). We are expanding this approach to model the brain of other animals (such as rodents and humans) and also investigating the use of such models in computational psychiatry, for example to identify potential biomarkers in brain disorders.

    Neural dynamics across multiple scales in the brain. (A) A computational model linking four levels of description: a microscopic circuit, a cortical column with layers 2/3 and 5/6, an inter-areal (V1-V4) network, and a large-scale cortical network constituted by 30 cortical areas of the macaque brain. (B) Neural dynamics of pyramidal neurons in layers 2/3 and 5/6 as predicted by the model. (C) The model predicts frequency-dependent interactions between brain areas. Model predictions at each level may be constrained by neuroanatomical data and validated with electrophysiological findings from multiple labs. Modified from Mejias et al., Science Advances 2016.

    Research line #3: Recurrent neural network for perception and cognition

    Recent work has shown that properly trained recurrent neural networks (RNNs) may be effectively used as a model to explain the computations underlying different perceptual and cognitive functions in the brain (Yang et al. PLoS Comput. Biol. 2015; Yang et al., Nature Neuroscience 2019). Our lab is currently developing biologically plausible RNNs (for example, by explicitly considering excitatory and inhibitory neurons, as in our preliminary work by Dora et al.) to model behavioral tasks such as multisensory integration and decision making. Our goal is to use these models to link the behavioral output observed in animals with underlying neural computations which give rise to such behavior.

    Training RNNs on perceptual and cognitive tasks. (A) A sketch of a recurrent neural network (RNN) constituted by excitatory and inhibitory units. The network may receive several inputs (for example, from different sensory modalities) and its output informs behavioral choices. (B) After training the network with synthetic data similar to the one used in real behavioral tasks, the network reproduces the behavioral patterns of animals trained in the same tasks. An inspection of the underlying, post-training circuitry may reveal computational strategies linked to the task (Dora et al., in preparation).

    Research line #4: Neural mechanisms of sensory prediction

    A final cornerstone of the lab is the understanding of the neural mechanisms of sensory prediction. There is abundant evidence of neural circuits in the brain which generate an “internal model of the world” and generate predictions to match the incoming sensory stimulation. A well-known example which we previously studied is the electrosensory circuit of electric fish, which cancel out redundant or unimportant electrosensory signals (Mejias et al., 2013). A more general scenario includes the theory of predictive coding, by which our brains generate predictions to cancel out familiar sensory signals. As a consequence, novel and unpredictable stimuli –which are not cancelled —are the most effective for driving learning in higher brain areas. Our lab has a strong focus in exploring the neural mechanisms of sensory predictions, and how these predictions may be useful to explain multisensory representations in the brain.


  • Integration in the Cognitive and Systems Neuroscience Group

    Our view is that a close collaboration between experimental and computational neuroscientists is key to successfully unravel the secrets of the brain. Our lab is uniquely well positioned in this sense: we have close interactions with the Olcese, Bosman, Suzuki, de Gee and Pennartz labs, and together we form the Cognitive and Systems Neuroscience Group. The Mejias lab specializes in theoretical and computational approaches to study the brain at different scales –from small neural circuits to full-brain models. With our work, we complement the efforts of our experimental colleagues and also develop our own independent research lines sketched above.

  • Bio-sketch

    I am assistant professor and head of the Computational Neuroscience Lab at the Cognitive and Systems Neuroscience Group of the University of Amsterdam. With a background in physics and mathematics, I obtained a PhD in computational neuroscience from the University of Granada (Spain) in 2009. During my PhD, I was also visiting researcher at the Université Paris V René Descartes (now Université de Paris, France). I then went on to work as a postdoctoral researcher at the University of Ottawa (Canada) and New York University (USA), and as a visiting researcher at the East China Normal University/ NYU Shanghai (China), before joining the University of Amsterdam in 2017. The research of my team is focused on the theoretical and computational study of data-constrained multi-scale brain networks during perception and cognition, and the subsequent development of our advances into brain simulation technologies. Our interest spans several brain functions, including working memory, multisensory integration and predictive coding, as well as brain disorders impairing such functions. I often partake in committee or organizational duties for leading conferences like Cosyne and the OCNS annual meeting, and I currently serve as external faculty member at the Institute Carlos I for Theoretical and Computational Physics in Granada and the European Institute of Theoretical Neuroscience in Paris.

  • Team

    The lab is currently constituted by the following people (digits indicate joint supervision with (1) Matthan Caan, (2) Cyriel Pennartz, (3) Sander Bohte, (4) Umberto Olcese, (5) Fernando Rosas):

    • Jorge Mejias (Principal Investigator)
    • Nikos Priovoulos (postdoc1)
    • Daniel Bodor (research software engineer)
    • Giulia Crocioni (research software engineer)
    • Parva Alavian (PhD student2)
    • Matthias Brucklacher (PhD student2,3)
    • Kwangjun Lee (PhD student2,3)
    • Giulia Moreni (PhD student2)
    • Yulia Demutskaia (MSc student)
    • Erencan Tatar (MSc student)
    • Xinhe Wu (MSc student4)
    • Qingkai Zhu (MSc student5)

    Recent alumni: Abhirup Bandyopadhyay (postdoc), Rares Dorcioman (MSc student), Mengli Feng (MSc student), Thomas Wierda (MSc student), Jochem Pannekoek (BSc student), Licheng Roy Zou (BSc student).

    Do you want to join the lab? Candidates at any level (Master, PhD, postdoc) can contact me by email. We usually have several projects available for interested candidates, and you are also welcome to bring your own ideas and interests to the table. For PhD and postdocs, I am happy to support fellowship applications if we have sufficient time for planning.


    Funding and support. The lab's current and past support comes from the following funding sources:

    • Amsterdam Neuroscience Proof of Concept "Understanding cortico-striatal network dynamics underlying compulsive behavior".
    • Horizon EU Project "Virtual brain twin for personalized treatment of psychiatric disorders".
    • ABC Talent Grant "The cerebellar stripes: cognitive units?"
    • eScience Spearhead Project "Training artificial neural networks for behavioral strategies in neuroscience".
    • ABC Project Grant “Translational biomarkers for compulsivity across large-scale brain networks”.
    • NWA-ORC consortium “Perceptive acting under uncertainty: safety solutions for autonomous systems”.
    • Human Brain Project, SGA3 Task 2.1: “Data-driven model of multisensory object recognition in cortical systems”.


    Collaborators. We work together with other labs and researchers across the Netherlands and abroad, including:

    • At the CSN group: Cyriel Pennartz, Conrado Bosman, Umberto Olcese, Mototaka Suzuki, Jan Willem de Gee.
    • The Netherlands, beyond CSN: Sander Bohte (CWI), Simon van Gaal (UvA), Matthan Caan (AUMC), Ingo Willuhn (NIN), Niels Cornelisse (VU), Chris de Zeeuw, Laurens Bosman and Mario Negrello (EMC).
    • Europe, beyond the Netherlands: Shirin Dora (Ulster University, UK), Henry Kennedy (INSERM, France), Viktor Jirsa (AMU, France), Miguel A. Munoz and Joaquin Torres (UGR, Spain).
    • And beyond: Xiao-Jing Wang (New York University), John D. Murray (Yale University), Guangyu Robert Yang (MIT), Ronaldo Nunes (UFABC), Andre Longtin and Leonard Maler (University of Ottawa), Songting Li and Douglas Zhou (Shanghai Jiao Tong University).
  • Publications


    • Beerendonk, L., Mejías, J. F., Nuiten, S. A., de Gee, J. W., Fahrenfort, J. J., & van Gaal, S. (2024). A disinhibitory circuit mechanism explains a general principle of peak performance during mid-level arousal. Proceedings of the National Academy of Sciences of the United States of America, 121(5), Article e2312898121. [details]
    • Lee, K., Dora, S., Mejias, J. F., Bohte, S. M., & Pennartz, C. M. A. (2024). Predictive coding with spiking neurons and feedforward gist signaling. Frontiers in Computational Neuroscience, 18, Article 1338280. [details]



    • Mejías, J. F., & Wang, X. J. (2022). Mechanisms of distributed working memory in a large-scale network of macaque neocortex. eLife, 11, Article e72136.


    • Homberg, J. R., Adan, R. A. H., Alenina, N., Asiminas, A., Bader, M., Beckers, T., Begg, D. P., Blokland, A., Burger, M. E., van Dijk, G., Eisel, U. L. M., Elgersma, Y., Englitz, B., Fernandez-Ruiz, A., Fitzsimons, C. P., van Dam, A-M., Gass, P., Grandjean, J., Havekes, R., ... Genzel, L. (2021). The continued need for animals to advance brain research. Neuron, 109(15), 2374-2379. [details]
    • Lindeman, S., Hong, S., Kros, L., Mejias, J. F., Romano, V., Oostenveld, R., Negrello, M., Bosman, L. W. J., & De Zeeuw, C. I. (2021). Cerebellar Purkinje cells can differentially modulate coherence between sensory and motor cortex depending on region and behavior. Proceedings of the National Academy of Sciences of the United States of America, 118(2), Article e2015292118. [details]
    • Nunes, R. V., Reyes, M. B., Mejias, J. F., & de Camargo, R. Y. (2021). Directed functional and structural connectivity in a large-scale model for the mouse cortex. Network Neuroscience, 5(4), 874-889. [details]
    • Torres, J. J., Muñoz, M. A., Cortés, J. M., & Mejías, J. F. (2021). Special Issue on Emergent Effects in Stochastic Neural Networks with Application to Learning and Information Processing. Neurocomputing, 461, 632-634. [details]




    • Joglekar, M., Mejias, J. F., Yang, G. R., & Wang, X-J. (2018). Inter-areal balanced amplification enhances signal propagation in a large-scale circuit model of the primate cortex. Neuron, 98.


    • Garcia del Molino, L., Yang, G. R., Mejias, J. F., & Wang, X-J. (2017). Paradoxical response reversal of top-down modulation in cortical circuits with three interneuron types. eLife, 6, Article e29742.
    • Melanson, A., Mejias, J. F., Jun, J. J., Maler, L., & Longtin, A. (2017). Nonstationary Stochastic Dynamics Underlie Spontaneous Transitions between Active and Inactive Behavioral States. eNeuro, 4(2).


    • Mejias, J. F., Murray, J. D., Kennedy, H., & Wang, X-J. (2016). Feedforward and feedback frequency-dependent interactions in a large-scale laminar network of the primate cortex. Sciences advances, 2(11).


    • Mejias, J. F., & Longtin, A. (2014). Differential effects of excitatory and inhibitory heterogeneity on the gain and asynchronous state of sparse cortical networks. Frontiers in Computational Neuroscience, 8.
    • Mejias, J. F., Payeur, A., Selin, E., Maler, L., & Longtin, A. (2014). Subtractive, divisive and non-monotonic gain control in feedforward nets linearized by noise and delays. Frontiers in Computational Neuroscience, 8.


    • Bol, K., Marsat, G., Mejias, J. F., Maier, L., & Longtin, A. (2013). Modeling cancelation of periodic inputs with burst-STDP and feedback. Neural Networks, 47, 120-133.
    • Mejias, J. F., Marsat, G., Bol, K., Maler, L., & Longtin, A. (2013). Learning Contrast-Invariant Cancellation of Redundant Signals in Neural Systems. PLoS Computational Biology, 9(9).





    • Mejias, J. F., & Torres, J. J. (2009). Maximum Memory Capacity on Neural Networks with Short-Term Synaptic Depression and Facilitation. Neural Computation, 21(3), 851-871.


    This list of publications is extracted from the UvA-Current Research Information System. Questions? Ask the library or the Pure staff of your faculty / institute. Log in to Pure to edit your publications. Log in to Personal Page Publication Selection tool to manage the visibility of your publications on this list.
  • Ancillary activities
    • No ancillary activities