MY RESEARCH INTERESTS

The main motivation of my research is to gain a deeper understanding of the biological processes underlying computation in neuronal networks. My work focuses on

elucidating the functional roles and interactions of distinct inhibitory neuron populations (e.g., PV, SST, VIP, and L1 neurons),
studying learning and memory through biologically plausible plasticity rules at excitatory and inhibitory synapses,
and, more recently, establishing links either to animal behavior or to computations performed in artificial neural networks.

In my research, I use

numerical simulations of recurrent network models at different levels of abstraction, ranging from population-level models to spiking neuron models with dendritic compartments,
mathematical tools such as mean-field descriptions of circuit dynamics,
the analysis of calcium imaging and voltage imaging data provided by experimental collaborators,
and I highly value close collaboration with experimentalists (see my projects below).

Ongoing (unpublished) projects

Behavioral state-dependent population input–output transformations revealed by voltage imaging

A fundamental principle of neural computation is that neurons integrate synaptic inputs and transform them nonlinearly into spike trains. While population recordings reveal coordinated spiking activity, they rarely capture subthreshold membrane voltage, and (heroic) paired patch-clamp studies suffer from low throughput. This leaves a gap between the classical single-cell understanding of input–output relationships and large-scale population dynamics.

To bridge this gap, I analyzed high-speed in vivo two-photon voltage imaging data from layer 1 (L1) neurons in mouse S1 (Cohen lab, Harvard) and from L2/3 neurons in mouse V1 (Ji lab, Berkeley), and combined these data with data-driven modeling approaches.

I have presented this work as a poster at COSYNE 2025 in Montreal and at the GRC 2025 in Maine. Part of this analysis has been published in Zhong et al. (2025, Nature Methods).

State-dependent dynamics and plasticity gating in inhibitory circuits

Neuronal circuits characterized by a diversity of inhibitory neuron types can appear dauntingly complex. Their operation depends on precise synaptic connectivity, neuronal nonlinearities, and external contextual signals. Experimental work has leveraged optogenetic tools to probe circuit dynamics; however, despite these advances, a comprehensive theoretical framework that elucidates how connectivity, plasticity, nonlinearities, and contextual signals jointly shape circuit dynamics is still lacking.

In this work, we study an inhibitory circuit model that incorporates somatic and dendritic nonlinearities, allowing us to identify the key parameters governing circuit dynamics. We focus on how the state of parvalbumin (PV), somatostatin (SST), and vasoactive intestinal peptide (VIP) interneurons influences circuit activity and responses to perturbations, providing a theoretical testing ground for observations from optogenetic perturbation experiments.

This work has been presented as a poster at the Sculpted Light in the Brain (SLB) Conference 2024 in Paris and at SfN 2024 in Chicago.

Interaction of excitatory and inhibitory plasticity in spiking networks

In this work, led by Michelle Miller (PhD student, Doiron lab), we study the connectivity structures that emerge from interacting excitatory and inhibitory plasticity rules in the presence and absence of spiking correlations.

Disinhibition underlies reward-based differential assembly formation

In this project, led by Satchal Postlewaite (PhD student, Doiron lab) in collaboration with the Oswald lab at the University of Chicago, we study reward-based assembly formation gated by a disinhibitory circuit in the mouse piriform cortex.

Published projects

Untangling the role of distinct inhibitory neurons in cortical circuits

Models with a single inhibitory interneuron class have successfully explained a wide range of cortical phenomena. However, such models typically address how inhibition supports a single function or network dynamic. It is now clear that cortical inhibition is highly diverse, with molecularly defined interneuron classes occupying distinct positions within cortical circuits.

An attractive hypothesis is that interneurons are functionally homogeneous within classes, while different classes perform distinct computational roles.

We show that, in recurrently connected E-PV-SST networks, SST-mediated modulation can simultaneously increase neuronal gain and network stability (Bos*, Miehl* et al., 2024, eLife).

Assembly formation via excitatory plasticity and assembly-based computations

A prominent hypothesis suggests that neuronal assemblies - groups of strongly connected neurons - form the basic building blocks of perception and memory by encoding representations of specific concepts.

While experimental evidence supports the existence and formation of such assemblies, computational models often fall short of showing how assemblies can be flexibly learned and combined for real-world computations. We present a biologically grounded framework in which nonlinear dendritic compartments and functionally distinct inhibitory populations enable assemblies to form, stabilize, and combine across multiple areas without catastrophic forgetting (Onasch*, Miehl* et al, 2025, bioRxiv).
We have also shown that assemblies can form spontaneously in the absence of structured external input (Montangie et al., 2020, PLoS CB).
We review the literature on assembly formation and its computational implications in neural circuits (Miehl*, Onasch* et al., 2023, Journal of Physiology).

Functional roles of inhibitory plasticity: stability, control of excitatory plasticity, and beyond

Synaptic changes underlying learning and memory are believed to be implemented through activity-dependent synaptic plasticity. While many studies have focused on plasticity at excitatory synapses, fewer have examined long-term plasticity at inhibitory synapses. In my main PhD work, I investigated the functional roles of inhibitory synaptic plasticity and its interaction with excitatory plasticity.

We show that a novel form of inhibitory plasticity, which depends nonlinearly on the postsynaptic firing rate, is sufficient to homeostatically stabilize firing rates and excitatory weight dynamics while simultaneously enabling the learning of new representations (Miehl & Gjorgjieva, 2022, PLoS CB).
In collaboration with the Berry lab at Princeton, we suggest that inhibitory plasticity underlies differential neuronal responses to familiar versus novel stimuli in cortex (Schulz*, Miehl* et al., 2021, eLife).
In collaboration with the Froemke lab at NYU, we studied how homosynaptic and heterosynaptic plasticity at excitatory and inhibitory synapses can establish a set point for cortical excitation–inhibition balance (Field et al., 2020, Neuron).
We summarize the recent literature on inhibitory plasticity and outline future directions and open questions (Wu*, Miehl* et al., 2022, Trends in Neurosciences).