TY  - GEN
KW  - Computational Neuroscience
KW  -  Neurowissenschaften
UR  - https://archiv.ub.uni-heidelberg.de/volltextserver/34191/
Y1  - 2024///
ID  - heidok34191
AV  - public
TI  - Computational Modeling of Time perception and its Dopaminergic Modulation
CY  - Heidelberg
A1  - Ravichandran-Schmidt, Pirathitha
N2  - Coordinated movements, foraging, and other higher-order cognitive tasks such as speech, music, or decision-making are impossible without precise timing. Computational models of interval timing, which ranges from a few hundred milliseconds to several minutes, are expected to provide key insights into the underlying mechanisms of timing, which are to date still largely unknown. So far, existing models have only partially replicated key experimental observations, namely the psychophysical law (linear relation between subjective and objective durations), the dopaminergic modulation, and the scalar property, i.e., the linear increase of the standard deviation of temporal estimates with objective durations. Among a number of brain regions, which, based on experimental observations in humans, might take part in time perception, here I focus on the prefrontal cortex (PFC) as a candidate for interval timing. Previously, various computational models for interval timing were proposed, namely, state-space model (Buonomano, 2000), ramping activity model (Durstewitz, 2003), synfire chains (Hass et al., 2008), and striatal beat model (Miall, 1989; Matell & Meck, 2004). Here, I test two of those four models within a computational PFC model for their ability to replicate experimental observations on time perception by incorporating the state-space model and the ramping activity model into a data-driven PFC model (Hass et al., 2016). 
I show that the combination of the state-dependent and the PFC model into the state-space PFC model, successfully encodes time up to 750 ms and, within this range, reproduces all key experimental observations. Analyzing the underlying mechanisms, I find that the representations of different intervals rely on the natural heterogeneity in the parameters of the network, leading to stereotypic responses of subsets of neurons. Furthermore, we propose a theory for the mechanism underlying subsecond timing in this model, based on correlation and ablation experiments as well as mathematical analyses explaining the emergence of the scalar property and Vierordt?s law. 
The ramping activity model was previously proposed as a time perception model making use of slowly increasing firing rates saturating at different time points through a calcium-dependent after-depolarizing (AD) current. For the ramping PFC model, the calcium-dependent AD current was incorporated into the PFC model and different readout methods for time estimation were conceived and tested for their explicit use of the ramping property. By counting the number of neurons above respective thresholds, a method was found that makes use of the ramping firing rates for time estimation and successfully reproduces all three timing properties in intervals ranging from 500 ? 1500 ms. 
The state-dependent PFC model as well as the ramping PFC model proposed in this work constitute the first data-driven models of interval timing in the range of hundreds of milliseconds to several seconds that have been thoroughly tested against a variety of experimental data. In accordance with the idea of multiple mechanisms responsible for different scales of time perception in nervous systems as proposed in the literature, here, I describe two computational models for interval timing operating on complementary time scales that are in line with experimentally observed connectivities and firing statistics of the prefrontal cortex. With this, the two proposed models provide an ideal starting point for further investigations of interval timing.
ER  -