A piece-wise linear recurrent neural network identifies generalizable dynamics from neural activity during decision-making in rodents
Alterations in impulsivity characterize several neuropsychiatric disorders, however the neural processes that underlie an impulsive choice have not been identified. A recent body of literature suggests that there may be generalizable aspects of neural representations across animals. Whether this is also true for dynamic rather than just static (like geometrical) properties of neural…
Recent research has uncovered alterations in impulsivity among various neuropsychiatric disorders, but the neural processes behind impulsive decision-making remain unclear. A growing body of literature indicates that certain neural representations may be generalizable across different animal species. However, the applicability of these findings to dynamic properties of neural representations, particularly in the context of impulsive behaviors, is still uncertain.
This study aimed to determine if there are generalizable features of latent dynamics that can differentiate impulsive from non-impulsive choices across multiple datasets and animals.
Using a deep learning-based latent factor model, the researchers constructed neural recordings obtained from the anterior cingulate cortex of rats while they performed a delay discounting task. This task is widely used to assess impulsivity and is known to be critical for the performance of this task in rats. To analyze the data, the researchers employed piecewise linear recurrent neural networks (PLRNNs), which are capable of performing nonlinear latent factor analysis across various animals and experimental sessions.
The PLRNN model proved to be more effective than linear approaches, explaining a significantly larger percentage of variance in fewer dimensions. The researchers then cross-validated the model's performance on test data and following the disruption of recurrent connectivity or auto-regressive latent dynamics. These tests confirmed that the PLRNN consistently embedded generalizable neuronal dynamics within its structure.
Additionally, the study found that the latent dynamics exhibited strong rotational features and progressed faster for impulsive choices.
In conclusion, the model successfully captured properties of latent dynamics that are consistent across multiple datasets and effectively distinguished impulsive choices from non-impulsive ones. These findings suggest that the identified generalizable features of latent dynamics may provide valuable insights into the neural basis of impulsive behaviors in neuropsychiatric disorders.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
