Enhanced Tactile Coding in Rat Neocortex Under Darkness

Abstract Sensory systems are known for their adaptability, responding dynamically to changes in environmental conditions. A key example of this adaptability is the enhancement of tactile perception in the absence of visual input. Despite behavioral studies showing visual deprivation can improve tactile discrimination, the underlying neural mechanisms, particularly how tactile neural representations are reorganized during visual deprivation, remain unclear. In this study, we explore how the absence of visual input alters tactile neural encoding in the rat somatosensory cortex (S1). Rats were trained on a custom-designed treadmill with distinct tactile textures (rough and smooth), and local field potentials (LFPs) were recorded from S1 under light and dark conditions. Machine learning techniques, specifically a convolutional neural network, were used to decode the high-dimensional LFP signals. We found that the neural representations of tactile stimuli became more distinct in the dark, indicating a reorganization of sensory processing in S1 when visual input was removed. Notably, conventional amplitude-based analyses failed to capture these changes, highlighting the power of deep learning in uncovering subtle neural patterns. These findings offer new insights into how the brain rapidly adapts tactile processing in response to the loss of visual input, with implications for multisensory integration. Competing Interest Statement The authors have declared no competing interest. Footnotes The revised manuscript explores how tactile neural encoding in the rat somatosensory cortex (S1) reorganizes in response to the removal of visual input. The key findings suggest that, under darkness, tactile representations of textures become more distinct, indicating rapid functional reorganization in S1. Several important revisions and updates were made based on reviewer feedback: 1.Addressing Arousal Effects: While gait was controlled, the authors acknowledge the potential influence of arousal or exploratory behavior, especially in the dark condition. New analyses, including locomotor-speed traces and trial order reversals, show that movement speed remained consistent across conditions, supporting the interpretation that sensory reorganization, rather than arousal, primarily contributed to the observed neural differences. 2.Improved Machine Learning Analyses: The convolutional neural network (CNN) model was enhanced with 5-fold cross-validation to assess its performance and generalization. Additional permutation tests confirmed the significance of the enhanced separability of textures in darkness, ruling out random effects. Supplementary analyses also detailed ability of CNNs to detect high-dimensional features not captured by traditional amplitude-based metrics. 3.Statistical Rigor: The study added more robust statistical methods, such as permutation tests, to quantify neural separability. Silhouette scores were recalculated to provide reliable measures of cluster separability, supporting the conclusion that neural representations were more distinct in the dark. 4.Expanded LFP Analysis: Traditional analyses of LFPs failed to capture subtle differences in neural activity, but the use of deep learning revealed high-dimensional patterns. The study now includes detailed analyses of LFP features and highlights the advantages of CNNs over simpler methods like SVMs, which could not effectively classify the conditions. 5.Clarification of Terminology and Figure Legends: The manuscript improves consistency in terminology, particularly using "attribution score" and "population dynamics" for LFP analyses, and refines figure legends for clarity. In conclusion, the revisions strengthen the manuscript by providing clearer explanations, more rigorous statistical analyses, and a focused discussion on the observed neural reorganization, which could inform future studies on sensory adaptation.