Biology subjects

Ranganath, A.

Publications and source records attributed to Ranganath, A..

3 recordsLinked to original sources

bioRxiv2026

Complementary prefrontal and thalamic representational dynamics during response planning

Advance contextual information can guide behavioral responses, but how prospective action representations unfold across prefrontal and thalamic neuronal populations of the cognitive control network remains unclear. Here, we trained mice on a contextual response-planning task in which auditory cues either predicted the upcoming instructed movement or left response identity unresolved until a later instruction. Predictive contexts improved accuracy and produced subthreshold, directionally biased movements, indicating that mice used contextual information before instruction onset. Extracellular recordings revealed prospective response-side information in prelimbic cortex (PL) and mediodorsal thalamus (MD), but with distinct dynamics. MD represented prospective response side earlier during the context epoch and showed a planning-related temporal advance after instruction onset. In contrast, PL exhibited enhanced instruction-epoch coding and stronger cross-epoch generalization, which could be localized to sparse neuronal subpopulations. These complementary representational dynamics suggest distinct but coordinated roles for prefrontal-thalamic circuits in using advance information to support flexible action planning.

neuroscience↗

bioRxiv2026

Closed-loop sensory feedback enables fast and reliable instrumental acquisition in head-fixed mice

Instrumental learning typically requires hundreds to thousands of trials in which subjects learn to link motor responses to sensory cues. In standard rodent protocols, response accuracy is reported only at trial end, preventing subjects from correcting erroneously initiated responses. We hypothesized that within-trial, closed-loop sensory feedback would accelerate instrumental learning by providing real-time information about response correctness. Head-fixed mice performed a two-alternative forced-choice task by rotating a choice wheel in response to sensory cues. Mice received either no feedback (n = 18), auditory feedback (n = 16), or audiovisual feedback (n = 4) coupled to wheel movements. Feedback-receiving mice required significantly fewer trials to reach 70 % accuracy criterion (median: 3186, 4918 and 7329 trials for multimodal, unimodal and no feedback, respectively; p = 0.0245) and showed higher accuracy when modifying choices (expert stage: 17 %, 11 % and 9 % accuracy in trials with modified choices for multimodal, unimodal and no feedback, respectively; p=1.04x10-). Only feedback mice displayed movement refinements across training (p = 8.02x10-, p = 4.07x10-{superscript 1} and p = 0.2602 for multimodal, unimodal and no feedback, respectively). In summary, closed-loop sensory feedback accelerated instrumental acquisition, demonstrating its value as routine training protocol. HIGHLIGHTSO_LIMice provided with feedback require fewer trials to reach expert stage in an instrumental learning task C_LIO_LIMice provided with feedback perform with higher accuracy in trials involving changes of mind C_LIO_LIMovement trajectories of mice provided with feedback undergo refinement as training advances C_LIO_LISensory feedback can be used as a training aid to accelerate instrumental learning C_LI

animal behavior and cognition↗

bioRxiv2025

SLAB: Simultaneous Labeling And Binding affinity prediction for protein-ligand structures

Machine learning models are often used as scoring functions to predict the binding affinity of a protein-ligand complex. These models are trained with limited amounts of data with experimentally measured binding affinity values. A large number of compounds are labeled inactive throughs single-concentration screens without measuring binding affinities. These inactive compounds, along with the active ones, can be used to train binary classification models, while regression models are trained using compounds with binding affinities only. However, the classification and regression tasks are often handled separately, without sharing the learned feature representations. In this paper, we propose a novel model architecture that jointly performs regression and classification objectives, aiming to maximize data utilization and improve predictive performance by leveraging two complementary tasks. In our setup, the regression yields the binding affinity, whereas the classification task yields the label as active or inactive. We demonstrate our method using PDBbind, the standard 3D structure database, as well as a dataset of flavivirus protease compounds with binding affinity data. Our experiments show that the new joint training strategy improves the accuracy of the model, increasing applicability in various practical drug screening scenarios.

pharmacology and toxicology↗