Brain-wide representations of prior information in mouse decision-making

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neural representations of prior information are not well understood.
• Prior information influences decision-making, but its brain-wide encoding is unclear.

Purpose of the Study:

• To investigate the neural encoding of prior probabilities.
• To determine the brain-wide distribution of subjective prior representations.

Main Methods:

• Utilized large-scale Neuropixels recordings and widefield calcium imaging from mice.
• Trained mice on a task with alternating prior probabilities (0.2 and 0.8) of a visual stimulus.
• Analyzed neural activity across various brain regions during decision-making.

Main Results:

• Mice successfully estimated and utilized the changing prior probabilities to enhance decision accuracy.
• The subjective prior was encoded in a significant portion (20-30%) of brain regions.
• Widespread encoding spanned sensory (LGN, V1), motor (M1, M2), and high-level cortical areas (dACC, vlOFC).

Conclusions:

• The brain widely represents subjective prior information, consistent with distributed Bayesian inference models.
• Prior encoding is not limited to decision-making areas but is present across multiple processing levels.
• Large-scale neural recordings provide critical insights into brain-wide prior encoding mechanisms.