Title
Neural mechanisms of predictive processing: a collaborative community experiment through the OpenScope program
Leaders
Prof. Alexander Maier (maierav)
Collaborators
N/A
Project description
The OpenScope Community Predictive Processing Project studies how the brain implements predictive processing:
Does the cortex learn to predict upcoming sensory input, and then signal prediction errors when reality violates expectation?
The entire project is run as an Open Science Team Project. Anyone can join and contribute, and there are both online resources as well as weekly online meetings to offer support. Terabytes of cutting-edge data are waiting to be analyzed by people like you!
The central question we aim to answer is whether different kinds of violation (a sensory oddball, a broken sensorimotor contingency, an omitted stimulus) are computed by distinct specialized circuits (H0) or by a common canonical mechanism repeated across the brain (H1).
To decide between these hypotheses, the same battery of "mismatch" (oddball) paradigms is recorded at three spatial scales in mouse visual cortex, so error signals can be compared from single spikes (Neuropixels) up to population (2-photon imaging) and dendritic (glutamate vesicle imaging) activity. One paradigm, three different techniques.
The conceptual and methodological background is laid out in a community white paper:
Aizenbud et al. (2025), Neural mechanisms of predictive processing: a collaborative community experiment through the OpenScope program. arXiv:2504.09614
Link to project repository/sources
https://github.com/maierav/ai_oscp_neuro
Concerete goals with specific tasks for Brainhack Vanderbilt 2026
Milestone 1 (Low Complexity): Develop an ingestion and visualization script to parse the raw data stored in NWB format. Deliverables include interactive time-series plots. Try your preferred technique first (e.g., single neuron spiking or Ca2+-imaging), and see if you can extend to the other two techniques. Single session examples first, then session averages. The latter will require a bit of book keeping since not all sessions are exactly identical in design.
• Milestone 2 (Medium Complexity): Calculate receptive fields and orientation tuning as well as other basic visual responses for all three techniques and create grand averages across all tested subjects and sessions. Take note of spatial structure, such as laminar differences in the spiking response or spatial patterns for the imaging responses. This step will require us to align multiple types of data, such as average responses and neuronal location or neuron type.
• Milestone 3 (High Complexity): Go all in and compute mismatch (oddball) responses. Choose one stimulus type and one technique first, then aim to expand to all stimulus types across all three techniques. As a final step, we will compare how response variance might relate to some of the neuronal properties (such as tuning) that we calculated previously.
Good first issues
1. issue one:
Data Cleaning & Integrity Check (Python recommended): Write an ingestion script to load the raw NWB files from the DANDI online archive. Because neural data is noisy, run basic checks that allow for evaluation of signal-to-noise, such as visual responses to the receptive field mapping stimuli.
2. issue two:
Signal Preprocessing & Visualization (Python recommended): Implement a script that plots example responses as time series as well as average time series within and across sessions.
Skills
Python (Intermediate to Advanced): Familiarity with time series, matrix-based neural time-series data, and basic signal processing (mean, median, SEM, CI, t-test, ANOVA etc.).
• Data Visualization (Basic to Intermediate): Ability to create labeled plots and 2D visualizations.
• Non-Coding Systems Neuroscience Knowledge (Basic): Ability to interpret behavioral tasks, know or learn about the sampled neocortical brain areas.
Onboarding documentation
https://allenneuraldynamics.github.io/openscope-community-predictive-processing/
NOTE: There is a lot of information collected on this web site. You do not need to read, let alone know, all of this information. The FAQ will probably answer most of your questions: https://allenneuraldynamics.github.io/openscope-community-predictive-processing/faq/
What will participants learn?
Participants will gain hands-on experience processing and manipulating intracranial neurophysiology data using free, open source data and computational tools.
Participants will develop practical skills in implementing time-resolved analyses on noisy biological data, alongside advanced feature extraction and signal processing techniques toward well-defined open scientific questions.
Participants will also gain familiarity with the OpenScope Community Project, and gain the opportunity to stay on as collaborators, with potential future co-authorship on peer-reviewed publications.
Public data to use
https://allenneuraldynamics.github.io/openscope-community-predictive-processing/data-access/
Number of collaborators
1
Credit to collaborators
All contributors will be prominently listed on the project’s primary repository documentation. Outstanding or prolonged contributions that go significantly beyond what has already been established by the community project qualify for co-authorship on future academic manuscripts stemming from this work.
Image
Project Summary
Predictive coding theorizes that brains predict sensory signals. Using open science data, we will test whether one computation underlies all predictive coding.
Type
pipeline_development, visualization
Development status
2_releases_existing
Topic
data_visualisation, hypothesis_testing, neural_networks, physiology, reproducible_scientific_methods, systems_neuroscience
Tools
NWB
Programming language
Python
Modalities
behavioral, Neurophysiology, other
Git skills
1_commit_push
Anything else?
No response
Things to do after the project is submitted and ready to review.
Title
Neural mechanisms of predictive processing: a collaborative community experiment through the OpenScope program
Leaders
Prof. Alexander Maier (maierav)
Collaborators
N/A
Project description
The OpenScope Community Predictive Processing Project studies how the brain implements predictive processing:
Does the cortex learn to predict upcoming sensory input, and then signal prediction errors when reality violates expectation?The entire project is run as an Open Science Team Project. Anyone can join and contribute, and there are both online resources as well as weekly online meetings to offer support. Terabytes of cutting-edge data are waiting to be analyzed by people like you!
The central question we aim to answer is whether different kinds of violation (a sensory oddball, a broken sensorimotor contingency, an omitted stimulus) are computed by distinct specialized circuits (H0) or by a common canonical mechanism repeated across the brain (H1).
To decide between these hypotheses, the same battery of "mismatch" (oddball) paradigms is recorded at three spatial scales in mouse visual cortex, so error signals can be compared from single spikes (Neuropixels) up to population (2-photon imaging) and dendritic (glutamate vesicle imaging) activity. One paradigm, three different techniques.
The conceptual and methodological background is laid out in a community white paper:
Aizenbud et al. (2025), Neural mechanisms of predictive processing: a collaborative community experiment through the OpenScope program. arXiv:2504.09614
Link to project repository/sources
https://github.com/maierav/ai_oscp_neuro
Concerete goals with specific tasks for Brainhack Vanderbilt 2026
Milestone 1 (Low Complexity): Develop an ingestion and visualization script to parse the raw data stored in NWB format. Deliverables include interactive time-series plots. Try your preferred technique first (e.g., single neuron spiking or Ca2+-imaging), and see if you can extend to the other two techniques. Single session examples first, then session averages. The latter will require a bit of book keeping since not all sessions are exactly identical in design.
• Milestone 2 (Medium Complexity): Calculate receptive fields and orientation tuning as well as other basic visual responses for all three techniques and create grand averages across all tested subjects and sessions. Take note of spatial structure, such as laminar differences in the spiking response or spatial patterns for the imaging responses. This step will require us to align multiple types of data, such as average responses and neuronal location or neuron type.
• Milestone 3 (High Complexity): Go all in and compute mismatch (oddball) responses. Choose one stimulus type and one technique first, then aim to expand to all stimulus types across all three techniques. As a final step, we will compare how response variance might relate to some of the neuronal properties (such as tuning) that we calculated previously.
Good first issues
1. issue one:
Data Cleaning & Integrity Check (Python recommended): Write an ingestion script to load the raw NWB files from the DANDI online archive. Because neural data is noisy, run basic checks that allow for evaluation of signal-to-noise, such as visual responses to the receptive field mapping stimuli.
2. issue two:
Signal Preprocessing & Visualization (Python recommended): Implement a script that plots example responses as time series as well as average time series within and across sessions.
Skills
Python (Intermediate to Advanced): Familiarity with time series, matrix-based neural time-series data, and basic signal processing (mean, median, SEM, CI, t-test, ANOVA etc.).
• Data Visualization (Basic to Intermediate): Ability to create labeled plots and 2D visualizations.
• Non-Coding Systems Neuroscience Knowledge (Basic): Ability to interpret behavioral tasks, know or learn about the sampled neocortical brain areas.
Onboarding documentation
https://allenneuraldynamics.github.io/openscope-community-predictive-processing/
NOTE: There is a lot of information collected on this web site. You do not need to read, let alone know, all of this information. The FAQ will probably answer most of your questions: https://allenneuraldynamics.github.io/openscope-community-predictive-processing/faq/
What will participants learn?
Participants will gain hands-on experience processing and manipulating intracranial neurophysiology data using free, open source data and computational tools.
Participants will develop practical skills in implementing time-resolved analyses on noisy biological data, alongside advanced feature extraction and signal processing techniques toward well-defined open scientific questions.
Participants will also gain familiarity with the OpenScope Community Project, and gain the opportunity to stay on as collaborators, with potential future co-authorship on peer-reviewed publications.
Public data to use
https://allenneuraldynamics.github.io/openscope-community-predictive-processing/data-access/
Number of collaborators
1
Credit to collaborators
All contributors will be prominently listed on the project’s primary repository documentation. Outstanding or prolonged contributions that go significantly beyond what has already been established by the community project qualify for co-authorship on future academic manuscripts stemming from this work.
Image
Project Summary
Predictive coding theorizes that brains predict sensory signals. Using open science data, we will test whether one computation underlies all predictive coding.
Type
pipeline_development, visualization
Development status
2_releases_existing
Topic
data_visualisation, hypothesis_testing, neural_networks, physiology, reproducible_scientific_methods, systems_neuroscience
Tools
NWB
Programming language
Python
Modalities
behavioral, Neurophysiology, other
Git skills
1_commit_push
Anything else?
No response
Things to do after the project is submitted and ready to review.
Hi @brainhack-vandy/project-monitors my project is ready!