Projects Dataset: IBL behavior

Summary

Eric DeWitt discusses the International Brain Laboratory's behavioral dataset available for NeuroMatch Academy projects. The IBL is a collaboration of experimentalists and theorists exploring behavior and neural data through a two-alternative forced-choice task with mice. The dataset includes over six million choices from 200 mice and is accessible via Data Joint and Open Neurophysiology Environment. DeWitt highlights the comprehensive nature of the dataset, which allows for diverse behavioral analyses, and acknowledges the contributions of the behavioral working group in standardizing tasks and data collection.

Takeaways

• 😀 The International Brain Laboratory (IBL) is a collaboration of experimentalists and theorists studying behavior and neural data.
• 😀 IBL's main task involves a two-alternative force trace detection task that is learned through an automated training protocol.
• 😀 Behavioral data collection utilizes a rig where head-fixed mice respond to visual stimuli by moving a wheel.
• 😀 The training protocol consists of several stages, starting with 50/50 stimulus probability and gradually introducing contrasts.
• 😀 After initial learning, biased stimulus blocks are introduced to assess changes in performance.
• 😀 Over 129 mice from nine labs contributed to the dataset, yielding almost four million choices in the initial dataset.
• 😀 Current updates indicate about 200 mice and nearly six million choices in the new dataset being prepared.
• 😀 Data can be accessed via Data Joint, an open-source framework for data analysis, and the Open Neurophysiology Environment (ONE).
• 😀 The dataset includes comprehensive information for behavioral analysis, such as stimulus on-time, animal choice, and feedback.
• 😀 A core behavioral working group at IBL standardized the tasks and developed the automated training system.

Q & A

What is the main purpose of the International Brain Laboratory (IBL)?

The IBL aims to explore a common behavioral task collaboratively among experimentalists and theorists to develop theoretical models linking behavior and neural data.

What type of behavioral task is being analyzed in the IBL dataset?

The task analyzed is a two-alternative forced-choice detection task, which is learned through an automated training protocol.

How do the animals interact with the task during the experiment?

Animals move a wheel to center a visual stimulus on a screen, receiving a water reward for correct choices and a noise burst for incorrect ones.

What technology is used to record behavioral data in the experiments?

The experiments use high-speed video cameras and neuro pixels to capture data from the animals while they perform the task.

How is the difficulty of the task adjusted?

Difficulty is controlled by varying the contrast of the stimuli and introducing blocks with biased stimulus presentations.

How many mice and choices were involved in the dataset described?

The initial dataset included 129 mice across nine labs, resulting in almost four million choices, with a newer dataset consisting of around 200 mice and nearly six million choices.

What are the two mechanisms available for accessing the IBL dataset?

The data can be accessed via Data Joint, an open-source framework for data analysis, and the Open Neurophysiology Environment (ONE), a standard query system.

What additional information does the IBL dataset provide for analyses?

The dataset includes detailed information such as stimulus onset times, animal choices, contrasts, feedback, and variations across different animals.

Who are the key contributors to the IBL behavioral dataset?

The behavioral working group consists of core members who standardized tasks, developed training systems, and collected the dataset, although many others from the IBL contributed.

How does the IBL plan to support NeuroMatch students in using the dataset?

IBL will provide a single login for NeuroMatch students and produce example collaboration notebooks to facilitate access and demonstrate simple queries on the data.