Known Data Issues [Updated November 2024)

[rachelstephlee]

Historical Issues for animal behavior

1. Lick issue

TLDR:
Issue is present for any data collected prior to May 6, 2024.

• Smart mice have figured this out and found a heuristic for learning the task
• This is found in all task types (coupled/uncoupled, baited/unbaited), at every stage.

How to find issue:

• Not every session has the issue, but please double check lick pattern

1. Check at ITI if there's an interlick interval peak around 1s or 2s
2. Check if there's a peak of lick rate around the start of no-lick delay.

Issue has been fixed for data collected after May 6, 2024.

• After May, all mice that were trained in the previous set up had a major behavior performance drop because of longer delay time (an aversive thing for the mouse) for early licking.
• Animals that have started training after May have better performance.

Discussion here:
2024-04-30: AIND Behavior Meeting · AllenNeuralDynamics/aind-behavior-blog · Discussion #343

Slides for problematic licking pattern:
Presentation1.pptx

2. Motivation Issue

TLDR:
Issue is present in all data prior to July 4, 2024

• We have a shorter go cue, mouse were not able to hear the go cue because they were too old (and lose hearing for high frequency).

How to find issue:

• Data is potentially usable, as long as lick latency to go cue is normal (90% <500ms), and response rate is higher than 70% in the first 300 trials.

Issue has been fixed after July 4th, 2024

• The fix was to increase the volume of go cue.

Discussion here:
https://github.com/AllenNeuralDynamics/aind-behavior-blog/issues/452

Pipeline Processing issues

Any NWBs created prior to October 2024 are not to be trusted and have alignment issues.

Check the pipeline and the merge history. Rachel's conservative estimate is: https://github.com/AllenNeuralDynamics/NWB-Packaging-FiberPhotometry-Base-Capsule/commit/ae0afd7776527a1781308696283fc1da2327f6e1