Questions for Datta Lab

[pauladkisson]

In optoda_raw_data/closed_loop_behavior.parquet,

• [ ] What are the different experiment_types and how do they relate to the paper figures? Especially excitation, excitation_photometry, and excitation_pulsed?
• [x] Is there any difference between 'chr2' and 'chrimson' for opsin?
• [x] Is feedback_status an onset signal, offset signal, or something else?
• [x] In the paper (Methods: Closed-loop stimulation behavioural paradigm), syllables are described as being targeted per-session i.e. for each session, only 1 syllable is targeted with opto-DA stim. But, in the dataframe, some sessions have multiple different values for target_syllable. Could you please explain this data anomaly?

In dlight_raw_data/dlight_photometry_processed_full.parquet,

• [ ] What is the logic for session_number, session_total_number, and stim_session?
• [ ] What is the source of (or rationale for) the discrepancies in the kinematic data? ex. velocity_height != height_ave_mm.diff(2) * 30 / 2
• [x] What are width_mm and length_mm?
• [ ] What is the relationship between syllable IDs in Extended Data Figure 2, predicted_syllable in the dataframe, and the usage mapping in syllable_stats_offline.toml?

In optoda_raw_data/closed_loop_behavior_velocity_conditioned.parquet,

• [ ] What is the difference between target_syllable and trigger_syllable?
• [ ] There are 9 different mouse_ids but the paper only references 5 animals, why?
• [ ] What does 'mean' refer to in trigger_syllable_scalar_threshold?
• [ ] No area field, but based on the paper it seems like the area should be 'snc (axon)' like in the optoda dataframe. Can you confirm?

In GDrive/.../tdt_data_xxx.dat

• [x] Could you clarify the source of this data? Is it the raw output of the TDT ActiveX Interface or converted somehow?
• [x] What are the loading specifications? Is there a standard library for loading and/or documentation from TDT?
• [x] After loading, the photometry_dict has several keys including 'pmt00', 'pmt00_x', 'pmt00_y', 'pmt01', ... can you describe these channels? Does one correspond to the signal and the other the UV reference? x and y?
• [x] Clarification of LED1Amp?

Video files

• [x] The depth.avi video has very coarse resolution, but the output from moseq results_00.mp4 has finer resolution in which you can actually see the mouse throughout. Is MoSeq using the ir.avi video along with depth.avi to generate its output?

In GDrive/.../proc/results_00.h5,

• [x] area_mm states that it should be 'Mouse area (mm)' but shouldn't it be in mm^2? Also, area_mm != length_mm * width_mm. What exactly is this field referring to?

In GDrive/_aggregate_results_ahrmm_scalar_01/session_20211108103843-429331 (velocity modulation?)

• [ ] The SubjectName is dls-ai32-6, which doesn't match any in the processed .parquet files NOR any of the ids listed in the maps in the dopamine-reinforces-spontaneous-behavior/rl_analysis/io/df.py file. Is this an excluded mouse? If so, where can I find a list/map of exclusions. I was thinking that the df.py file covered them, but apparently not.

In the keypoint.p files,

• [ ] there are multiple unused keypoint variables ex. positions_mean, positions_std, etc. do we want to store all of these different keypoint representations or only the ones ultimately used in the notebook for plotting figures (positions_median)?

In the provided sex_map,

• [ ] there are four missing subjects: {'5894', '5893', 'vta-nacc-ctrl-7', '5892', '5891'}. What are their sexes?

[kaijfox]

Linking the reference_frame question from #20:

The reference_frame is a description of the "zero-position" for a Spatial Series like the mouse's position or orientation. For example, "top left", "bottom right", etc.

The question then being what our reference frame is for position/orientation columns in the parquet files.

[pauladkisson]

@kaijfox, please see new questions about extra raw data. Thanks!

[wingillis]

Hi @pauladkisson, I can attempt to answer these questions in bits and pieces over time.

I'll start with the first section - optoda_raw_data/closed_loop_behavior.parquet

• experiment_types: excitation is the main closed-loop reinforcement experiment we performed (see fig. 3a-d for setup); excitation_photometry is the same experimental setup as excitation but in animals what we also measured dLight in (also related to fig. 3a-d); excitation_pulsed follows the same experimental protocol as the closed-loop reinforcement experiment, but we performed ~25Hz pulsed stimulation for 3s instead of the default 250ms continuous stimulation (see fig. 3m).
• 'chr2' mice are DAT-Cre::Ai32 mice, whereas 'chrimson' mice are DAT-Cre mice with a Cre-dependent ChrimsonR virally expressed in the DLS.
• feedback_status refers to stimulation onset. A value of 1 = stimulation, 0 = catch trial (i.e., we detected the target, but didn't stim), -1 = everything else
• Each session in the dataframe has a unique uuid associated with them. If after you've verified that there are sessions with the same uuid but different target_syllables, what is likely happening is that the session is on a recording day that occurred between the two stim sessions for that week. We duplicate those sessions in the dataframe to group "session sequences" (where a sequence contains pre-stim, stim, and post-stim sessions for a target syllable) more easily, and the middle session is the only one shared with both stim sessions (i.e., post-stim for 1 sequence, pre-stim for the other).

[wingillis]

Next, the video files section:

• the depth.avi file contains the raw depth video collected during a session. This means the value associated with each pixel corresponds to the distance the camera is from that point in physical space in millimeters. Thus the mouse won't be clearly visible without some pre-processing. The results_00.mp4 files are a visualization of the extraction process, and thus have intentionally isolated the mouse from the rest of the environment. The ir.avi file is only used for aligning the photometry signal to behavior.

[pauladkisson]

Hey @wingillis,

Thank you for the clarifications! For the experiment_type, is there a difference between excitation_photometry and reinforcement_photometry?

[pauladkisson]

Linking the reference_frame question from #20:

Circling back around to this. Based on moseq-extract code it seems like the reference is 'top left' of the image, but could use a confirmation.