Meeting Notes

• Date: 2025-10-14
• Time: 09:00 (PT)
• Location: Teams Meeting
• Presentations: SLAP2 data analysis

Agenda

1. Brief update on data collection by @jeromelecoq
2. Slap2 analysis by @Dedalus9

Meeting Recording

Meeting Notes

Update on Neuropixel Data Collection: Jerome provided an update on the start of real mice work for neuropixel data, with initial recordings expected in early November, and discussed recent debugging efforts with Carter and Ryan to streamline data packaging.

Timeline and Expectations: Jerome announced that the team is beginning real mice work for neuropixel data this week, with the first recordings anticipated in early November. The sensory motor design, zebra, trippy, and receptor fields will be included in these recordings.

Data Packaging Debugging: Jerome described recent debugging activities with Carter and Ryan from the Neuropixel team to ensure efficient data packaging, aiming to expedite the process for upcoming recordings.

Electrode Placement and Area Identification: Karim inquired about the localization of electrode subsets in the neuropixel recordings, and Severine explained that a software tool will provide files indicating which channels correspond to specific brain areas, such as V1 or LGN.

Planning Neuropixel Experiments: Karim requested guidance on planning neuropixel experiments for targeting specific brain areas, and Jerome suggested posting on the forum or reaching out directly for further discussion and standardization.

Comparison of PCA Analyses on SLAP2 Data: Nicholas presented a comparison between his PCA analysis methodology on SLAP2 data and Jesse's approach, highlighting differences in temporal scale and feature engineering, with Jerome, Alexander, and Karim contributing to the discussion on methodological implications and future directions.

Nicholas's PCA Methodology: Nicholas explained his preprocessing pipeline, which computes orientation tuning vectors for pre and post blocks to capture relative preference changes in each ROI, focusing on block-level temporal changes rather than instantaneous activity.

Comparison with Jesse's Approach: Nicholas contrasted his method with Jesse's, noting that Jesse's PCA is based on stimulus presentations and ROI activity, capturing instantaneous differences, while his approach summarizes changes across entire blocks.

Interpretation of Results: Nicholas showed that both analyses suggest the oddball block modulates ROI activity, with his method making these changes more evident at the block level, and discussed the complementary nature of the two approaches.

Discussion on Feature Engineering and Assumptions: Jerome and Nicholas discussed the implications of assuming orientation tuning in the analysis, agreeing that both agnostic and feature-engineered approaches are valuable for understanding network dynamics and learning.

Future Directions: Incorporating Spatial Analysis: Nicholas, Jerome, and Karim emphasized the need to include spatial information in future analyses, suggesting that examining the spatial arrangement of ROIs and their relation to adaptation sources could provide deeper insights.

Suggestions for Further Analysis and Modeling: Alexander and Karim proposed extending the analysis to interneurons, comparing inputs and outputs, and developing mathematical models to distinguish between feedforward and feedback adaptation, with group discussion on technical implementation and hypothesis testing.

Analysis of Interneurons and Output Comparison: Alexander suggested repeating the PCA analysis for interneurons and comparing results to pyramidal cells, as well as relating input data to spiking output from calcium imaging, to identify potential discrepancies and deepen understanding of orientation tuning.

Visualization of Principal Components: Alexander recommended plotting principal components to better interpret their shapes and contributions, rather than treating them as black boxes, which Nicholas agreed could complement the analysis.

Modeling Feedforward and Feedback Adaptation: Karim advocated for constructing mathematical models that integrate both feedforward adaptation and feedback mechanisms, using spatial analysis to help distinguish the sources of observed changes in the data.

Technical Discussion on Spatial PCA: Jerome and Nicholas discussed methods for running PCA across ROIs to capture spatial patterns in orientation tuning changes, considering approaches for scoring and visualizing spatial distributions of principal component weights.

Hypothesis Testing with Simulated Data: Nicholas proposed using mathematical models to generate simulated data under different hypotheses (e.g., apical dendrite involvement) and applying the same PCA analysis to both simulated and empirical data to validate interpretations.

Coordination on Data Sharing and Analysis: Carter announced the imminent upload of new SLAP2 data files to DANDI, and Jerome encouraged Nicholas and Jesse to continue spatial analyses on both new and existing datasets, with Carter planning to notify the group via the forum.

Data Upload Timeline: Carter informed the group that five new SLAP2 data files would be uploaded to DANDI within a day, and he would post an announcement on the forum once available.

Ongoing Analysis Efforts: Jerome encouraged Nicholas and Jesse to proceed with spatial analyses on both the new and older SLAP2 files, emphasizing the importance of integrating these datasets into ongoing research.

Collaboration and Communication with External Groups: Farzaneh updated the group on ongoing discussions with Jeff, who created a forum for further exchange, and plans to share outcomes and comments on GitHub, aiming to reconcile differing results between groups.

Forum and Meeting Coordination: Farzaneh reported that Jeff, who could not attend the meeting, set up a forum for continued discussion, and they plan to meet separately to address differences in results, with updates to be posted on GitHub.