DANDI: archive and share.
The DANDI Archive provides public access to neurophysiology datasets, including NWB recordings and associated behavior. PhotoBench is designed to complement this established infrastructure.
DRIFTLESSPhotoBench.
A community benchmark
for fiber photometry.
This is a demo. Driftless PhotoBench will become fully operational in October 2026.Upload raw data in NWB, prepared with our forthcoming standalone conversion GUI and complete experimental metadata. Contributions will be open data for anyone to download and reuse.
MMiN 2026 · TALK COMPANION
Better comparisons.
Better experiments.
Explore the signal. Connect the behavior.
Ask what a timescale really tells us.
dLight3.8 / PFCREAL RECORDINGS
Large reward Small rewardTrial averages · IM-1138
02FLiP examples
162FLiP reward trials
01Contributing laboratory
Real data. Open questions.
A first step toward shared standards.
A first step toward shared standards.
TALK COMPANION
Monitoring Molecules in Neuroscience 2026· Amsterdam · 8 September
FROM THE TALK / ALI MOHEBIWhat are we talking about
What are we talking about
when we talk about timescales?
A longer response and a longer reward-history horizon answer different questions. Explore what changes when we change the measurement.
01Release + clearance
02Sensor response
03Optics + sampling
04Analysis window
The response persists.
Simultaneous dLight3.8 · PFC + NAcc · 60 Hz
PFC60% of peak
Retained at 9–11 s · peak 1.32% dF/F
NAcc24% of peak
Retained at 9–11 s · peak 47.61% dF/F
21 long-ITI rewards per animal; 2 animals total. Raw 470 nm, local −5 to 0 s median baseline; no isosbestic regression. 100 ms display averages; peaks and late fractions from unsmoothed 60 Hz source traces. Sensor identity: lab-supplied.
NOW CHANGE THE QUESTION
Which region is “slower”?
Fit an exponential to the full-session autocorrelation. The regional ordering depends on the lag interval.
PFC
0.71 s
NAcc
6.85 s
NAcc has the longer fitted τ here.
Source fits, rounded to 0.01 s. Lower fit lag also varies (0.04–0.10 s); these reproduce the talk’s intervals. Full-session autocorrelation includes task timing. This is not an estimate of reward-history influence.
Explore the integration window nextName the quantity. State the window. Preserve the context. A persistent fluorescence response does not by itself establish a longer cortical reward-history horizon.
02 / EXPLORE
Your signal, in context.
RECORDING EXPLORER
Source-derived demodLight3.8 / Prefrontal cortex
81 matching trials
Signal + behavior
TRIAL 20 / IM-1138 Filtered mean Selected trial Lick contacts0.00 s
Large reward · long ITI
Reward responses
Trial means Large · n = 21 Small · n = 60
Every trial matters
−3 sReward ↑+8 s
Click a row to inspect. Gray cells: neighboring-reward exclusions.
DriftlessFLIP / TimeHarp488 nm excitationSpecies: not suppliedTiter: not supplied
Intensity relative to each trial’s −3 to 0 s median. Means exclude source-flagged neighboring rewards; the selected trial retains them for inspection. These sessions differ in optogenetic pulse duration. Lick contacts include both ports; reward-tracking is not a measure of a reward-history horizon.
03 / FIND YOUR TRADEOFFLess noise.
Less noise.
At what cost?
Longer integration improves precision, but can blur a fast response. Move the slider and follow the same recording through five analysis windows.
IM-1138 · source analysisNoise–resolution playground
100 ms
100 ms200 ms500 ms1 s2 s
Cross-validated peak / SEM10.5×
Peak amplitude retained100%
Boxcar −3 dB bandwidth4.43Hz
All 81 session trials. Exported cross-validated peak / SEM, not baseline-noise SNR. Bandwidth describes the integration filter, not sensor kinetics.
Shared archives. Focused photometry benchmarks.
We plan to collaborate with DANDI to connect PhotoBench’s fiber photometry benchmarks with the broader neurophysiology data-sharing ecosystem.
PhotoBench: benchmark and compare.
Our focus is photometry-specific quality assessment, noise–resolution tradeoffs and comparisons across sensors, laboratories and experimental conditions, with acquisition and behavioral context preserved.
Have a fuller dataset?
We welcome complete sessions and larger studies as well as representative raw recording samples. We can work with you on NWB preparation, DANDI sharing and selecting recordings for PhotoBench benchmarks. Discuss your dataset with the lab.
DANDI collaboration and integration are planned. This demo does not transfer or publish data to DANDI.
BUILT TOGETHERYour next recording could
Your next recording could
be someone else’s reference.
Fully operational in October 2026. Contribute raw data in NWB with experimental metadata, openly available for anyone to download and reuse.
Sensor + preparation Signal + behavior Raw NWB + open access
Demo now · full service in October 2026
Keep the signal honest.
Show units, filtering, exclusions and analysis provenance. Frequency content alone does not establish sensor frequency response.
Match the experiment.
Compare sensor, region, species, behavior and hardware. The talk’s sensor summary has two animals per sensor; the interactive FLiP examples use dLight3.8. Neither supports a clean sensor ranking.
Make context portable.
Submit raw recordings in NWB. Our standalone GUI will collect sensor, virus, animal, anatomy, hardware and behavioral metadata during conversion. Contributions will be openly downloadable.
Demo excerpts supplied by Mohebi & Associates. Lifetime estimates are omitted pending source QC review. Species, viral titer and expression interval are not inferred. FLiP sensor labels come from acquisition metadata; regional and sensor-summary identities are lab-supplied.
Data provenance and analysis sources
Talk framing follows the revised MMiN plan (7 September 2026). Regional curves and autocorrelation fits come from the paired PFC–NAcc analysis; the sensor summary uses selected best sessions. FLiP reward traces and noise-window summaries come from the per-session analysis exports.
NWB photometry extensionAnalysis/pfc_nacc_same_sensor/outputs/panel_traces.csvSHA-256: 4eab2b8ba653854766093310a95c2f2bd99a29175557e81dde392d0dceb85e66Analysis/pfc_nacc_same_sensor/outputs/kinetics_summary.txtSHA-256: 08b090c9c1e17d8fc4f8095e26b80d56c4551217f2d08ca02f1c92ba73302c4aAnalysis/sensor_comparison/outputs/selected_sessions_summary.csvSHA-256: 1e0f6a410b9dcfc9d9e4ed1993995f7b00cbce8d97334302ba624455b3efe311Figs/trials_aligned.csvSHA-256: a4e089706cbafc9f3425f1407190c7baaafbc2b8952f689fcd434cb29997d9c8Figs/behavior_events_aligned.csvSHA-256: 671b83b6bb429acca235a9bcf520cf8555af2b5defc0288e359f14eee6f95550Figs/reward_aligned_traces.csvSHA-256: c564414b414465d9d76faaa212f602b809112151fbc759d6ea619cede49777b6Figs/Noise_Bandwidth_Optimization/reward_aligned_curves.csvSHA-256: 7d76e5fc76f83f7458ddabb6fd2aa6e51c351126804af56cca4bdd4fe36cbc57Figs/Noise_Bandwidth_Optimization/reward_retention_summary.csvSHA-256: 5ae998981e611383eab75a4a53676f20a58eb55a1b8fd33a3867c8df4af33402
Driftless PhotoBenchA community resource by Mohebi & Associates