{ "dandiset": "002013", "page": "dandisets/002013/", "generated": "2026-10-19T14:45:08+00:00", "caution": "A failure to reproduce is not a finding that the paper is wrong. It may be a failure of the attempt.", "outcome_counts": { "reproduced": 7, "partially_reproduced": 1, "not_reproduced": 0, "blocked": 5 }, "outcome_counts_with_outside_data": { "reproduced": 1, "partially_reproduced": 0, "not_reproduced": 0, "blocked": 0 }, "outcome_counts_elsewhere": { "reproduced": 0, "partially_reproduced": 0, "not_reproduced": 0, "blocked": 0 }, "attempts": [ { "id": "002013-amjad-2025-01", "dandiset": "002013", "dandiset_version": "0.261012.0921", "paper": "10.1021/acschemneuro.5c00071", "paper_title": "Microinvasive Probes for Monitoring Electrical and Chemical Neural Activity in Nonhuman Primates", "date": "2026-10-15", "kind": "agent", "accountable": "CatalystNeuro", "agent": { "harness": "Claude Code (GitHub Actions)", "model": "claude-opus-5-5", "protocol": "pilot-7" }, "tier": "agent_verified", "entry": "002013-amjad-2025", "entry_title": "Microinvasive probes for dopamine, LFP and spikes in monkeys (Amjad et al. 2025)", "description": "Tests the findings of Amjad, Mahajan, Choi, Shrivastav, Murray, Somich, Coyne and Schwerdt (2025) on dandiset 002013, which holds the three example sessions of the paper. It checks the recording parameters and task structure, the dopamine response to large and small reward cues at site c8dg, beta-band LFP suppression by reward and cue side at c3bs-c3a, beta bursts around the cue, and the isolation of the sorted unit of monkey P, and compares the dandiset with the paper's Zenodo deposit.", "notebooks": [ { "path": "reproduce_amjad_2025.ipynb", "title": "Reproducing Amjad et al. (2025) from dandiset 002013", "executed": "notebooks/002013-amjad-2025/reproduce_amjad_2025.html" } ], "outcomes": [ { "claim": "insertion-yield", "statement": "Insertion yields were low, 21.4% (3 of 14 sensors) in monkey T and 33% (3 of 9) in monkey P, counted as recording sites that stayed functional after insertion until the end of the study.", "source": "Results and Discussion, second paragraph", "outcome": "blocked", "blocked_by": "data_missing", "notebook": null, "notes": "The dandiset holds one or two sites per session and no list of the implanted sensors or of which kept working, in its files or its metadata. None of the ten file names of the paper's Zenodo deposit, printed in the data check, points to such a list.", "deviations": [], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "chronic-duration", "statement": "The s-\u00b5IPs kept recording for more than a year after implantation; the example sites were recorded 483 days (c8dg), 496 days (c3bs-c3a) and 390 days (monkey P) after implantation.", "source": "Figure 3 to 5 legends; Results and Discussion", "outcome": "blocked", "blocked_by": "metadata_insufficient", "notebook": null, "notes": "The subject, device and electrode fields of the files hold no implantation date, and none of the mentions of implantation in the files' general section or the dandiset metadata (all printed in the data check) comes with a date, so days post-implant cannot be computed. The two Monkey T sessions are 13 days apart, which matches the difference between the 483 and 496 days of the legends.", "deviations": [], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fscv-waveform", "statement": "FSCV applied a triangular waveform from -0.4 to 1.3 V at 400 V/s, every 100 ms (10 Hz).", "source": "Methods, FSCV and EPhys Recordings", "outcome": "partially_reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "The applied voltage runs from -0.347 to 1.340 V at 400.5 V/s, one scan every 0.1 s. Three of four values are within tolerance; the lowest potential is outside -0.4 +- 0.04 V. The dandiset's own description (-0.35 to 1.34 V) agrees with the data, so the Methods state the nominal waveform. The data of the applied voltage have conversion 1 and offset 0, so the values are read as stored.", "deviations": [ "Measured on session 09/13/2024 only, the only FSCV session in the dandiset." ], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "ephys-acquisition", "statement": "EPhys was sampled at 30 kHz in monkey P and 32 kHz in monkey T, and band-pass filtered from 0.1 to 7500 Hz.", "source": "Methods, FSCV and EPhys Recordings", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "Monkey P's raw timestamps have a median interval of 33.333 \u00b5s (30 kHz) and Monkey T's series a rate of 32000 Hz. The Neuralynx settings of both Monkey T channels give a low cut of 0.1 Hz and a high cut of 8000 Hz, inside the tolerance of 7500 Hz but not equal to it; the dandiset's description also says 0.1 to 7500 Hz (printed). Monkey P's file stores no filter settings; its notes, printed in full, and its electrode columns name no filter or cut-off.", "deviations": [], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "shape-task-structure", "statement": "In the shape task, forced trials made up 25% of all trials, and the cue-reward association was reversed after blocks of 50 to 300 trials.", "source": "Methods, Behavioral Tasks", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "Forced trials are 358 of 1468 (24.4%). The eight complete blocks of the cue-reward association are 71 to 273 trials long; the first and last blocks, cut by the file, are 61 and 194 trials.", "deviations": [ "Blocks are found from the image codes of successful forced trials, as choice trials carry no feedback code; a reversal is placed after the last forced trial of the earlier association, so block edges are uncertain by the gap between forced trials." ], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "zenodo-link", "statement": "The dandiset reproduces the processed data the paper deposited on Zenodo; the trial selections and the beta power of the paper's example trial computed from the dandiset agree with the paper's own files.", "source": "Data Availability Statement (doi:10.5281/zenodo.14826579)", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "Trials are matched by number. The decoded selections agree with the four Zenodo lists for 569 of 570, 602 of 602, 594 of 594 and 577 of 578 trials; the only difference is trial 1501, the last of the session, absent from the Zenodo large-reward and right lists. The Zenodo example \"trial35\" is the 35th trial of the large-reward list (trial 84). Beta power computed here correlates with its trace at r = 0.943 at a lag of 6 ms, with a median amplitude ratio of 0.943, so the dandiset's LFP, its trial times and the beta pipeline used here agree with the paper's files. The Zenodo example's raw LFP variable is a copy of its time variable, so the LFP trace itself could not be compared. The criterion tests only r and was written after r = 0.94 was known. The amplitudes differ (maximum 2184 \u00b5V\u00b2 here against 1560 \u00b5V\u00b2 in the Zenodo trace), so the paper's filter or envelope probably differs from the one used here.", "deviations": [ "The beta filter (4th-order Butterworth, forwards and backwards, squared Hilbert envelope) is a common choice; the paper's code is not public, and r = 0.943 suggests its filter differs somewhat." ], "disagreed": false, "data": "with_inputs", "where": null, "proposed": null }, { "claim": "fig3a-dual-channel", "statement": "Dopamine was measured from two neighbouring channels of one dual-channel s-\u00b5IP (sites c8dg and c8ds, 0.1 to 0.2 mm apart), and both showed dopamine increases after the value cue in a single forced large-reward trial.", "source": "Figure 3A; Results and Discussion", "outcome": "blocked", "blocked_by": "data_missing", "notebook": null, "notes": "The dandiset holds FSCV from site c8dg only. Site c8ds is not in it, and the Zenodo deposit holds processed FSCV of c8dg only; its one FSCV file is FSCV_trlists_c8dg_09132024_firsthalf.mat (file list printed in the data check). The c8dg half can be tested and holds; the mean of the re-referenced dopamine over 0 to 3 s is above zero in 56 of 85 large-reward forced trials (65.9%), and in 58 of 85 with a baseline of the mean over -1 to 0 s.", "deviations": [], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fig3a-redox-peaks", "statement": "The single-trial FSCV colour plot shows dopamine redox current, with changes near 0.6 V (oxidation) and -0.2 V (reduction) after the value cue.", "source": "Figure 3A", "outcome": "blocked", "blocked_by": "metadata_insufficient", "notebook": null, "notes": "The raw FSCV current is in the dandiset, but its timing could not be linked to the trials. Its scan times are an exact 0.1 s grid starting at the first trial start and running 24120 s, past the last trial; at 0.6 V it does not match the stored oxidation current of any of six rows tested at the stored times (|r| below 0.06) or anywhere in the record (best r 0.16 to 0.20 at offsets that differ from row to row); and it shows no large-against-small contrast at the stored value cue times (p = 0.46). The same search at 0.50 V, 0.70 V and -0.19 V gives best r 0.17 to 0.22 at offsets that differ from row to row. Computed at the stored times (not deciding), the background-subtracted voltammogram of 164 large-reward forced trials has its largest rising-sweep value at -0.24 V (at 0.6 V it is -0.033) and its most negative value at the switching potential, 1.34 V, with an SD of 0.042 over the scan; it does not meet the criterion. A colour plot of a forced large-reward trial therefore cannot be placed in the trial.", "deviations": [], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fig3b-beta-bursts", "statement": "Single-trial bipolar LFP from site c3bs-c3a shows beta bursts (one to a few cycles of high beta power) whose frequency changes around the cue and other task events.", "source": "Figure 3B; Results and Discussion", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "With the threshold at the 75th percentile (356 \u00b5V\u00b2) and bursts of at least 50 ms, the burst rate of 570 large-reward trials is 1.51 per s in the second before the value cue and 1.21 in the second after it (Wilcoxon p = 9.5e-10). The rate changes around the cue at the 50th and 90th percentiles too (p below 1e-40), rising at the 50th and falling at the 90th, so its direction depends on the threshold while the change does not. The rate dips just after the cue (0.51 per s in 0 to 0.25 s) and falls again from 6 s after it, after the feedback. On the smoothed power 95.1% of runs above the 75th percentile last at least 50 ms (median 206 ms), so the minimum length does little; on the unsmoothed envelope the rate also falls after the cue (3.07 to 2.11 per s, p = 6.8e-34). At the deciding threshold the fall in bursts coincides with the drop in beta power after the cue, so it is not separate evidence from the suppression of fig4b-beta-reward.", "deviations": [ "The paper shows a single trial and does not define a burst; the threshold, minimum length and test were chosen here before the analysis." ], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fig4a-dopamine-reward", "statement": "Trial-averaged dopamine at site c8dg is higher after large-reward than after small-reward value cues, in successful forced trials of the shape task.", "source": "Figure 4A", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "Over 0 to 3 s after the value cue, the re-referenced dopamine has a median of 1.28 in 85 large-reward trials and 0.19 in 80 small-reward trials (Mann-Whitney p = 0.023; 3 trials of 168 have no valid samples in the window). The difference lies in 1.5 to 3 s (p = 0.0072), not 0 to 1.5 s (p = 0.41). The effect is weak. Before the cue the groups differ the other way (p = 0.064). Without re-referencing (the stored da) there is no difference (p = 0.3), nor in the re-referenced oxidation current (p = 0.41). It holds in trials 1 to 367 (p = 0.013) and not in trials 368 to 734 (p = 0.51). With a baseline of the mean over -1 to 0 s in place of the sample at the cue the result is the same (1.38 against 0.10, p = 0.022).", "deviations": [ "Only the first 734 of 1468 trials of the session have trial-aligned FSCV in the dandiset, so 168 successful forced trials are used.", "The stored da is not referenced at the value cue; it is re-referenced here by subtracting its value at the cue, which equals background subtraction there only if the PCA projection was linear and the nulling (NaN) masks are ignored.", "Small reward is inferred from the absence of the big-reward code, confirmed by the cue condition codes 35 and 36 in all 303 successful forced trials.", "The paper gives no window or test; 0 to 3 s and the Mann-Whitney test were chosen before the analysis." ], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fig4b-beta-reward", "statement": "Beta-band LFP power at c3bs-c3a in the direction task is more suppressed (stronger event-related desynchronization) after large-reward than after small-reward value cues, in a brief window after the cue.", "source": "Figure 4B", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "Over 0 to 0.5 s after the value cue, median beta power is 275 \u00b5V\u00b2 in 570 large-reward trials and 327 \u00b5V\u00b2 in 602 small-reward trials (ratio 0.84, Mann-Whitney p = 1e-08), with no difference before the cue (ratio 1.006, p = 0.82). It holds relative to pre-cue power (ratio 0.82, p = 1.1e-06) and on each side (left p = 0.0016, right p = 1e-07). It is confined to a brief window, as the legend says; over 0 to 1 s large is slightly higher (ratio 1.04, p = 0.09) and over 0.25 to 0.75 s there is no difference (p = 0.66). The 0.25 s smoothing does not carry the difference from before the cue; on the unsmoothed envelope it holds over 0 to 0.5 s (ratio 0.834, p = 1.9e-08) and 0.125 to 0.5 s (ratio 0.671, p = 4.8e-19); in 0.1 s bins large and small do not differ before the cue (ratios 0.95 to 1.05), large is lower from 0.1 to 0.4 s (0.85, 0.59, 0.60) and higher from 0.5 s on (1.32 to 1.49). The criterion, lower power over 0 to 0.5 s, is met; the bins fit a deeper suppression after large-reward cues and equally an earlier suppression with an earlier rebound, so the part of the claim saying the suppression is greater is not shown. The LFP's 1497 joins do not step (largest change at a join 20.3 \u00b5V, below the 99.9th percentile of 26.0 \u00b5V over all samples), and 118 of 1172 successful trials have one within -0.5 to 1 s of the cue, 23 to 36 in each reward and side condition. Trial 1501, absent from the Zenodo lists, is included; one trial of 570 cannot change the result.", "deviations": [ "The paper does not give the window highlighted in Figure 4B or a test; 0 to 0.5 s and the Mann-Whitney test were chosen before the analysis." ], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fig4c-beta-side", "statement": "With the reward fixed at large, beta-band LFP power is more suppressed when the cue is on the left (contralateral) than on the right (ipsilateral).", "source": "Figure 4C", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "Among large-reward trials, median beta power over 0 to 0.5 s is 245 \u00b5V\u00b2 for left cues and 301 \u00b5V\u00b2 for right cues (282 and 288 trials, ratio 0.81, Mann-Whitney p = 1.6e-06). Before the cue left is slightly higher (ratio 1.095, p = 0.013), so relative to pre-cue power the difference is larger (ratio 0.71, p = 2.1e-09). It holds over 0 to 1 s (p = 0.019), not over 0.25 to 0.75 s (p = 0.75), and also in small-reward trials (p = 1.5e-10). On the unsmoothed envelope left is lower over 0 to 0.5 s (ratio 0.807, p = 1.7e-06) and 0.125 to 0.5 s (ratio 0.688, p = 3.2e-17), most from 0.1 to 0.3 s (ratios 0.54 and 0.55).", "deviations": [ "Left is taken as contralateral, as the Methods state; the files do not say which hemisphere was implanted.", "Window and test chosen before the analysis, as for fig4b-beta-reward." ], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fig5a-spike-reward", "statement": "Spikes recorded from an s-\u00b5IP in the caudate of monkey P (site c33, 390 days post implant) fire at a different rate for large and small reward during the value cue fixation of the direction task.", "source": "Figure 5A; Results and Discussion", "outcome": "blocked", "blocked_by": "data_missing", "notebook": null, "notes": "Monkey P's file has no trials table and no events, so value cue times and reward sizes are not in the dandiset. The paper's Zenodo deposit has the session's raw Neuralynx events (session126_events.mat) as TTL values without names; their codes differ from those of the Monkey T sessions, so they could not be decoded (metadata insufficient). The notebook now reads that file, with 20812 events with 36 distinct TTL values and no names; the trial-start code 128 does not occur and the value cue code 23 occurs 20 times.", "deviations": [], "disagreed": false, "data": "dandiset", "where": null, "proposed": null }, { "claim": "fig5bc-unit-isolation", "statement": "The s-\u00b5IP recorded a spike-sorted unit in monkey P, shown by its average waveform and its interspike interval histogram.", "source": "Figure 5B and 5C", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_amjad_2025.ipynb", "notes": "Plexon unit 1 has 31247 spikes at 2.61 Hz. One of 31246 interspike intervals is under 1 ms (0.003%), and its mean waveform trough (-30.6 \u00b5V) is 5.66 times the raw noise level (5.41 \u00b5V, 300 to 6000 Hz, 600 s). The units table's unit 0 (unsorted and poor-quality spikes merged) has 11.65% of intervals under 1 ms and is not a unit.", "deviations": [ "The noise level and the two thresholds are common choices; the paper shows the waveform and histogram without numbers.", "The paper names the site c33; the file names it cl3. The session is taken as the paper's." ], "disagreed": false, "data": "dandiset", "where": null, "proposed": null } ], "findings": [ { "kind": "format_error", "statement": "Each event is stored about five times in the events tables of both Monkey T sessions (85665 stored times, 17686 distinct, in 09/13/2024; 59702 and 12434 in 09/26/2024), once for every overlapping 60 s trial window that contains it." }, { "kind": "metadata_insufficient", "statement": "The trials tables hold only times, no trial conditions. Conditions must be decoded from the event codes. Session 09/13/2024 has no small-reward code, and the labels of codes 35 and 36 (\"right cue reward condition 1/2\") do not say that they code large and small reward, which the data show in all 303 successful forced trials." }, { "kind": "data_missing", "statement": "The trial-aligned FSCV table of session 09/13/2024 has rows for only the first 734 of its 1468 trials (from a source file named \"firsthalf\"), and the dandiset has FSCV from c8dg only; the second site of Figure 3A (c8ds) is missing." }, { "kind": "metadata_insufficient", "statement": "The trial-aligned FSCV columns (da, ph, m, iox) carry no units, and da is not referenced at the value cue as the Methods describe; the reference point is not stated." }, { "kind": "metadata_error", "statement": "The raw FSCV current is labelled amperes, but its values range from -1527 to 1557; the unit is likely nA or the data are not current." }, { "kind": "metadata_error", "statement": "The raw FSCV scan times are an exact 0.1 s grid starting at the first trial start and running 24120 s, past the last trial, and the raw current cannot be matched to the trial-aligned oxidation current at any offset or to the task at the stored times. The conversion notes say the times were interpolated between trial-start codes; the data do not support that they are on the trial clock." }, { "kind": "metadata_error", "statement": "The bipolar LFP of session 09/26/2024 stores values already in volts (around 1e-4) with a conversion factor of 3.05e-8, so that read as NWB prescribes it is in picovolts." }, { "kind": "metadata_error", "statement": "The raw ephys of session 09/26/2024 has starting_time 0 and no timestamps, but it matches the stored LFP (r 0.92 to 0.98) only when read 9831.35 s later on the trial clock, consistently from the first trial to the last." }, { "kind": "metadata_error", "statement": "In Monkey P's file the spike times (6800 to 18785 s) are 6800.666 s later than the raw signal's timestamps (0 to 11985 s); the spike-triggered average of the raw signal shows the spike only with that shift. The file does not say the two use different clocks." }, { "kind": "format_error", "statement": "In Monkey P's units table, unit 0 merges the unsorted (0) and poor-quality (-1) thresholded spikes, including 7 spike times below 1 s and 3 repeated times; the poor-quality spikes of the thresholded table are not in time order." }, { "kind": "data_missing", "statement": "Monkey P's file has no trials or events, so the reward-related firing of Figure 5A cannot be tested. The paper's Zenodo events for this session are raw TTL values without a code map." }, { "kind": "metadata_insufficient", "statement": "No implantation dates are given, so the days post-implant of each session cannot be checked." }, { "kind": "metadata_error", "statement": "Monkey P's site is named cl3 in the file and dandiset description, c33 in the Figure 5 legend; the files' related_publications cite the bioRxiv preprint, not the article; the acquisition high cut is 8000 Hz in the Neuralynx settings, not the 7500 Hz the description states." }, { "kind": "other", "statement": "The applied FSCV waveform runs from -0.347 to 1.340 V, as the dandiset description says; the paper's Methods give -0.4 to 1.3 V." } ], "verifications": [ { "date": "2026-10-15", "kind": "agent", "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebook)", "reviewed": "The first executed notebook with the conclusion cells and the summary removed, the helper module and the claims file, with the paper's full text (PMC12183687). The verifier read the archive's file list and metadata through the DANDI API and tested the helpers on synthetic inputs (beta_power, burst_onsets, window_mean, codes_per_trial, rereference, fscv_window_mean, best_lag, reward_blocks). It had no attempt record and took the claims from the claims file.", "agreed": 14, "of": 14, "disagreements": [] }, { "date": "2026-10-15", "kind": "agent", "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebook)", "reviewed": "The full executed notebook of the first revision, the helper module, the claims file, the attempt record, the README and its own round-1 report, with the paper's full text (PMC12183687).", "agreed": 14, "of": 14, "disagreements": [] } ] } ], "entries": [ { "name": "002013-amjad-2025", "type": "reproduction", "title": "Microinvasive probes for dopamine, LFP and spikes in monkeys (Amjad et al. 2025)", "description": "Tests the findings of Amjad, Mahajan, Choi, Shrivastav, Murray, Somich, Coyne and Schwerdt (2025) on dandiset 002013, which holds the three example sessions of the paper. It checks the recording parameters and task structure, the dopamine response to large and small reward cues at site c8dg, beta-band LFP suppression by reward and cue side at c3bs-c3a, beta bursts around the cue, and the isolation of the sorted unit of monkey P, and compares the dandiset with the paper's Zenodo deposit.", "dandisets": [ "002013" ], "papers": [ "10.1021/acschemneuro.5c00071" ], "derived_from": [], "notebooks": [ { "path": "reproduce_amjad_2025.ipynb", "title": "Reproducing Amjad et al. (2025) from dandiset 002013", "executed": "notebooks/002013-amjad-2025/reproduce_amjad_2025.html" } ] } ] }