Validation / Papers / Gouwens 2019

Classification of electrophysiological and morphological neuron types in the mouse visual cortex

Electrophysiology · research paper · pyABF, h5py and SciPy (Python), through the patch-clamp adapter

How to read this page. In this validation, a script plays the scientist. It gives the answers that we wrote before the run, from the methods of the paper. A known value comes from the paper, from a tutorial or from a check that we ran. This page has no combined run of the paper yet.

Run of 9 October 2026, claude-haiku-5-5: 14 of 14 values computed, 14 of 14 correct in the final answer

The paper

Gouwens NW, Sorensen SA, Berg J, Lee C, Jarsky T, Ting J, et al. Classification of electrophysiological and morphological neuron types in the mouse visual cortex. Nature Neuroscience 22(7):1182-1195 (2019). doi:10.1038/s41593-019-0417-0

Related sources:

Data: Allen Cell Types Database, cell electrophysiology recordings (NWB files) and the feature table of each cell, api.brain-map.org. Allen Institute Terms of Use. link

What it measured

Whole-cell current-clamp recordings of mouse visual cortex neurons. The paper defines each electrophysiological feature (rheobase, F-I slope, input resistance, membrane time constant, sag, spike shape) and classifies the cells. The Cell Types Database publishes the value of each feature for each cell. The article prints distributions, not the value of one cell.

Data

Allen Cell Types Database NWB files of the cells 567987585, 595511209 and 563220277. fetch.sh downloads them from api.brain-map.org (16 to 21 MB each) and checks the SHA-256 of each one. Size: 55 MB.

License: Allen Institute Terms of Use (citation required). The benchmark downloads the files and does not copy them into the repository.

Data source

The instruction

A script sends this message as the scientist.

ScientistFor three cells, give the rheobase, the F-I slope, the input resistance, the membrane time constant and the upstroke to downstroke ratio of the first spike at rheobase, from the Long Square sweeps.

Basis: The Methods of the paper and the Allen Institute electrophysiology white paper: the feature definitions of the ipfx package. The database fields are ef__threshold_i_long_square, ef__f_i_curve_slope, ef__ri, ef__tau and ef__upstroke_downstroke_ratio_long_square.

The decisions

The model asks questions during a run. A script gives these answers to the questions of the model. We wrote the answers before the run.

**Table 1 |** Answers that a script gives to the questions of the model.
Decision	Value	Source
Definition of the spike threshold	`dvdt_fraction_of_upstroke`	The Allen Institute definition (ipfx spike_detector.refine_threshold_indexes, 5 percent of the mean upstroke).
Spike threshold cutoff for dV/dt	`20`	The ipfx default of dv_cutoff, 20 mV per ms.
Lowest spike peak	`-30`	The ipfx default of min_peak, -30 mV.
Baseline window before the step	`0.92,1.02`	The 0.1 s before the step, which starts at 1.02 s (the ipfx baseline interval).
Sweeps to include	`all`	The request limits the sweeps to the Long Square label.
Liquid junction potential correction	`0`	The database values are not corrected, so the benchmark does not correct.
Deflection that gives the input resistance	`peak`	ipfx subthresh_features.input_resistance uses the peak deflection.
Other questions of the agent	`Use the values in the decision record.`	Not in the paper. The benchmark answers each free question with this text.

Known values

The tolerance is the largest difference from the known value that we accept. We set it before the run. Exact: the number must be the same.

**Table 2 |** Known values for Gouwens 2019.
Value	Known value	Tolerance	Source
`c1_rheobase`Cell 567987585: rheobase, pA Source of the known value Printed in the paperWhere: Cell Types Database, cell 567987585, field ef__threshold_i_long_square. 230. The article does not print it.Check: The adapter gives 230.Note in the list of known values: published	230	exact	Printed in the paper
`c1_fi_slope`Cell 567987585: F-I slope, Hz per pA Source of the known value Printed in the paperWhere: Same record, field ef__f_i_curve_slope. 1.07166364262164.Check: The adapter gives 1.07166365.Note in the list of known values: published	1.0717	± 0.005	Printed in the paper
`c1_ri`Cell 567987585: input resistance, MOhm Source of the known value Printed in the paperWhere: Same record, field ef__ri. 129.843786358833.Check: The adapter gives 129.843763.Note in the list of known values: published	129.84	± 0.5	Printed in the paper
`c1_tau`Cell 567987585: membrane time constant, ms Source of the known value Printed in the paperWhere: Same record, field ef__tau. 8.25874032258183.Check: The adapter gives 8.223. The difference is 0.4 percent and its cause is not known.Note in the list of known values: published	8.26	± 0.3	Printed in the paper
`c1_ratio`Cell 567987585: upstroke to downstroke ratio Source of the known value Printed in the paperWhere: Same record, field ef__upstroke_downstroke_ratio_long_square. 1.14581526994923.Check: The adapter gives 1.14581527.Note in the list of known values: published	1.1458	± 0.003	Printed in the paper
`c2_rheobase`Cell 595511209: rheobase, pA Source of the known value Printed in the paperWhere: Cell 595511209, field ef__threshold_i_long_square. 70.Check: The adapter gives 70.Note in the list of known values: published	70	exact	Printed in the paper
`c2_fi_slope`Cell 595511209: F-I slope, Hz per pA Source of the known value Printed in the paperWhere: Same record, field ef__f_i_curve_slope. 0.115153631284916.Check: The adapter gives 0.115153630.Note in the list of known values: published	0.1152	± 0.001	Printed in the paper
`c2_ri`Cell 595511209: input resistance, MOhm Source of the known value Printed in the paperWhere: Same record, field ef__ri. 187.187567353249.Check: The adapter gives 187.187566.Note in the list of known values: published	187.19	± 0.5	Printed in the paper
`c2_tau`Cell 595511209: membrane time constant, ms Source of the known value Printed in the paperWhere: Same record, field ef__tau. 32.555055567042.Check: The adapter gives 32.259, which is 0.9 percent lower.Note in the list of known values: published	32.56	± 0.6	Printed in the paper
`c2_ratio`Cell 595511209: upstroke to downstroke ratio Source of the known value Printed in the paperWhere: Same record, field ef__upstroke_downstroke_ratio_long_square. 3.87456911022343.Check: The adapter gives 3.87456911.Note in the list of known values: published	3.8746	± 0.004	Printed in the paper
`c3_rheobase`Cell 563220277: rheobase, pA Source of the known value Printed in the paperWhere: Cell 563220277, field ef__threshold_i_long_square. 90.Check: The adapter gives 90.Note in the list of known values: published	90	exact	Printed in the paper
`c3_fi_slope`Cell 563220277: F-I slope, Hz per pA Source of the known value Printed in the paperWhere: Same record, field ef__f_i_curve_slope. 0.207532397408207.Check: The adapter gives 0.207532393.Note in the list of known values: published	0.2075	± 0.001	Printed in the paper
`c3_ri`Cell 563220277: input resistance, MOhm Source of the known value Printed in the paperWhere: Same record, field ef__ri. 128.515630960464.Check: The adapter gives 128.515623.Note in the list of known values: published	128.52	± 0.5	Printed in the paper
`c3_ratio`Cell 563220277: upstroke to downstroke ratio Source of the known value Printed in the paperWhere: Same record, field ef__upstroke_downstroke_ratio_long_square. 3.22588602314736.Check: The adapter gives 3.22588602.Note in the list of known values: published	3.2259	± 0.004	Printed in the paper
`c3_tau` (reference)Cell 563220277: membrane time constant, ms (the tool gives no fit for one of the three sweeps) Source of the known value Printed in the paperWhere: Same record, field ef__tau. 14.8685687379896.Check: The adapter gives 15.79 from 2 sweeps (the tool gives no fit for the third), so the item is a reference.Note in the list of known values: published	14.87	± 0.5	Printed in the paper

Latest scored run

Model: claude-haiku-5-5. Runs for each paper and model: 1. Blind mode: on. Status: answer. 250 s. Computed: 14 of 14 values. Reported: 14 of 14 values. The result file is bench/results/papers-2026-10-09-ephys-haiku.md. This run is not in the totals of the page of papers.

Computed: a logged number is within the tolerance. Reported: the final answer states the value, as the claim check measures. The table copies the cells of the result file.

**Table 3 |** Items of the run of claude-haiku-5-5.
Item	Expected	Computed	Reported
`c1_rheobase`Cell 567987585: rheobase, pA	230 exact	pass 230 (n6 metrics.rheobase_pa, entry 71)	pass 230 via tolerance (n14, claim check 144)
`c1_fi_slope`Cell 567987585: F-I slope, Hz per pA	1.0717 ±0.005	pass 1.071664 (n6 metrics.fi_slope_hz_per_pa, entry 71)	pass 1.072 via tolerance (n14, claim check 144)
`c1_ri`Cell 567987585: input resistance, MOhm	129.84 ±0.5	pass 129.8438 (n11 metrics.input_resistance_mohm, entry 95)	pass 129.8 via tolerance (n11, claim check 144)
`c1_tau`Cell 567987585: membrane time constant, ms	8.26 ±0.3	pass 8.223051 (n11 metrics.tau_ms, entry 95)	pass 8.223 via tolerance (n11, claim check 144)
`c1_ratio`Cell 567987585: upstroke to downstroke ratio	1.1458 ±0.003	pass 1.145815 (n6 metrics.upstroke_downstroke_ratio, entry 71)	pass 1.146 via tolerance (n5, claim check 144)
`c2_rheobase`Cell 595511209: rheobase, pA	70 exact	pass 70 (n6 table.rows[9][2], entry 71)	pass 70 via tolerance (n14, claim check 144)
`c2_fi_slope`Cell 595511209: F-I slope, Hz per pA	0.1152 ±0.001	pass 0.1151536 (n7 metrics.fi_slope_hz_per_pa, entry 75)	pass 0.115 via tolerance (n7, claim check 144)
`c2_ri`Cell 595511209: input resistance, MOhm	187.19 ±0.5	pass 187.1876 (n12 metrics.input_resistance_mohm, entry 99)	pass 187.2 via tolerance (n12, claim check 144)
`c2_tau`Cell 595511209: membrane time constant, ms	32.56 ±0.6	pass 32.25924 (n12 metrics.tau_ms, entry 99)	pass 32.26 via tolerance (n12, claim check 144)
`c2_ratio`Cell 595511209: upstroke to downstroke ratio	3.8746 ±0.004	pass 3.874569 (n7 metrics.upstroke_downstroke_ratio, entry 75)	pass 3.875 via tolerance (n7, claim check 144)
`c3_rheobase`Cell 563220277: rheobase, pA	90 exact	pass 90 (n6 table.rows[10][2], entry 71)	pass 90 via tolerance (n14, claim check 144)
`c3_fi_slope`Cell 563220277: F-I slope, Hz per pA	0.2075 ±0.001	pass 0.2075324 (n8 metrics.fi_slope_hz_per_pa, entry 79)	pass 0.208 via tolerance (n8, claim check 144)
`c3_ri`Cell 563220277: input resistance, MOhm	128.52 ±0.5	pass 128.5156 (n13 metrics.input_resistance_mohm, entry 103)	pass 128.5 via tolerance (n13, claim check 144)
`c3_ratio`Cell 563220277: upstroke to downstroke ratio	3.2259 ±0.004	pass 3.225886 (n8 metrics.upstroke_downstroke_ratio, entry 79)	pass 3.226 via tolerance (n8, claim check 144)
`c3_tau` (reference)Cell 563220277: membrane time constant, ms (the tool gives no fit for one of the three sweeps)	14.87 ±0.5	match 15 (n12 table.rows[0][0], entry 99)	match 15 via tolerance (n14, claim check 144)

Notes

Notes in the case file

The feature values come from the Cell Types Database, which is the data release of the paper. The adapter reproduces the rheobase, the F-I slope, the input resistance and the upstroke to downstroke ratio of each cell to 6 or more digits, because it follows the ipfx definitions. The membrane time constant differs by 0.4 to 6 percent. Cell 563220277 has no time constant fit for one sweep.