Dandiset 001712

Reproducing Davatolhagh et al. (2025) from dandiset 001712 7

Notebook reproduce_davatolhagh_2025.ipynb

• ✕Not reproducedFigure 1E

  Reaction time (time to first movement) is longer at lower contrasts in both genotypes.

  Over 750 trials with a movement onset after the stimulus, the Spearman correlation of absolute contrast with reaction time is -0.053 (p = 0.15); the median at 0% (0.089 s) exceeds that at 100% (0.081 s), the second part of the criterion. Medians are 0.081 to 0.095 s at every contrast, and 161 trials have no movement onset after the stimulus (149 with an onset at or before it and 12 with no onset). With reaction times above 0.15 s only (226 trials) the correlation is -0.226 (p = 0.00062), so the relation appears among slower responses. Of the 899 onsets, 149 fall at or before the stimulus and 346 within 0.08 s after it, so many may mark movements begun before the stimulus could guide them. The response time (choice registration minus stimulus onset) is shorter at higher contrast over all 911 choice trials (rho -0.134, p = 5.2e-05; medians 0.326 s at 0% and 0.281 s at 100%), a variant that does not decide the outcome. The paper's text calls reaction time the time to first movement and does not say how that movement is detected or how earlier movements are treated.

  How the test differed from the paper’s method

  • One session of one mouse.
  • Trials with a movement onset at or before the stimulus onset are left out; the paper does not say how it treats them.
• ✕Not reproducedFigure 3B; Results, third section

  Visual cortex (VISp) responds robustly to contralateral visual stimuli.

  Under the criterion, the right-minus-left and left-minus-right peaks both fall at the left edge of the frame (columns 0 and 2, rows 0 and 539), on the same side of the estimated midline (column 316), with trial tests at p = 0.053 and 0.14. These corner pixels are darker in the mean image (534 and 926) than 95% of the frame (5th percentile 1538), outside the imaged brain; the spatial components cover the whole frame with no mask. In a variant restricted to bright pixels 20 or more from the edge, the right-minus-left peak lies left of the midline (column 199) and the left-minus-right peak right of it (column 481), p = 8.1e-15 and 1.6e-17; the variant does not decide the outcome. It does not show that the lateralization is visual. At these two pixels the side difference exceeds 3 standard errors from 0.128 s after the stimulus, while the median movement onset is 0.086 s. Only 27 trials (13 right, 14 left) have no movement before 0.4 s; on them the difference at the two pixels points the same way (p = 0.02 and 0.091), but a masked peak search on these trials alone puts the peaks on the reverse sides (columns 550 and 72, p = 0.0062 and 0.012), which is not explained here. The slow trials exclude only wheel movement. In the same window paw speed and camera motion energy rise on 0.90 to 0.96 of all trials at 25 or 100% contrast, and wheel velocity, the right paw in both side cameras and the left camera's motion energy differ between right and left stimulus trials (p = 0.0005 or less); on the 27 slow trials paw speed rises on 0.44 to 0.78 of trials and the right paw in the right camera still differs by side (p = 0.014). Neither VISp nor the hemispheres can be named without an atlas registration.

  How the test differed from the paper’s method

  • Tested in image coordinates, with a midline estimated from the mean image, as the dandiset has no atlas registration.
  • The peak search of the criterion did not exclude pixels outside the brain; the masked search is a variant.
• ◐Partially reproducedFigure 4B and 4C; Results, fourth section

  The full linear encoding model explains more than half of single-trial cortical variance (cvR2 0.60 +/- 0.02 in KO and 0.54 +/- 0.02 in WT).

  Full-model cvR2 is 0.471 pooled in pixel space (reading a), 0.422 as the mean over pixels (b) and 0.126 as the mean over components (c); minus the intercept-only model (the time-point columns alone), 0.123, 0.133 and 0.078. None is within 0.486 to 0.660; reading (a) is above 0.3, which the criterion counts as partial. Without the body camera's video components the pooled cvR2 is 0.414. Variants that do not decide the outcome are as follows. With the video components of the left and right cameras added, as in the paper (encoding_all_cameras.ipynb), cvR2 is 0.574 (a) and 0.525 (b), within the range, and 0.208 (c). With folds that keep each trial's frames together and each frame once per trial, it is 0.422, 0.363 and 0.047 with the body camera only, and 0.472, 0.405 and 0.046 with all three cameras, below the range. Over the masked pixels of the fig3b variant (pixels brighter than the median of the mean image and 20 or more from the edge) reading (b) is 0.434. The paper's Results give 0.60 and 0.54 as the full model's cvR2 and say it accounts for more than half of the variance, while its Figure 4C legend says the full model is shown with the trial time regressor subtracted out. Under that reading the values here are 0.123 and 0.133 (a, b) with the body camera and 0.226 and 0.236 with all three cameras, far below the range.

  How the test differed from the paper’s method

  • One session of one mouse of unknown genotype; the expected range spans the WT and KO means.
  • No video components from the left and right cameras in the design that decides the outcome (left out for runtime; the body camera's are included; a second notebook adds them as a variant), paw speeds from Lightning Pose where the paper uses DeepLabCut, ROI motion energies assumed to be the whisker pads and body.
  • The subjective prior is an action-kernel model fitted here; trial windows, the penalty selection (generalized cross-validation) and the readings of cvR2 were chosen here, as the paper's widefield model code is not public.
• ◐Partially reproducedFigure 4C and 4D; Results, fourth section

  Movement predictors account for most of the explained variance; shuffling them reduces cvR2 by much more than shuffling task predictors (delta R2 for movement 0.40 +/- 0.02 in KO and 0.33 +/- 0.02 in WT; task shuffle beta -0.02).

  Shuffling movement predictors lowers cvR2 by 0.104 (a), 0.119 (b) and 0.067 (c); shuffling task predictors by 0.017, 0.015 and 0.013. Movement dominates under every reading and in every fold (smallest movement drop 0.065, largest task drop 0.019), but the movement drop is below 0.297 to 0.440 under every reading. Without the body camera's video components the movement drop is 0.042 against 0.021 for task pooled, and 0.010 against 0.015 over components. Variants that do not decide the outcome are as follows. With folds by trial and each frame once per trial, the movement drop is 0.049, 0.060 and -0.005 against 0.017, 0.016 and 0.012 for task, so over components task exceeds movement, which the criterion would count as not reproduced. With all three cameras the movement drop is 0.216, 0.232 and 0.161 against 0.008, 0.008 and 0.007 for task, still below the range; with folds by trial, 0.109, 0.115 and 0.008 against 0.008, 0.008 and 0.000.

  How the test differed from the paper’s method

  • As for fig4c-full-cvr2; the missing side-camera video components likely lower the movement contribution.
• ✓ReproducedFigure 1C; Results, first section

  Mice of both genotypes show sigmoidal psychometric curves, with the proportion of rightward choices increasing as the stimulus contrast increases on the right.

  P(right) rises at every step of signed contrast, from 0.112 at -100% to 1.000 at +100% (Spearman rho 1.000 over nine levels, span 0.887). The lapse-and-erf fit gives sigma 17.04% contrast (bootstrap 95% interval 13.98 to 20.47), bias -6.77% and left lapse 0.116. The claim covers both genotypes; this is one mouse of unknown genotype.

  How the test differed from the paper’s method

  • One session of one mouse, where the paper pools 10 sessions from each of 19 mice.
  • The criterion (Spearman rho and span) was set here; the paper shows the curves without a test.
• ✓ReproducedFigure 2B and 2C; Results, second section

  Mice use the block prior; rightward choices are more frequent in right-likely blocks than in left-likely blocks, most clearly when contrast is low, shifting the psychometric curve toward the likely side (shown for WT, and present but smaller in KO).

  At 0% contrast P(right) is 0.831 (74 of 89) in right-likely blocks and 0.507 (37 of 73) in left-likely blocks, difference 0.325, one-sided Fisher p = 9.2e-06. The difference is 0.182 to 0.458 at every signed contrast except +100% (0.000), and the psychometric bias moves by 19.41% contrast between blocks. Whether the shift is smaller in KO than in WT is not testable here.

  How the test differed from the paper’s method

  • One session of one mouse; the test (Fisher exact at 0% contrast) was chosen here.
• ✓ReproducedFigure 2E to 2G

  After a block switch, mice adapt their choices toward the newly likely side, and the proportion of choices toward the block side rises from an initial value A to an asymptote B with a time constant tau.

  Over 14 switches between biased blocks, choices toward the new block side are 0.543 in trials 0 to 4 and 0.827 in trials 20 to 39 (one-sided Fisher p = 2.7e-06). The fit gives A = 0.480, B = 1.000 (at its bound) and tau = 25.12 trials (SE 16.14), so tau is poorly determined. Resampling the switches gives 95% intervals of 0.279 to 0.643 for A, 0.836 to 1.000 for B and 6.41 to 38.37 trials for tau; B exceeds A in every draw and tau is 40 or more in 0.018 of them. A maximum-likelihood fit on single trials gives A = 0.479, B = 1.000, tau = 25.03. At contrasts of 12.5% or less, 0.452 against 0.820, tau 21.57.

  How the test differed from the paper’s method

  • The paper's parametric curve is not given; B - (B - A) exp(-t / tau) over trials 0 to 39, all contrasts, was chosen here.