Static report generated 2026-10-25 from urea_calibration_546-10.py (marimo notebook; code not shown).

Urea Calibration (546-10 filter) — 2026-10-24, 3 mM HEPES

Dataset: 2026-10-24_full_run_16bit/urea_samples/ — one BSA-passivated stepped-height channel (10 sections, 330 → 30 um nominal), urea + urease pre-mixed off-chip, imaged through the 546-10 bandpass filter in the randomized order from Urea Assay Prep - Printable.pdf (5 frames per sample). ROIs come from channel_rois.json (see channel_roi.py). What changed vs. 2026-10-02: the working solution is 3 mM HEPES + 0.08 % phenol red (was 10 mM). With less buffer, the same amount of ammonia from urea hydrolysis should move the pH, and so the phenol red absorbance, further. 0mM was imaged first this time, right after the light/dark frames. Scope: urea data only. The Allura Red frames are not used here, so no relative channel heights enter the analysis. The sigmoid shape parameters (c50, h) do not depend on path length, so each section is fit on its own. Pipeline
  1. Average the 5 frames per condition; T = (I - dark) / (light - dark) with the averaged 546-10 light/dark frames.
  2. Background (illumination) correction for every frame: b = median T in the clear substrate strips above and below the channel (first/last 40 crop rows); A = -log10(<T>_ROI / b).
  3. QA — blanks, replicate scatter vs. frame noise, path-length linearity.
  4. Per-section 4-parameter sigmoid fits a(c) = a_min + (a_max - a_min) / (1 + (c / c50)^-h).
  5. Residuals vs. time (time since mixing = imaging time + a constant).
  6. How well is h constrained? Profile over h, fixed-a_max fits, and a joint fit over the thinner sections.
  7. Comparison with the 10/02 run (10 mM HEPES, same filter) and earlier runs.
Concentrations: process.txt records 9.5 mg urea weighed for the top stock (prep sheet: 9.0 mg; confirmed for this run), so the top stock is ≈31.7 mM, not 30 mM, and every sample is ×1.056 its label. The toggle below switches between corrected and label concentrations. Concentrations refer to the urea sample before 1:1 mixing with urease.

☑ Use mass-corrected urea concentrations (x 1.0556: 9.5 mg weighed vs 9.0 mg nominal)

↑ topStep 1 — Background (illumination) correction

b is the median transmittance of the clear substrate above and below the channel. It should be 1 when the illumination matches the light frame. Error bars span the 5 individual frames of each burst. 0mM was imaged first, ≈1 min after the light frame.
Figure 1
conditiontimebb range in burstdA_bg = -log10 b
0mM14:33:241.00130.00047-0.0006
11bmM14:35:240.9990.000230.0004
2mM14:37:020.99920.000260.0003
18mM14:41:341.00020.00037-0.0001
5amM14:43:231.00160.00013-0.0007
23cmM14:45:451.00120.00028-0.0005
9mM14:47:231.00185e-05-0.0008
blank114:48:011.00210.00021-0.0009
5bmM14:48:531.00190.00017-0.0008
1mM14:50:041.00120.00016-0.0005
11amM14:51:261.00170.00028-0.0007
23amM14:53:121.00210.00012-0.0009
14mM14:55:151.00210.00016-0.0009
5cmM14:56:391.00160.00016-0.0007
blank214:58:011.00290.00011-0.0012
11cmM14:59:031.00270.00016-0.0012
3mM15:00:131.00240.00031-0.001
23bmM15:01:281.00240.00023-0.001
30mM15:02:471.00270.00017-0.0012
7mM15:04:221.00328e-05-0.0014
blank315:05:051.00359e-05-0.0015

↑ topStep 2 — Section absorbance

A = -log10(<T>_ROI / b) from the 5-frame average, in acquisition order. Concentrations in the plot follow the toggle above. Lines connect replicate means.
Figure 2
conditiontimec (mM)330 um250 um200 um150 um120 um90 um70 um50 um40 um30 um
0mM14:33:2400.1330.10280.0810.0620.04960.03690.0290.02050.01570.0115
11bmM14:35:2411.6110.38540.31250.25580.20210.16580.12530.10090.07180.05550.0406
2mM14:37:022.1110.15490.12040.09550.0730.05880.0440.03450.02440.0190.0137
18mM14:41:34190.32060.25660.20830.16340.13310.10070.08080.05760.04450.0327
5amM14:43:235.2780.16780.13130.1040.080.06510.04890.03970.02940.02310.018
23cmM14:45:4524.2780.37990.30890.25210.19970.16430.12420.10120.07270.05570.042
9mM14:47:239.50.22140.17530.13970.10790.08820.06610.05410.03870.02940.0225
blank114:48:010.00130.00070.00070.00080.00060.00060.00220.00220.0010.0021
5bmM14:48:535.2780.18430.14430.11450.08840.07180.05390.04440.03180.02370.0184
1mM14:50:041.0560.13860.1090.08670.06660.05370.04020.03370.02430.01790.0142
11amM14:51:2611.6110.25850.20610.16590.12860.1050.07920.06480.04640.03460.0266
23amM14:53:1224.2780.36730.29750.24240.19160.15780.11860.09690.06940.05290.0399
14mM14:55:1514.7780.27220.21950.17740.13820.11280.08520.070.05040.0380.0287
5cmM14:56:395.2780.14320.11190.08870.06820.05510.0410.03460.02530.01840.0146
blank214:58:010.00080.00040.00050.00060.00090.00040.00260.00250.00110.0021
11cmM14:59:0311.6110.22960.18450.14810.11480.09410.07060.05840.04190.03170.024
3mM15:00:133.1670.16320.12740.10090.07770.0630.04690.03910.02830.0210.0165
23bmM15:01:2824.2780.3330.26770.21720.17070.13970.10550.08630.06210.04720.0355
30mM15:02:4731.6670.40050.32680.26830.2130.17610.1330.10850.0780.05990.045
7mM15:04:227.3890.2050.16110.12790.09860.08040.06040.05010.03590.02710.0207
blank315:05:050.00060.00040.00020.00060.00070.00050.00260.00240.00150.0021

↑ topStep 3 — QA

  • Blanks should read A ≈ 0 in every section (the light frame was taken with blank buffer).
  • Replicate scatter vs. frame noise: the SD of the independently prepared a/b/c samples (5, 11, 23 mM) compared with the SD of A across the 5 frames of one burst. A large ratio means preparation (or timing) variation dominates, not camera noise.
  • Path-length linearity: if A were exactly proportional to channel height, A_i / A_90um would be flat in concentration for each section.

Blanks

blanktime330 um250 um200 um150 um120 um90 um70 um50 um40 um30 um
blank114:48:010.00130.00070.00070.00080.00060.00060.00220.00220.0010.0021
blank214:58:010.00080.00040.00050.00060.00090.00040.00260.00250.00110.0021
blank315:05:050.00060.00040.00020.00060.00070.00050.00260.00240.00150.0021

Replicate scatter vs. 5-frame noise

label (mM)quantity330 um250 um200 um150 um120 um90 um70 um50 um40 um30 um
5replicate SD0.02070.01630.0130.01010.00840.00650.00490.00330.00290.0021
5frame SD (median)0.00030.00030.00040.00020.00030.00030.00030.00030.00030.0002
5ratio6551.73640.827.525.216.710.811.112.1
5replicate CV %12.512.612.712.813.113.612.411.413.412.3
11replicate SD0.08290.06850.05770.04690.03860.02940.02290.01610.0130.0089
11frame SD (median)0.00020.00020.00030.00020.00040.00020.00030.00040.00020.0001
11ratio477.9282171.5232.1107.4132.888.745.356.1164.3
11replicate CV %28.529.230.431.631.832.130.730.232.129.4
23replicate SD0.02430.02130.0180.0150.01280.00960.00770.00540.00430.0033
23frame SD (median)0.00020.00030.00030.00030.00030.00030.00030.00040.00030.0001
23ratio102.664.869.351.940.531.424.61413.424.1
23replicate CV %6.87.37.688.38.38.188.38.5

Path-length linearity

Figure 3

↑ topStep 4 — Per-section sigmoid fits

Each section is fit independently to all 21 sample points (replicates kept separate, 0 mM included, unweighted least squares). Standard errors come from the covariance matrix scaled by the residual variance. corr(c50,h) and corr(a_max,h) show how strongly the parameters trade off: values near ±1 mean the data cannot pin them down separately, which is expected when the upper plateau is not reached.
sectiona_mina_maxc50hSE a_minSE a_maxSE c50SE hcorr(c50,h)corr(a_max,h)RMSEat bound
330 um0.1370.417312.40492.13460.02360.07823.90291.1023-0.725-0.8850.03329
250 um0.10620.339912.49612.1390.01960.06583.95851.1079-0.73-0.8860.0277
200 um0.08390.279312.6872.13530.01650.05734.17131.1282-0.741-0.8880.02333
150 um0.06390.225213.1682.07530.01350.05224.83271.1282-0.773-0.8960.01896
120 um0.05110.189213.61261.98690.01140.04915.61051.0975-0.8-0.9030.01573
90 um0.03810.142413.50232.0030.00870.03635.42881.1017-0.794-0.9020.01194
70 um0.03050.120214.2041.84910.00710.0356.72491.0106-0.831-0.9110.0094
50 um0.02160.089615.13251.73090.00520.03018.46570.9546-0.86-0.9190.0066
40 um0.01640.068915.3191.740.00410.02448.95170.9968-0.863-0.920.00527
30 um0.01210.058318.85471.4720.00310.0317.11550.8645-0.912-0.9330.0037
Figure 4
Figure 5

↑ topStep 5 — Residuals vs. time

Assuming all samples were mixed with urease at about the same moment (as on 10/02; not recorded for this run), time since mixing is the imaging time plus a constant. If the reaction were still progressing, residuals would trend upward with time. Residuals are normalized by each section's measured span (A at the top concentration minus A at 0 mM) so sections can be pooled. The right panel uses only the a/b/c replicates: each replicate's deviation from its group mean, averaged over sections, which is model-free.
Figure 6

↑ topStep 6 — How well is h constrained?

Profile over h: for each fixed h, refit a_min, a_max, c50 and record the residual sum of squares. The 95 % interval is where RSS / RSS_min stays below 1 + F(0.95; 1, n-4) / (n-4). A flat profile means the data do not determine h. Fixed a_max: without an upper plateau, a_max is set mostly by extrapolation. Holding a_max = A_0 + k (A_top - A_0) for several k shows how much h depends on that assumption. Joint fit: shared c50 and h across the selected sections, with each section keeping its own a_min and a_max. Thinner sections have lower absorbance and the most nearly linear path-length response (Step 3).
Figure 7
sectionh (free fit)95% CI low95% CI highCI hits grid edge
330 um2.1350.64True
250 um2.1390.64True
200 um2.1350.64True
150 um2.0750.64True
120 um1.9870.64True
90 um2.0030.64True
70 um1.8490.64True
50 um1.7310.64True
40 um1.740.64True
30 um1.4720.64True
Figure 8
sectionh (k=1.1)c50 (k=1.1)h (k=1.25)c50 (k=1.25)h (k=1.5)c50 (k=1.5)h (k=2)c50 (k=2)h (k=3)c50 (k=3)h (k=5)c50 (k=5)
330 um2.0212.91.71151.4319.31.1829.71.01570.9131.3
250 um2.01131.7115.21.4319.51.19301.0157.40.9131
200 um2.0113.21.7115.51.4419.81.230.31.0357.50.92129.8
150 um2.0213.41.7215.71.46201.2230.51.0457.30.93127.6
120 um213.61.7115.81.4520.21.2130.81.0457.90.93128.8
90 um213.51.7115.81.4520.21.2130.91.04580.93129.3
70 um1.9513.51.6715.81.4220.31.1931.31.0259.60.91134.7
50 um1.9313.61.6615.91.4120.51.1831.61.0160.40.91137.1
40 um1.9513.71.6716.11.4220.61.231.61.0359.80.92134
30 um1.8913.81.6316.21.3920.91.1732.41.0162.10.9141.2

Sections in joint fit: 90um, 70um, 50um, 40um, 30um

fitsectionshSE hc50 (mM)SE c50corr(c50,h)RMSE
selected90 um, 70 um, 50 um, 40 um, 30 um1.8660.43714.212.84-0.8290.00795
all330 um, 250 um, 200 um, 150 um, 120 um, 90 um, 70 um, 50 um, 40 um, 30 um2.0930.32912.761.29-0.7510.01829

↑ topStep 7 — Comparison with 10/02 (10 mM HEPES) and earlier runs

The 10/02 run used the same 546-10 filter, sample series, and analysis, so it isolates the buffer change (different device, so heights may differ by a few %). Its section absorbances are read from ../2026-10-02_urea_and_allura_red/urea_calibration_546-10.json and get the same mass-correction choice as this run (toggle above). Earlier runs stored G = blank-subtracted absorbance referred to the 90 um section (using Allura-derived relative heights). The comparable quantity here is ΔA_90 = A_90(c) - A_90(0 mM). The 9/26 run used the 545-30 filter, so its absolute values are not comparable, but its curve shape is. The right panel normalizes every curve to its value at 20 mM (the highest concentration all runs share). 9/11 and 9/26 concentrations are as labeled.
Figure 9
runc50 (mM)hnote
10/24 (546-10, 3 mM), 90 um free fit13.52.003per-section fit of A (not dA)
10/24 (546-10, 3 mM), joint fit: selected14.211.86690 um, 70 um, 50 um, 40 um, 30 um
10/24 (546-10, 3 mM), joint fit: all12.762.093330 um, 250 um, 200 um, 150 um, 120 um, 90 um, 70 um, 50 um, 40 um, 30 um
10/02 (546-10, 10 mM), 90 um free fit10001.245as saved (mass-corrected conc.)
10/02 (546-10, 10 mM), joint fit: selected10001.25890 um, 70 um, 50 um, 40 um, 30 um
10/02 (546-10, 10 mM), joint fit: all624.041.166330 um, 250 um, 200 um, 150 um, 120 um, 90 um, 70 um, 50 um, 40 um, 30 um
9/11 Set 1 (546-10)8.063.694
9/11 Set 2 (546-10)229.831.029bounded fit (a_max<=1.0, c50<=500) - see divergence caveat above
9/26 (545-30)1251.181.116

3 mM vs. 10 mM HEPES, section by section

Mean ΔA = A(c) - A(0 mM) for every section, this run (solid) vs. 10/02 (dashed). Ratio panel: ΔA_10/24 / ΔA_10/02 at the concentrations both runs share; a ratio above 1 means the weaker buffer gives more color change per mM urea.
Figure 10
section1.06 mM2.11 mM3.17 mM5.28 mM7.39 mM9.5 mM11.6 mM14.8 mM19 mM24.3 mM31.7 mM
330 um0.811.9361.3050.7821.2651.2371.6181.0831.091.0360.924
250 um0.8971.5961.1930.7291.2041.181.5941.0631.0530.9930.874
200 um1.0091.5951.1450.6881.1121.0921.5061.0051.0110.9360.824
150 um1.1011.6111.1860.7051.1221.0911.510.9890.9970.9270.816
120 um1.2631.721.2490.7471.1581.1341.5531.0110.9960.9280.818
90 um1.2641.8071.1950.7351.1231.0891.4990.9830.9720.8910.773
70 um2.9972.1161.5450.9191.2991.1991.6161.0621.0080.9430.81
50 um3.2851.9031.7270.9861.2761.2151.5841.0560.9790.920.792
40 um5.1293.0851.5860.9511.1311.2021.5420.9890.9640.8760.755
30 um12.8084.842.0691.2531.3311.3111.6111.0540.970.9150.774

↑ topStep 8 — Sensitivity: refit without outlier samples

11b (imaged second, 2 min into the run) reads A ≈ 0.125 at 90 um, close to the 23 mM level and far above 11a (0.079) and 11c (0.071). This section repeats the fits of Steps 4–6 with the samples chosen below removed (default: 11bmM) and shows them next to the all-sample results. Everything above this section still uses all samples.

Samples to exclude: 11bmM

Excluded: 11bmM (17 of 18 samples kept)
sectionc50 allc50 exclh allh exclh 95% CI allh 95% CI excla_max alla_max exclRMSE/span % allRMSE/span % exclat bound (excl)
330 um12.420.962.1351.7080.60-4.000.88-2.790.4170.53212.44.9
250 um12.521.372.1391.7070.60-4.000.90-2.790.340.43912.44.8
200 um12.6921.822.1351.720.60-4.000.90-2.790.2790.36412.54.7
150 um13.1722.642.0751.7210.60-4.000.92-2.790.2250.29612.64.7
120 um13.6124.451.9871.6510.60-4.000.90-2.720.1890.25712.44.7
90 um13.524.192.0031.6560.60-4.000.88-2.720.1420.19312.44.7
70 um14.227.691.8491.5170.60-4.000.86-2.530.120.17211.84.6
50 um15.1331.411.7311.4340.60-4.000.84-2.360.090.13411.54.4
40 um15.3231.031.741.4720.60-4.000.86-2.470.0690.10211.94.5
30 um18.8549.981.4721.2480.60-4.000.80-2.250.0580.104114.6
Figure 11
Figure 12

Summary: all samples vs. excluded removed

quantityall samplesexcluded removed10/02 (10 mM)
joint fit (selected): c50 (mM)14.2127.081000
joint fit (selected): h1.8661.5471.258
joint fit (selected): RMSE0.007950.003050.00161
joint fit (all): c50 (mM)12.7622.05624.04
joint fit (all): h2.0931.6911.166
joint fit (all): RMSE0.018290.007110.00326
90 um free fit: c5013.50224.1861000
90 um free fit: h2.0031.6561.245
trend of median residual (%/10 min)-5.62-1.19
p0.0770.379
replicate-only trend (%/10 min)-16.92-8.62
p0.0020.013
Saved: urea_calibration_546-10.json