BCECF pH Sensor

David Garenne

BCECF Ratiometric pH Calibration

Estimate pH in myTXTL cell-free reactions from BCECF fluorescence.

BCECF is a ratiometric pH dye: its emission at 535 nm is excited at two wavelengths (490 nm, 440 nm). The ratio of those reads (490/440) tracks pH independently of absolute dye concentration.

This analysis uses the Nucleus CDK plate-reader pipeline:

Load the calibration plate (BCECF standards at known pH)
Compute the 490/440 ratio and fit a standard curve
Load the kinetic data plate
Standardize fluorescence ratio to pH via the curve
Fit kinetics with a sigmoid + drift + baseline-offset model

Install note. nucleus-cdk declares pyarrow>=18,<19, and pyarrow publishes no cp314 wheels, so a plain pip install fails on Python 3.14. Nothing in cdk imports pyarrow, so we install with --no-deps; the packages the CDK actually uses come from environment.yml.

!pip install --no-deps nucleus-cdk==0.6.0rc2

# Surface a failed install here, rather than as a confusing ModuleNotFoundError
# in the import cell below.
import importlib.metadata as md
assert md.version("nucleus-cdk") == "0.6.0rc2", f"got {md.version('nucleus-cdk')}"
print("nucleus-cdk", md.version("nucleus-cdk"))

Requirement already satisfied: nucleus-cdk==0.6.0rc2 in /opt/homebrew/anaconda3/lib/python3.12/site-packages (0.6.0rc2)
nucleus-cdk 0.6.0rc2

%load_ext autoreload
%autoreload 2

import pandas as pd

from cdk import logging
log = logging.setup_logging(logging.ERROR)

from cdk.instruments.platereader import load_platereader_data

The autoreload extension is already loaded. To reload it, use:
  %reload_ext autoreload

1. Load calibration data¶

Gen5-exported endpoint data paired with a platemap. The platemap specifies the pH of each standard in a pH column, and marks the standard wells with Name == "BCECF standard".

cal = load_platereader_data(
    data_file="2026_3_10 BCECF pH measurement Calibration.xlsx",
    platemap_file="2026_3_10 BCECF pH measurement Calibration Platemap.csv",
    platereader="biotek",
)
print(cal)

PlateReaderResult with the following 1 blocks:
0: (24, 35) endpoint read with reads: 440,535 (Fluorescence), 490,535 (Fluorescence) (Plate 'Plate 5')

                    INFO     Parsing plate block 1                 biotek.py:190

                    INFO     Read metadata section, found 1 read   biotek.py:276
                             block(s).

                    WARNING  Failed to parse the data grid:       biotek.py:1430
                             Duplicate names are not allowed.

                    WARNING  Parsed data grid did not contain any  biotek.py:334
                             data.

                    WARNING  Failed to parse the data grid:       biotek.py:1430
                             Duplicate names are not allowed.

                    WARNING  Parsed data grid did not contain any  biotek.py:334
                             data.

                    WARNING  Failed to parse the data grid:       biotek.py:1430
                             Duplicate names are not allowed.

                    WARNING  Parsed data grid did not contain any  biotek.py:334
                             data.

                    INFO     Parsed 1 segment(s) successfully.     biotek.py:406

                    INFO     Found 0 segment(s) that could not be  biotek.py:407
                             associated with existing read data.

PlateReaderResult with the following 1 blocks:
0: (24, 35) endpoint read with reads: 440,535 (Fluorescence), 490,535 (Fluorescence) (Plate 'Plate 5')

# Take the block we want for further analysis
calibration_data = cal[0]

2. Calibration curve¶

Compute the BCECF ratio (490,535 / 440,535) and fit a standard curve over the "BCECF standard" wells — a linear regression of pH against 490/440 ratio.

cal_ratio = calibration_data.ratio("490,535", "440,535")
curve = cal_ratio.fit_standard_curve(known="pH", standard_name="BCECF standard")

curve.plot()
print("slope, intercept =", curve.result.params)
print("R^2 =", round(curve.result.r_squared, 4))

slope, intercept = [  2.67495596 -13.79554277]
R^2 = 0.9112

3. Load kinetic data¶

The 24 h kinetic run. The export contains the 440 and 490 kinetic blocks we want plus a summary block we don’t, so we select just the kinetic blocks.

sample_result = load_platereader_data(
    data_file="2026_3_25-bcecf-measurement-data.xlsx",
    platemap_file="2026_3_25-bcecf-measurement-platemap.csv",
    platereader="biotek",
)
print(sample_result)

PlateReaderResult with the following 2 blocks:
0: (2790, 38) kinetic read with reads: 440,535 (Fluorescence) (Plate 'Plate 1')
1: (2790, 38) kinetic read with reads: 490,535 (Fluorescence) (Plate 'Plate 1')

                    INFO     Parsing plate block 1                 biotek.py:190

                    INFO     Read metadata section, found 1 read   biotek.py:276
                             block(s).

                    WARNING  Failed to parse the data grid: Error biotek.py:1430
                             tokenizing data. C error: Expected                 
                             14 fields in line 4, saw 33

                    WARNING  Parsed data grid did not contain any  biotek.py:334
                             data.

                    INFO     Parsed 2 segment(s) successfully.     biotek.py:406

                    INFO     Found 0 segment(s) that could not be  biotek.py:407
                             associated with existing read data.

PlateReaderResult with the following 2 blocks:
0: (2790, 38) kinetic read with reads: 440,535 (Fluorescence) (Plate 'Plate 1')
1: (2790, 38) kinetic read with reads: 490,535 (Fluorescence) (Plate 'Plate 1')

sample_data = sample_result.take("kinetic")

4. Convert to pH and fit kinetics¶

Compute the 490/440 ratio on the data plate, then standardize(curve) to estimate pH from that ratio.

pH traces sit on a non-zero baseline (~6–8), so we fit sigmoid_drift_offset — a sigmoid with a linear drift term and a fitted constant baseline c. The plain sigmoid_drift model has no constant term and would mis-estimate both steady state and maximum velocity on a pH trace.

pH = sample_data.ratio("490,535", "440,535").standardize(curve)
kinetics = pH.fit_kinetics(fit_function_name="sigmoid_drift_offset")
kinetics.summary

Loading...

kinetics.plot() overlays each fit on its data. Facet on "Well" to get one fit per trace, matching the original analysis: the fit itself is always per-well, but the default facet (Name) averages the fit parameters across the three replicate wells and draws a single dashed line per condition.

Each panel shows the calculated pH trace (points), the model fit (dashed), the maximum-velocity tangent, and the steady-state and completion markers.

kinetics.plot(col="Well")

<seaborn.axisgrid.FacetGrid at 0x1645916d0>

Interpreting the kinetics¶

The fit is in pH units, so the derived metrics read directly as pH:

Max Velocity — peak rate of acidification/alkalinization, in pH per hour.
Lag Time — time before the response takes off.
Steady State — the plateau pH. Because sigmoid_drift_offset fits a baseline c, this sits on the real pH baseline rather than near zero.
Completion Time — when the response reaches its completion threshold.
R² / Good Fit — fit quality; inspect wells flagged Good Fit == False.

Note that well C5 does not fit (Good Fit == False); its trace is effectively flat, so the sigmoid model has no transition to fit.

kinetics.plot_response(x="Max Velocity", y="Steady State", hue="Name")

<seaborn.axisgrid.FacetGrid at 0x167545090>

Abbreviations¶

CDK: cell development kit