Plate Reader Normalizer
Plate readers measure light from hundreds of tiny wells at once, and each brand saves the numbers in its own awkward layout. Open the export here to see each plate as a coloured grid, check its control wells, and get how strongly each compound worked and at what dose.
Open the export your plate reader wrote, check each plate's controls and Z′, and get percent inhibition and IC50s without re-keying the grid into Excel.
Open plate reader exports
Drop CSV, TXT or Excel exports from BMG, SoftMax Pro, Tecan, EnVision or Gen5 anywhere on this page. Several files open together.
Every plate as a heatmap
Each plate, read and kinetic timepoint, with the value of any well on hover or focus.
Controls and QC
Mark controls by dragging across wells or by rule. Z′, signal to background, control CVs and edge effects per plate.
Curves and export
Four-parameter fits and IC50s for a dilution layout, and one tidy CSV of every well.
About this data
What it reads
Each file is read in the layout its software writes. The layouts were checked against real exports published as parser test data, chiefly Benchling's open-source allotropy converters.
- BMG Labtech MARS (CLARIOstar, PHERAstar, Omega, and SMART Control workbooks): the plate-layout CSV, with the test name, the ID1 plate name, and each “Raw Data” or calculated grid. A layout grid of sample codes (S, X, N, P, B) and a standard concentration grid are read too.
- Molecular Devices SoftMax Pro text export: every Plate: block, in plate or time format, endpoint or kinetic, one read per wavelength, in UTF-8 or the UTF-16 SoftMax Pro writes. Group: tables give sample names, control groups and concentrations.
- Tecan i-control and SparkControl Excel or CSV: the 8 × 12 or 16 × 24 matrix under a “<>” corner cell, kinetic tables with Cycle Nr. and Time [s], and well lists, one read per label. Layouts from wellr, flopr and fpcountr example files.
- Revvity (PerkinElmer) EnVision CSV: each plate and repeat, its results and calculated results, including the unlabelled 16 × 24 block some protocols write.
- Agilent BioTek Gen5 text export: the Results grid, where several reads share one row letter and are named in the last column, and the Layout grid's well IDs and Conc/Dil values.
Plate size is the smallest standard plate (96, 384 or 1536 wells) that holds every well in the file, or the size the file names if larger. Readings that are not numbers, such as OVRFLW, Masked or >3.5, are left empty rather than guessed.
The arithmetic
Percent inhibition is 100 × (negative-control mean − well) ÷ (negative-control mean − positive-control mean), so the negative controls average 0% and the positive controls 100%. Percent of control is 100 minus that: the negative controls average 100%.
Z′ is 1 − 3 (σp + σn) ÷ |μp − μn|, from Zhang, Chung and Oldenburg, Journal of Biomolecular Screening, 1999, with sample standard deviations of the two control groups. Their reading: 0.5 or more is an excellent assay, between 0 and 0.5 a marginal one, and 0 or less a screen that cannot separate hits from noise. Signal to background is the higher control mean over the lower. Each control's CV is its standard deviation over its mean.
Edge effects compare sample wells (not controls or excluded wells) on the outer ring of the plate with those inside it, and name the row and column whose mean sits furthest from the median. Evaporation and temperature gradients usually show up as low outer rows and columns.
The dose-response fit is the four-parameter logistic y = bottom + (top − bottom) ÷ (1 + 10(log IC50 − log x) × Hill), fitted by Levenberg–Marquardt least squares to the normalized values of each series. An IC50 outside the tested concentrations is marked with ~. Concentrations come from the file (SoftMax Pro groups, Gen5 Conc/Dil, BMG standard concentrations) or from a rule: a top concentration, a dilution factor and how many points run along each row or down each column.
The export
Tidy CSV writes one row per well, per read, per timepoint, for every plate open: plate, well, row, col, read, timepoint, raw, normalized, role, and the concentration when a dilution layout is set. Normalized values use each plate's own controls. The file opens straight into R, Python, Spotfire or a pivot table.
The samples
The three sample exports were generated by Arca for this page, laid out exactly as the vendors' software writes them: a 384-well CLARIOstar fluorescence plate (MARS CSV), two 96-well SoftMax Pro absorbance plates with groups and concentrations (UTF-16 text), and a 384-well Tecan Spark workbook with a six-cycle kinetic read and an endpoint read. Each has vehicle and reference-inhibitor control columns and 10-point, 1:3 dilution series with known IC50s, measurement noise and a slight edge effect. The compounds are fictional.
Frequently asked questions
- Are my plate files uploaded?
- No. The files are read in a background thread of this browser tab, and nothing is sent to Arca or anyone else. Close the tab and they are gone. Only the control and dilution rules are kept, in the page address, so a saved or shared link brings back the layout but not the data.
- My reader software is not listed. Will it open?
- Probably. Besides the five vendor layouts, the page looks anywhere in a CSV, text file or workbook for a grid with column numbers across the top and row letters down the side, a bare block of numbers 12, 24 or 48 wide, or a list with a well column (A1, B12). Old binary .xls workbooks do not open; save them as .xlsx or CSV first.
- Which control is negative and which is positive?
- The negative control is 0% effect: vehicle or DMSO wells, full signal in an inhibition assay. The positive control is 100% effect: a reference inhibitor at a saturating dose, or no enzyme. The arithmetic does not care which reads higher, so the same marking works for gain-of-signal assays.
- Why does my IC50 differ from GraphPad Prism's?
- The model is the same variable-slope four-parameter logistic Prism calls log(inhibitor) vs. response, fitted by least squares with all four parameters free. Differences come from constraints (Prism users often fix the top at 100 and the bottom at 0), from weighting, or from which wells are excluded. The 95% interval here is the usual approximate one, from the curvature of the fit.
- Is this validated for GLP or GMP work?
- No. It is a fast way to check plates and get numbers into a tidy table. Reportable results should come from your validated analysis system.