Quantification method and reproducibility

Before you start

1. Measurement uses 8/16-bit integer grayscale or RGB pixels from original PNG or single-page TIFF files. JPEG can be used for figures but not as scientific input to this method. Unsupported transparency, multipage, floating-point or CMYK inputs receive explicit errors.

2. Review the signal and non-overlapping background in each lane. Count each source pixel once when its center is inside the region; rotated regions do not use interpolated preview intensities.

3. For dark bands, net integral S = N × B − ΣI; for light bands, S = ΣI − N × B. N is the signal pixel count and B the background mean. Negative net values are retained. The sample ratio is target net integral divided by its matching reference net integral.

Display settings are separate from measurement

Brightness, contrast, gamma, black/white points and display inversion do not alter originals or enter original-pixel integration. 2.0 retains algorithm identifier native-local-bg-v3. Historical results retain their original method. Review and recalculate after changing sources, regions or reference relationships.

Input changes and recalculation

After replacing a source, moving or rotating signal/background regions, changing target/reference relations, or changing lane groups or repeat identities, review and calculate from the current inputs. Historical analyses retain their original inputs and algorithm and are not results for the edited setup. Calculate rechecks source files. Missing or changed sources are excluded from current ratios, n and statistics; recalculate after restoring them. A malformed project should produce an explicit error. Keep the original and restore a readable backup instead of renaming it or deleting fields to hide corruption.

Which pixels enter the calculation

Regions use original pixel centers and half-open local coordinates: left/top boundaries are included, right/bottom boundaries are excluded. Adjacent lanes cannot count the same pixel twice. Grayscale uses native integers. RGB uses encoded values, I = 0.2126R + 0.7152G + 0.0722B, without linearization. Measurement ignores ICC display profiles and first canonicalizes EXIF orientation.

Bit depth and input limits

8-bit and 16-bit inputs retain their native value ranges; 16-bit data is not first reduced to 8-bit. Only single-page integer PNG/TIFF inputs are accepted. CMYK, multipage, floating-point inputs and regions containing non-opaque sampled pixels cannot be measured. Re-export a supported original format from the acquisition software. Renaming JPEG or saving it as PNG cannot recover lost measurement information. Zero or the encoded maximum indicates an encoding endpoint, not proof of detector saturation.

Background variation and noise diagnostics

M is the background pixel count and must be at least two. Sample variance is s² = Σ(b−B)²/(M−1), with background SD = √s². Net-integral standard error is SE = √[N×s²×(1+N/M)], assuming independent, homoscedastic pixel noise. SNR = S/SE; SNR is left empty when SE is zero. SNR below 3 is only a review heuristic, not an experimentally validated LOD/LOQ. Gradients or correlated noise may violate these assumptions.

Results and diagnostic messages

Endpoint pixels no longer block ratios behind an endpoint-review gate. Invalid denominators and unmeasurable inputs still retain their reasons rather than producing Infinity or invented zeros. Noise messages are not validated detection limits, and no warning does not prove experimental validity.

What recalculation can establish

Raw sums and identically defined background subtraction can be compared with Fiji. Agreement does not validate background selection: an injected signal of 4000 can yield 7000 or 10400 with different regions on a gradient background. Real experiments still need evidence for the linear range, exposure and background choice.

Validate the real experiment separately

Before a comparison, record the dilution series, exposure range, background strategy and operator-review method. Retain unmodified originals and signal/background coordinates, and prespecify acceptable linearity, bias, variability and repeat agreement. Compare exposures, background locations and operators, reporting failures and exclusions. This is a validation plan to carry out, not a real-WB study completed by this candidate. Formula checks, original pixels, Fiji agreement, application operation and real experiments are five distinct levels.

Scope of the current local checks

The local 2.0.0 review on 20 October 2026 compared 28 specified synthetic 8/16-bit grayscale TIFF cases with Fiji (ImageJ 1.53t). All 364 numerical checks had zero absolute difference. A separate independent check covered 107 statistical formula and design safeguards. These results cover the recorded inputs and definitions. They do not validate ImageJ Gels, rolling-ball backgrounds, every RGB/PNG case or real blots, nor establish antibody specificity, biological independence or comparability across dates.

A short methods description

Western blot images were quantified using WB Figure Studio. Signals were corrected for lane-specific local background and normalized to the corresponding loading control. Add the actual biological replicate count, control baseline and statistical method.

Run statistics for the actual design

Confirm independent samples, design and between-run comparability before choosing a test. Tests do not run automatically. Available methods are Welch t for two independent groups, Welch ANOVA for multiple independent groups, paired t for explicitly paired groups, and blocked ANOVA for complete blocks. Prespecified comparisons use Holm correction. Each group needs at least two valid independent units. Missing pairs or blocks are not silently dropped. Analyze original ratios or their natural logarithms, with log analysis requiring positive values.

Code, records and licensing

Exported records identify regions, pixel counts, background, sources and the algorithm. SHA identifies files and checks integrity; it is not tamper-proof certification. The license for the 2.0 scientific review package remains an ineffective draft, and no public download is offered. Historical materials apply only to their frozen versions. The app remains proprietary, and matching recalculations do not constitute scientific certification.