Demo study: LPS challenge, IL-6 and TNF-α

Simulated data with known ground truth (immunoassay.simulate.default_demo_study) · generated 2026-09-29 13:14 · immunoassay v0.1.0

1 · Plate acceptance

A plate passes when R² ≥ 0.98, ≥ 75% of standard levels recover within 80–120% with CV ≤ 20%, a quantifiable range exists, and ≥ 67% of controls with expected ranges fall inside them. Limits are in-well concentrations.

2 · Standard curves

Weighted (1/y^2) least-squares fits. 5PL is chosen over 4PL only when its AICc is lower by more than 2. Weighted residuals should scatter evenly around zero; a curved pattern means the model is wrong for the data. The plate map shows raw OD. Excluded wells are marked ✗.

IL6_P1 · IL-6

IL6_P2 · IL-6

TNFa_P1 · TNF-α

TNFa_P2 · TNF-α

Excluded wells

None.

3 · Controls and flagged samples

Flagged replicate groups

below_lloq / above_uloq: outside the validated range (not reported as numbers). below_lod: not distinguishable from blank. cv_high: replicate CV above 20%.

4 · Dilution linearity

Samples run at more than one dilution should agree after correction (±20%). A concentration that rises with dilution points to matrix interference or the high-dose hook effect.

5 · Inter-plate normalization

Bridge controls shared by every plate estimate a multiplicative plate factor, on the log scale with the average plate as reference. Corrected concentration = raw / factor.

6 · Group comparisons

Tests are run on log10-transformed concentrations. Shapiro-Wilk on every group decides between parametric tests (Welch t / Welch ANOVA) and rank tests (Mann-Whitney / Kruskal-Wallis). Omnibus p-values are Benjamini-Hochberg adjusted across analytes; pairwise p-values are Holm-adjusted within each analyte. Hollow points are censored values that were substituted.

Omnibus tests

Pairwise comparisons

7 · Settings and warnings

{
  "model": "auto",
  "weighting": "1/y^2",
  "include_zero_standard": true,
  "qc": {
    "max_cv_pct": 20.0,
    "recovery_low_pct": 80.0,
    "recovery_high_pct": 120.0,
    "edge_tolerance_pct": 5.0,
    "lod_sd_multiplier": 3.0,
    "outlier_z": 3.5,
    "max_std_exclusions": 2,
    "min_r_squared": 0.98,
    "min_std_pass_frac": 0.75,
    "min_control_pass_frac": 0.6666666666666666
  },
  "stats": {
    "transform": "log10",
    "censored": "substitute",
    "harmonize_lloq": true,
    "lloq_divisor": 1.4142135623730951,
    "alpha": 0.05,
    "normality_alpha": 0.05,
    "exclude_failed_plates": true,
    "value_col": "conc_norm"
  },
  "normalize": true,
  "linearity_max_deviation_pct": 20.0,
  "group_order": [
    "vehicle",
    "LPS",
    "LPS+drug"
  ],
  "control_ranges": [
    {
      "analyte": "IL-6",
      "sample_id": "CTRL_LO",
      "nominal": 40.0,
      "low": 30.0,
      "high": 50.0
    },
    {
      "analyte": "IL-6",
      "sample_id": "CTRL_MID",
      "nominal": 160.0,
      "low": 120.0,
      "high": 200.0
    },
    {
      "analyte": "IL-6",
      "sample_id": "CTRL_HI",
      "nominal": 600.0,
      "low": 450.0,
      "high": 750.0
    },
    {
      "analyte": "TNF-\u03b1",
      "sample_id": "CTRL_LO",
      "nominal": 20.0,
      "low": 15.0,
      "high": 25.0
    },
    {
      "analyte": "TNF-\u03b1",
      "sample_id": "CTRL_MID",
      "nominal": 70.0,
      "low": 52.5,
      "high": 87.5
    },
    {
      "analyte": "TNF-\u03b1",
      "sample_id": "CTRL_HI",
      "nominal": 250.0,
      "low": 187.5,
      "high": 312.5
    }
  ]
}