The Evidence Layer for Scientific AI

 Quick Analysis Plan Copied

 Long Analysis Plan

What the supplied evidence supports — and what it does not

The records contain quantitative LODs exclusively for Raman-family methods (conventional Raman, time-gated Raman, SERS). No FTIR, NMR, or NTA limit-of-detection values appear anywhere in the supplied claims, so a cross-technique LOD graph would require inventing data, which is refused.

Sourced Raman LODs:

• Polystyrene particles (785 nm confocal Raman): LOD ≈ 7.47–7.64 × 10¹² particles/mL for 50 nm particles and ≈ 6.09–6.24 × 10⁸ particles/mL for 1000 nm particles; 25 nm particles were not reliably detectable because of CCD saturation Raman particle LOD vs size
• SERS probiotics in yogurt: LOD 10⁴–10⁶ CFU/mL depending on strain (L. bulgaricus 10⁴; L. acidophilus 10⁵; S. thermophilus 10⁶; yogurt matrix 10⁵ CFU/mL) SERS probiotic LODs in yogurt
• SERS ricin in milk: LOD 4 μg/mL within 20 min SERS ricin detection in milk
• Raman heparin in bovine serum: detection limit ≈ 8 USP/mL with PLS (R² > 0.91, RMSEP ≈ 4 USP/mL) Raman heparin serum PLS LOD
• SERS M. paratuberculosis immunoassay: experimental LOD > 3 orders of magnitude below the theoretical ≈ 8 × 10⁵ cells/mL, likely inflated sensitivity from shed MAP2121c protein rather than whole bacteria SERS MAP immunoassay LOD caveat

The one defensible graph: Raman LOD vs particle size

The only internally consistent LOD series (single technique, single unit, single matrix) is the particle-size study, plotted on a log scale below.

Note: these are the prepared concentration series; the reported detected LODs (≈ 7.5 × 10¹² particles/mL for 50 nm; ≈ 6.2 × 10⁸ for 1000 nm) sit near the low-concentration end of each series Raman LOD data availability and methods.

What is missing for the requested cross-technique figure

To build the intended FTIR vs Raman vs NMR vs NTA comparison, the review must supply: (1) an FTIR LOD with matched analyte and units, (2) an NMR LOD, and (3) an NTA concentration LOD. Only then could normalized LODs (e.g., log particles/mL for the same model particles) be plotted. Directly mixing particles/mL, CFU/mL, μg/mL, and USP/mL on one axis would be a category error, not a comparison. Confidence in this limitation is high; confidence in any future cross-technique ranking would depend on identical analytes and matrices.