Truace
Subscribe

Machine learning for dairy pathogen detection and origin tracing under multivariable storage conditions

Source article: Sample-efficient machine learning for dairy pathogen detection and origin traceability under multivariable conditions using gold nanoparticle colorimetry

Storage temperature, humidity, and time jointly influence pathogen growth in dairy products, yet most rapid detection methods are validated under single-variable conditions that do not represent real supply chain scenarios. Testing across these combined conditions sharply increases the number of samples needed, whereas dairy safety studies usually work with small data sets, which limits the use of conventional machine learning. Here we combined gold nanoparticle colorimetric sensing, a time-temperature-humidity…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1875· cite & verify
Sample-efficient machine learning for dairy pathogen detection and origin traceability under multivariable conditions using gold nanoparticle colorimetry
The technical control of dairy products; a treatise on the testing, analyzing, standardizing and the manufacture of dairy products by Mojonnier, Timothy, 1876- Troy, Hugh Charles, 1867- joint author. Public domain
Trace impact readingContested
P 65The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 66The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

Researchers combined an aptamer-based gold nanoparticle color test, a framework tracking time, temperature and humidity, and an optimized Random Forest model to detect three bacterial pathogens and trace the origin of milk and milk powder. They used 2,592 measurement records from three independent replicates and samples from four Chinese regions.

The work matters because rapid dairy tests are usually validated under single conditions and machine learning struggles with small food-safety data sets. The approach could support recalls, origin fraud screening and source tracing at distribution checkpoints, but the authors describe it as proof-of-concept because the sampling was narrow.

Main points

  1. The assay used aptamers and needed no complex sample preparation, nucleic acid extraction, or cold-chain reagent storage.
  2. Detection limits were 104 cfu/mL for S. aureus, 108 for S. typhimurium and 107 for E. coli, with recoveries of 96.4-108.4%.
  3. The optimized model reached 94.6% accuracy for pathogen classification, with areas under the curve of at least 0.97 for both classification tasks.
  4. At 37°C pathogens were estimated to reach reference limits within about 4 h, versus 8-9 h at 25°C.

The gain

A sample-efficient Random Forest model trained on colorimetric readings predicted pathogen concentration and identified dairy pathogen species and production region with high accuracy under combined storage conditions.

The problem

Because each region was represented by a single commercial product per type and one production batch, the results are only proof-of-concept and may not generalize to wider supply chains.

The rundown

The colorimetric assay was tested in commercially sterilized fluid milk and milk powder from Shanghai, Yunnan, Inner Mongolia and Xinjiang. Models were evaluated against plate-count-calibrated concentrations and the inoculated species and region of purchase as reference labels.

Time-temperature-humidity monitoring showed that growth speed depended heavily on storage temperature. At 4°C pathogen levels stayed low for about the first 12 days of a 16-17-day monitoring period, and region-specific growth patterns enabled origin identification.

The Random Forest was tuned with a target-R2-guided hyperparameter strategy designed to work within the small sample sizes typical of dairy safety studies.

What this doesn’t fix

Each region was represented by a single commercial product per type and a single production batch, so the findings are proof-of-concept and not validated across wider products or batches.

Sources

  1. Peer-reviewedJournal of Dairy Science2026-10-09

The debate