Thai University RankingsRESEARCH RADAR
Experimental classifier evidence

Two-dimensional FTIR and machine learning detected molecular signatures of cadmium exposure in fish

Clarias batrachus fingerlings were exposed to multiple cadmium levels and muscle tissue analyzed with FTIR, 2DCOS and PCMW2DCOS. Lipid disruption dominated around 0.8-1.9 ppm, early protein changes appeared at 0.4-1.2 ppm, and carbohydrate alterations at 3.2-4.5 ppm. Amide-I shifts from alpha helices to beta turns supported protein instability. PCA-LDA achieved 96% accuracy, 96% F-score and MCC 0.94; linear and polynomial SVMs were reported around 96-97%, suggesting earlier stress detection than conventional one-dimensional FTIR.

01

Key findings

  • Dose windows showed lipid CH2/CH3, protein-secondary-structure and glycogen changes. PCA-LDA reached 96% accuracy/F-score and MCC 0.94; linear/polynomial SVMs were about 96-97%, versus 91% for RBF and 87% for sigmoid.
02

Why this matters globally

The approach could support heavy-metal biomonitoring in aquatic species and food safety if spectral standards are built and specificity is demonstrated across metals, diseases and environmental conditions.

03

Thai researcher contribution

B. Velmurugan is affiliated with Chaopraya University, linking a Thai institution to environmental toxicology, spectroscopy and machine learning. Individual author roles are not specified in the abstract.

04

Limitations to consider

Sample size, train-test splitting, cross-validation and external validation are absent from the abstract, preventing assessment of overfitting or leakage. SVM accuracy is inconsistently stated as 96% and 97%. Controlled exposure classification is easier than field samples confounded by diet, age, disease and multiple pollutants.

05

Verify the original sources

Spectroscopy LettersRead the original article

DOI: 10.1080/00387010.2026.2652071

KEEP EXPLORING

More Thai research to explore