Protein, Fat and Moisture Analyses of Fresh Fishmeal with an Antaris II FT-NIR Analyzer

Importance of the topic

Fishmeal is a globally important high-protein feed ingredient for aquaculture and livestock. Its composition—primarily protein, fat and moisture—influences nutritional value, storage stability and processing behavior. Conventional analyses (Kjeldahl for protein, Soxhlet for fat, and drying/Karl Fischer for moisture) are time-consuming, hazardous, and ill-suited for near‑real‑time process control. Demonstrating rapid, accurate alternatives that reduce chemical use, speed decision-making and enable at‑line or on‑line quality control is therefore highly relevant for producers and analytical laboratories.

Objectives and study overview

This application study evaluated whether a single FT‑NIR measurement using the Thermo Scientific Antaris II (laboratory) and Antaris MX (process) analyzers can replace traditional wet chemistry for routine quantification of protein, fat, moisture (water) and other constituents (ammonia, ash) in fresh fishmeal. The goal was to develop robust calibrations from production samples, assess prediction performance against reference methods, and demonstrate transferability from lab to process analyzer.

Methods and instrumentation

The study used 247 production-derived standards capturing seasonal and raw-material variability. Spectra were acquired on an Antaris II Method Development Sampling (MDS) system equipped with a sample cup spinner; pressing samples into a petri dish improved spectral consistency. Key spectral acquisition parameters were: spectral range 9000–4000 cm⁻¹, resolution 16 cm⁻¹, 32 scans, and ~12 s collection time. Spectral pretreatment applied the first derivative with a Norris smoothing (length=3, gap=3). Partial least squares (PLS) regression was used on the full spectral range to build calibrations with Thermo Scientific TQ Analyst software. After laboratory model development, calibrations were transferred to an Antaris MX process analyzer.

Used instrumentation

• Thermo Scientific Antaris II FT‑NIR analyzer with Method Development Sampling (MDS) sample cup spinner
• Thermo Scientific Antaris MX FT‑NIR process analyzer (for model transfer)
• Software: Thermo Scientific TQ Analyst (PLS calibration)

Main results and discussion

The FT‑NIR PLS models achieved strong agreement with primary reference methods across measured constituents. Summary of calibration performance:

• Protein: correlation coefficient ~0.96; RMSEC ≈0.57; RMSECV ≈0.71; calibration range ~68.7–72.3
• Fat (total lipid): correlation coefficient ~0.984; RMSEC ≈0.27; RMSECV ≈0.31; range ~8.3–10.5
• Ammonia (NH3): correlation coefficient ~0.97; RMSEC ≈0.012; RMSECV ≈0.015; range ~0.12–0.25
• Ash: correlation coefficient ~0.935; RMSEC ≈0.605; RMSECV ≈0.74; range ~12.0–15.67
• Water (moisture): correlation coefficient ~0.985; RMSEC ≈0.242; RMSECV ≈0.32; range ~5.6–9.3

Overall, correlation coefficients ranged from ~0.935 to 0.985, indicating excellent fits for most constituents (especially fat and moisture). The use of first‑derivative pretreatment and full‑range PLS leveraged overlapping overtone and combination bands typical of organic constituents in the NIR region. The study also noted that including real production variability (seasonal and geographic differences) in the calibration set improved model robustness. Calibration transfer to the Antaris MX process analyzer was reported as straightforward, supporting at‑line deployment.

Benefits and practical applications of the method

• Speed: single FT‑NIR measurement (~12 s acquisition) versus hours for Kjeldahl and Soxhlet workflows.
• Safety and sustainability: eliminates hazardous reagents and large solvent volumes used in Kjeldahl and Soxhlet methods, reducing chemical waste and exposure risk.
• Operational efficiency: enables near‑real‑time, at‑line or on‑line monitoring, allowing immediate process adjustments and reduced product loss from out‑of‑spec material.
• Transferability: chemometric models were shown to transfer from laboratory Antaris II to process Antaris MX hardware, facilitating scale‑up from development to production.
• Multiparameter capability: simultaneous prediction of protein, fat, water and minor components from one spectrum increases throughput and lowers per‑sample cost.

Future trends and potential applications

• Wider on‑line integration: embedding FT‑NIR analyzers in process lines for continuous monitoring and closed‑loop control of drying, pressing and blending stages.
• Improved model maintenance: deployment of calibration transfer strategies, robust validation procedures and adaptive or federated calibration approaches to manage batch, seasonal and site variability.
• Advanced chemometrics and AI: use of nonlinear models, variable selection, and machine learning approaches to further improve prediction accuracy for minor constituents and compositional extremes.
• Miniaturization and remote analytics: emergence of compact NIR probes and cloud‑based analytics for decentralised monitoring across multiple plants.
• Regulatory and QC adoption: wider acceptance of validated FT‑NIR methods as official or routine alternatives to wet chemistry in feed and food quality control.

Conclusion

The application demonstrates that FT‑NIR spectroscopy using the Antaris family can accurately and reproducibly quantify protein, fat, moisture and other constituents in fresh fishmeal. Calibrations built from production samples delivered strong correlation to reference methods, substantially faster results, and removal of hazardous wet chemistry steps. The ability to transfer models to process analyzers supports real‑time quality control and improved operational efficiency in fishmeal production.

Reference

Wiertz M. Protein, Fat and Moisture Analyses of Fresh Fishmeal with an Antaris II FT‑NIR Analyzer. Thermo Fisher Scientific Application Note 51873, 2010.