Prediction of total lipids and fatty acids in black soldier fly (Hermetia illucens L.) dried larvae by NIR-hyperspectral imaging and chemometrics

J P Cruz-Tirado; Matheus Silva Dos Santos Vieira; Ramon Sousa Barros Ferreira; José Manuel Amigo; Eduardo Augusto Caldas Batista; Douglas Fernandes Barbin

doi:10.1016/j.saa.2024.125646

Prediction of total lipids and fatty acids in black soldier fly (Hermetia illucens L.) dried larvae by NIR-hyperspectral imaging and chemometrics

Spectrochim Acta A Mol Biomol Spectrosc. 2024 Dec 19:329:125646. doi: 10.1016/j.saa.2024.125646. Online ahead of print.

Authors

J P Cruz-Tirado¹, Matheus Silva Dos Santos Vieira¹, Ramon Sousa Barros Ferreira¹, José Manuel Amigo², Eduardo Augusto Caldas Batista¹, Douglas Fernandes Barbin³

Affiliations

¹ Department of Food Engineering, School of Food Engineering, University of Campinas, Campinas, SP, Brazil.
² IKERBASQUE, Basque Foundation for Science, 48011 Bilbao, Spain; Department of Analytical Chemistry, University of the Basque Country UPV/EHU, 48080, Bilbao, P.O. Box 644, Basque Country, Spain.
³ Department of Food Engineering, School of Food Engineering, University of Campinas, Campinas, SP, Brazil. Electronic address: dfbarbin@unicamp.br.

PMID: 39718268
DOI: 10.1016/j.saa.2024.125646

Abstract

The unique fatty acid composition of BSF larvae oil makes it suitable for various applications, including use in animal feed, aquaculture, biodiesel production, biomaterials, and the food industry. Determination of BSF larvae composition usually requires analytical methods with chemicals, thus needing emerging techniques for fast characterization of its composition. In this study, Near Infrared Hyperspectral Imaging (NIR-HSI) (928 - 2524 nm) coupled with chemometrics was applied to predict the lipid content and fatty acid composition in intact black soldier fly (BSF) larvae. Partial Least Squares Regression (PLSR) and Support Vectors Machine Regression (SVMR) models, combined with two variable selection methods, Interval Partial Least Squares (iPLS) and Bootstrapping Soft Shrinkage (BOSS), were compared. PLSR reached a good performance to predict myristic acid with Root Mean Square Error in prediction (RMSEP) = 0.45 %, while SVMR reached values of Ratio to Prediction Deviation (RPD) > 3 to predict total lipid content, lauric acid, myristic acid, palmitic acid and oleic acid. In addition, selecting wavelength by BOSS improved PLSR models (6 - 15 % increases in RPD), while iPLS improved SVMR model to predict palmitic acid (16 % increases in RPD). The study emphasizes the advantages of NIR-HSI as a non-invasive, rapid method for lipid and fatty acid quantification, which can be highly valuable for industrial applications such as monitoring BSF larvae feeding systems to ensure high-quality oil production.

Keywords: Bootstrapping; Edible insects; Lipid content; Multivariate analysis; Near infrared.