High-throughput profiling of sweet potato vine biomass for cellulosic ethanol production using near-infrared spectroscopy and chemometrics

Abstract

Sweet potato (Ipomoea batatas L.) vines, despite their abundance as agricultural by-products, remain underexplored as lignocellulosic feedstocks for bioethanol production. To address this gap, the present study introduces a novel high-throughput phenotyping strategy that integrates near-infrared spectroscopy (NIRS) with chemometrics to rapidly evaluate the bioethanol potential of sweet potato vine biomass. A diverse panel of 115 germplasm accessions was analyzed to develop robust quantitative and qualitative NIRS models. Seven optimized partial least squares regression (PLSR) models, encompassing cellulose, hemicellulose, lignin, soluble sugar, hexose, pentose, and theoretical ethanol potential (TEP), exhibited exceptional accuracy, with determination coefficients (R²) of 0.92–0.96 (calibration), 0.90–0.95 (cross-validation), and 0.87–0.94 (external validation). The ratio of prediction to deviation (RPD) values ranged from 5.64 to 8.33, confirming strong predictive capacity. Additionally, a complementary partial least squares-discriminant analysis (PLS-DA) model achieved 98% calibration accuracy and 93% validation accuracy in classifying feedstock quality grades, enabling efficient germplasm screening. This study, for the first time, demonstrates that NIRS-based phenotyping can replace traditional wet chemistry methods for large-scale evaluation of sweet potato vine biomass. Our approach provides a paradigm for accelerating the development of dedicated bioenergy crops through rapid trait profiling and precision breeding.