NIRS Calibration Model for Evaluating Boiled Yam Texture (Hardness) as a High Throughput Method

The calibration statistics for the NIRS prediction of the textural attribute (hardness) of the boiled yam. The development of NIRS prediction models for rapid evaluation of texture (a key quality trait) of boiled yam was reported in this document. Thirty yam genotypes ( TDr and TDa) harvested from the experimental field of IITA were used for the experiment. Boiled yam samples were prepared by boiling in a steam cooker for 20 minutes following a standardized operating procedures developed in the project. The instrumental hardness was determined using Stable Micro Systems TA.XTplus texture analyzer to generate reference data, while the NIRS spectra data was also collected using a NIRS spectrophotometer (FOSS XDS Rapid content analyzer). For the preprocessing of the spectra data, Standard Normal Variate Detrend (SNVD) and Savitzky Golay were used as mathematical treatments. Prediction model was later developed using a modified partial least square (MPLS) algorithm. The NIRS model developed had a good coefficient of determination in calibration (R2cal) of 0.97. Another set 10 samples were used to validate the model giving a coefficient of determination in prediction(R2pred) of 0.92. This experiment has shown that NIRS can rapidly predict the texture of boiled yam, making the technology applicable in yam breeding for routine screening for textural attributes of boiled yam.