NIRS Calibration Database on fresh grated cassava for DM at NaCRRI Uganda

Cassava utilization for food and/or industrial products depends on inherent root dry matter content (DMC). Accordingly, there is need to develop NIRS models to aid breeding and selection of dry matter content as a critical industrial trait taking care of root sample preparation and cassava germplasm diversity available in Uganda. For calibration development (300 clones) were used. Grating was the main sample preparation method. Reference DMC data was obtained by oven-drying. DMC in calibration samples ranged from 20 to 45%. Modified Partial least square (MPLS) regression model was used with the pretreatment carried out by SNVD over the segment 1100-2500nm at a 2nm step. The number of factors used was 8 while a total number of 13 outliers were eliminated. The regression coefficients (R2) of 0.997 were obtained while the R2 for cross validation was 0.993. In addition, the RMSEC was 0.28 while the RMSECV was 0.43. These findings show the robustness of NIRS in estimation of dry matter content and thus justify its use in routine cassava breeding operations. WP3