Determination And Establishment Of NIR Rapid Detection Model On Macro-Nutrients For Mongolian Horse Milk

Horse milk is abundant in nutrients and has certain health functions,thus very welcomed by Mongolians and other northern people.Recent years Mongolian horse milk has been commercializing,however Mongolian horse milk lacks systematic nutrients data,and to date,no near-infrared(NIR)horse milk composition prediction model or system have been established yet.A total of 92 Mongolian raw horse milk and natural fermented horse milk/koumsis,together with 44 Holstein cow and 27 goat milk were collected from major Mongolian horse milk producing areas of Inner Mongolia.Routine nutrients such as protein,fat,lactose,moisture and ash of horse milk were determined according to the national standard methods;a desk top and a micro NIR Spectrometers were applied to milk NIR spectra scanning.NIR spectrum sector choice and pre-processing such as derivative,multiple scattering correction(MSC)and standard normal variate transform(SNV)on the NIR spectra were conducted by a Chemometrics software,and processed or transformed spectra were calibrated with the nutrient values by partial least square(PLS),and NIR milk nutrients prediction models were established and evaluated mainly by the coefficient determination(R2),Root mean square error of cross validation(RMSECV),relative percent deviation(RPD)and relative standard Deviation(RSD).As results,the optimal parameters of general prediction model of horse,cow and goat milk nutrients are far less ideal than individual modeling parameters,so it is necessary to establish a individual model for horse milk nutrients prediction.On desk top NIR spectrometer,the ideal model of horse milk were established by whole NIR spectrum section and second derivative transformation,and the parameters R2,RMSECV,RPD and RSD were:protein 0.9496,0.0538,4.4715 and 2.95%,fats 0.9673,0.0949,5.5592 and 9.25%,lactose 0.9710,0.2500,5.8122 and 2.86%and ash 0.9480,0.0150,4.4101 and 4.91%,respectively,as well as SNV and second derivative spectra processing to established moisture model,and its R2,RMSECV,RPD and RSD were 0.9514,0.4314,4.4101 and 0.41%respectively,all parameters meet the accurate prediction requirements.pH value has a certain influences on the precision of the models,the parameters of the general horse milk model,raw horse milk model and fermented horse milk model were improved successively in the order,i.e.the specific model is more precise.Further it has a certain possibility and feasibility to establish NIR model to predict pH value of the horse milk.Horse milk models established on Micro NIR spectrometer are inferior to Desk top NIR spectrometer,more meticulous optimizing work are needed.