Gas chromatography–mass spectrometry (GC–MS) coupled with the acceptance test, partial least-squares regression (PLSR) analysis, validation experiment, and preference test was used to identify the key aroma compounds in dog foods (DFs). Six DFs were evaluated by spraying six palatability enhancers onto a basal DF. The differently flavored palatability enhancers were prepared by the Maillard reaction using different protein sources and reaction conditions. The intake ratios of the six DFs were tested by six adult beagle dogs and were classified into high, medium, and low levels. A total of 55 volatile compounds were identified using headspace solid-phase microextraction (HS-SPME) GC–MS. Correlation analysis of the volatile compounds associated with intake ratios using partial least-squares regression (PLSR) found nine significantly positive and three significantly negative compounds that made a significant contribution to the palatability of DFs. Validation tests undertaken by adding three significantly positive compounds, one significantly negative compound, and one nonsignificant compound into the odorless matrix successfully verified the accuracy prediction of the PLSR model.
The nine significantly positive compounds were heptanal, nonanal, octanal, (E)-2-hexenal, (E,E)-2,4-decadienal, 2-pentylfuran, 4-methyl-5-thiazoleethanol, 2-furfurylthiol, and (E)-2-decenal. The contributions of nine key aroma compounds were further analyzed by the preference test. (E)-2-decenal, 2-furfurylthiol, and 4-methyl-5-thiazoleethanol showed higher first choice, consumption ratio, and unit contribution rate and were vital to the overall preferred aroma of DFs.