Julie E Gervis1, Kenneth K H Chui2, Jiantao Ma3, Oscar Coltell4,5, Rebeca Fernández-Carrión5,6, José V Sorlí5,6, Rocío Barragán5,6, Montserrat Fitó5,6,7, José I González5,6, Dolores Corella5,6, Alice H Lichtenstein1. 1. Cardiovascular Nutrition Laboratory, Jean Mayer USDA Human Nutrition Research Center on Aging, Tufts University, Boston, MA, USA. 2. Department of Public Health and Community Medicine, Tufts University School of Medicine, Boston, MA, USA. 3. Department of Nutrition Data Science, Friedman School of Nutrition Science and Policy, Tufts University, Boston, MA, USA. 4. Department of Computer Languages and Systems, University of Jaume I, Castellón, Spain. 5. CIBER Fisiopatología de la Obesidad y Nutrición, Instituto de Salud Carlos III, Madrid, Spain. 6. Department of Preventive Medicine and Public Health, University of Valencia, Valencia, Spain. 7. Cardiovascular Risk and Nutrition Research Group, Hospital del Mar Medical Research Institute (IMIM), Barcelona, Spain.
Abstract
BACKGROUND: Current approaches to studying relations between taste perception and diet quality typically consider each taste-sweet, salt, sour, bitter, umami-separately or aggregately, as total taste scores. Consistent with studying dietary patterns rather than single foods or total energy, an additional approach may be to study all 5 tastes collectively as "taste perception profiles." OBJECTIVE: We developed a data-driven clustering approach to derive taste perception profiles from taste perception scores and examined whether profiles outperformed total taste scores for capturing individual variability in taste perception. METHODS: The cohort included 367 community-dwelling adults [55-75 y; 55% female; BMI (kg/m2): 32.2 ± 3.6] with metabolic syndrome from PREDIMED-Plus, Valencia. Cluster analysis identified subgroups of individuals with similar patterns in taste perception (taste perception profiles); quantitative criteria were used to select the cluster algorithm, determine the optimal number of clusters, and assess the profiles' validity and stability. Goodness-of-fit parameters from adjusted linear regression evaluated the individual variability captured by each approach. RESULTS: A k-means algorithm with 6 clusters best fit the data and identified the following taste perception profiles: Low All, High Bitter, High Umami, Low Bitter & Umami, High All But Bitter and High All But Umami. All profiles were valid and stable. Compared with total taste scores, taste perception profiles explained more variability in bitter and umami perception (adjusted R2: 0.19 vs. 0.63, respectively; 0.40 vs. 0.65, respectively) and were comparable for sweet, salt, and sour. In addition, taste perception profiles captured differential perceptions of each taste within individuals, whereas these patterns were lost with total taste scores. CONCLUSIONS: Among older adults with metabolic syndrome, taste perception profiles derived via data-driven clustering may provide a valuable approach to capture individual variability in perception of all 5 tastes and their collective influence on diet quality. This trial was registered at https://www.isrctn.com/ as ISRCTN89898870.
BACKGROUND: Current approaches to studying relations between taste perception and diet quality typically consider each taste-sweet, salt, sour, bitter, umami-separately or aggregately, as total taste scores. Consistent with studying dietary patterns rather than single foods or total energy, an additional approach may be to study all 5 tastes collectively as "taste perception profiles." OBJECTIVE: We developed a data-driven clustering approach to derive taste perception profiles from taste perception scores and examined whether profiles outperformed total taste scores for capturing individual variability in taste perception. METHODS: The cohort included 367 community-dwelling adults [55-75 y; 55% female; BMI (kg/m2): 32.2 ± 3.6] with metabolic syndrome from PREDIMED-Plus, Valencia. Cluster analysis identified subgroups of individuals with similar patterns in taste perception (taste perception profiles); quantitative criteria were used to select the cluster algorithm, determine the optimal number of clusters, and assess the profiles' validity and stability. Goodness-of-fit parameters from adjusted linear regression evaluated the individual variability captured by each approach. RESULTS: A k-means algorithm with 6 clusters best fit the data and identified the following taste perception profiles: Low All, High Bitter, High Umami, Low Bitter & Umami, High All But Bitter and High All But Umami. All profiles were valid and stable. Compared with total taste scores, taste perception profiles explained more variability in bitter and umami perception (adjusted R2: 0.19 vs. 0.63, respectively; 0.40 vs. 0.65, respectively) and were comparable for sweet, salt, and sour. In addition, taste perception profiles captured differential perceptions of each taste within individuals, whereas these patterns were lost with total taste scores. CONCLUSIONS: Among older adults with metabolic syndrome, taste perception profiles derived via data-driven clustering may provide a valuable approach to capture individual variability in perception of all 5 tastes and their collective influence on diet quality. This trial was registered at https://www.isrctn.com/ as ISRCTN89898870.
Authors: Andre G Dias; Karen M Eny; Moira Cockburn; Winnie Chiu; Daiva E Nielsen; Lisa Duizer; Ahmed El-Sohemy Journal: J Nutrigenet Nutrigenomics Date: 2015-08-01
Authors: Suzen M Moeller; Jill Reedy; Amy E Millen; L Beth Dixon; P K Newby; Katherine L Tucker; Susan M Krebs-Smith; Patricia M Guenther Journal: J Am Diet Assoc Date: 2007-07
Authors: Valerie B Duffy; Andrew C Davidson; Judith R Kidd; Kenneth K Kidd; William C Speed; Andrew J Pakstis; Danielle R Reed; Derek J Snyder; Linda M Bartoshuk Journal: Alcohol Clin Exp Res Date: 2004-11 Impact factor: 3.455
Authors: Shakeela N Jayasinghe; Rozanne Kruger; Daniel C I Walsh; Guojiao Cao; Stacey Rivers; Marilize Richter; Bernhard H Breier Journal: Nutrients Date: 2017-07-14 Impact factor: 5.717
Authors: Julie E Gervis; Rebeca Fernández-Carrión; Kenneth K H Chui; Jiantao Ma; Oscar Coltell; Jose V Sorli; Eva M Asensio; Carolina Ortega-Azorín; José A Pérez-Fidalgo; Olga Portolés; Alice H Lichtenstein; Dolores Corella Journal: Nutrients Date: 2021-12-29 Impact factor: 5.717