Application of the deep learning algorithm in nutrition research – using serum pyridoxal 5′-phosphate as an example

Ma, Chaoran; Chen, Qipin; Mitchell, Diane C.; Na, Muzi; Tucker, Katherine L.; Gao, Xiang

doi:10.1186/s12937-022-00793-x

Table 2 R squares for pyridoxal 5′-phosphate prediction models, based on deep learning algorithm versus multivariable linear regression

From: Application of the deep learning algorithm in nutrition research – using serum pyridoxal 5′-phosphate as an example

	37 Food groups and supplement variables included^a				Further including non-dietary variables^b
	Deep learning algorithm		Multivariable linear regression		Deep learning algorithm		Multivariable linear regression
	Training	Test	Training	Test	Training	Test	Training	Test
All participants	0.46	0.41	0.21	0.15	0.43	0.47	0.25	0.18
Excluding users of vitamin B6 supplements	0.36	0.33	0.08	0.08	0.49	0.33	0.15	0.16
Including only users of vitamin B6 supplements	0.59	0.53	0.20	0.17	0.66	0.51	0.25	0.21
Including only variables identified by the stepwise model^d,e	0.45	0.41	0.21	0.15	0.52	0.38	0.25	0.18

^aVariables include energy intake, vitamin B6 supplement, citrus/melons/and berries, other fruits, fruit juice, dark green vegetables, tomatoes, other red and orange vegetables, potatoes, other starchy vegetables, other vegetables, beans and peas (vegetables), whole grains, refined grains, meat, cured meat, organ meat, poultry, seafood high in n-3 fatty acids, seafood low in n-3 fatty acids, eggs, soy products, nuts and seeds, beans and peas (proteins), milk, yogurt, cheese, oils, solid fats, added sugars, and alcoholic drinks
^bVariables in¹ and also age, sex, education, ethnicity, ratio of family income to poverty, adherence to physical activity guideline recommendations, smoking status, anti-hypertension medication use, cholesterol-lowering medication use, insulin treatment, anti-Diabetes medication use, systolic blood pressure, diastolic blood pressure, glucose, glycosylated hemoglobin, body mass index, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, total cholesterol, and C-reactive protein
^cVariables for 37-food groups, include added sugars, alcoholic drinks, cheese, milk, yogurt, fruit juice, other fruits, whole grains, oils, cured meat, legumes (proteins), nuts and seeds, poultry, seafood high in n-3 fatty acids, soy products, solid fats, legumes (vegetables), other vegetables, other red and orange vegetables, tomatoes, other starchy vegetables, and vitamin B6 supplement
^dVariables for the dietary and non-dietary model include age, sex, ratio of family income to poverty, smoking status, systolic blood pressure, cholesterol-lowering medication use, glucose, body mass index, high-density lipoprotein cholesterol, C-reactive protein, alcoholic drinks, mile, yogurt, fruit juice, other fruits, refined grains, whole grains, oils, cured meat, legumes (proteins), nuts and seeds, poultry, seafood high in n-3 fatty acids, soy products, solid fats, legumes (vegetables), other vegetables, other red and orange vegetables, tomatoes, other starchy vegetables, and vitamin B6 supplement

Back to article page

ISSN: 1475-2891

Contact us

General enquiries: journalsubmissions@springernature.com

Nutrition Journal

Contact us