Computational model for vitamin D deficiency using hair mineral analysis

Area of Science:

• Biochemistry
• Computational Biology
• Public Health

Background:

• Vitamin D deficiency is widespread in the Arabian Gulf, particularly affecting women.
• This deficiency is linked to underlying mineral imbalances, necessitating accurate assessment methods.
• Traditional tests for mineral status can be costly and less informative.

Purpose of the Study:

• To develop a predictive model for vitamin D deficiency using hair mineral analysis (HMA).
• To assess the association between mineral levels and vitamin D status in Kuwaiti women.
• To introduce a novel hybrid optimization algorithm for feature selection in predictive modeling.

Main Methods:

• Hair mineral analysis (HMA) was performed on 118 Kuwaiti women to determine mineral and vitamin D levels.
• A hybrid optimization algorithm (Bat Algorithm with Genetic Algorithm mutation) was developed for feature selection.
• Data balancing techniques (under-sampling, over-sampling, SMOTE) and an AdaBoost classifier were employed for prediction.

Main Results:

• The study successfully built a computerized model to predict vitamin D deficiency.
• The hybrid optimization algorithm improved feature selection for the predictive model.
• The AdaBoost classifier demonstrated good performance in detecting vitamin D deficiency using mineral data.

Conclusions:

• Hair mineral analysis is a viable and informative tool for assessing mineral imbalances associated with vitamin D deficiency.
• The developed predictive model shows promise for early detection of vitamin D deficiency in the target population.
• This approach offers a cost-effective and accurate method for identifying hidden mineral imbalances linked to vitamin D status.