Machine learning-based predictive model for type 2 diabetes mellitus using genetic and clinical data

Type 2 diabetes mellitus (T2DM) is one of the leading metabolic conditions attributed to complex interactions between environmental and genetic factors. Around 90% of diabetes mellitus cases are T2DM, which is estimated to affect 9.9% of the world's population. Inheritance patterns have been studied extensively in T2DM, with the knowledge that the relative genetic risk of disease increases if relatives are affected. Genome-wide association and genomic linkage studies, identifying novel genetic risk loci in T2DM, have further strengthened genotypic associations with diabetic phenotypes. These studies often link whole exome-sequencing data with clinical data from electronic health records (EHRs) to formulate genotype–phenotype associations. However, there is still a growing need to integrate genetic data from these studies with EHR data to devise predictive models that improve risk stratification in T2DM. Traditional registry-based methods have major limitations to the interpretability of studies and may not be sufficiently powered to create predictive models. However, the use of EHRs from large population–based databases together with the rise in artificial intelligence and machine learning algorithms can better address these limitations. Neural networks govern the hierarchy of artificial intelligence, machine learning, and deep learning models. Tasks that generally require human input are automated and can carry out higher levels of pattern recognition using refined models or algorithms. In T2DM, ML-based predictive models are promising for risk stratification in clinical practice. Integrating genetic data from genome studies with clinical data can drive risk predictions of adverse complications and personalize medical treatment. In this section, we will discuss the role of genetics in T2DM and how genome studies can strengthen ML-predictive models in clinical practice.