An InDepth Analysis of AI Techniques for Predicting Genetic Disorders

The integration of artificial intelligence (AI) with predictive methods and a deeper understanding of genetic diseases has the potential to greatly simplify and expedite the diagnosis of rare genetic diseases. Estimates indicate that over 5 million babies are born with genetic diseases annually in India, imposing a significant burden on the economic and social fabric of society. The identification of rare genetic disorders requires advanced genetic testing and specialized expertise due to their intricate nature. In this study, various machine-learning models were employed to analyze datasets related to genetic disorders and their subclasses. The first step in the analysis involved preprocessing the dataset by checking for any missing values. Subsequently, these absent data points were visually depicted within various categories like genetic disorders, subclasses of genetic disorders, samples of symptoms, inherited genes from both parents, birth defects, and others, with the aim of recognizing any discernible patterns or trends. After preprocessing, standardized features were selected and used as input for the machine learning models. The models used in this study included Deep Learning-Powered Advanced Genome Disorder Prediction Model, Random Forest, XG Boost, Baseline Gradient Boosting Classifier, Gradient Boosting with Hyperparameter Tuning and Cross Validation, Ensemble Learning with hyperparameter Tuning and Cross Validation. The predicted classes are compared against the true classes to categorize the outcomes into four groups using confusion matrix. The results of the analysis showed that Deep Learning-Powered Advanced Genome Disorder Prediction Model and XG Boost achieved the highest accuracy of 91% for genetic disorders. However random forest and Gradient Boosting Classifier achieved accuracy of 88% for genetic disorder subclasses.