Developing Algorithms for Predicting Gene-Disease Associations

It is evident that identifying gene-disease relationships is essential to elucidate disease genomics and progression as well as aiding in developing targeted therapies. This work aims at training and testing various machine learning approaches for predicting these relations, as well as considering various datasets, and their integration involving multi-omics data. The process of data collection has been carried out from Genomic, Transcriptomic, and Clinical data stored in various data repositories and then preprocessed for feature selection and fine-tuning of the model along with the development of Supervised, Unsupervised, and Integrated Models. The models were built and tested using cross-validated data sets; model efficiency was assessed with respect to accuracy, precision, recall and F1-score. The findings showed that the models from the second group of hybrids obtained a systematically higher accuracy than models from the first and subsequent groups, especially when analyzing multiple types of ‘omics data at once. Similar deployments on entirely separate datasets ensured the model’s reliability and portability across different contexts. Furthermore, interpretability features and visualization modules were included into the software to improve assessment and interactivity for users and scholars. This research will be beneficial for the field of genomics since it proposes tools that will help in the better understanding of disease and the identification of genes involved in particular diseases, making it possible to start the process of targeting important genes in order to develop cures for such diseases. The study results preaciously highlight the significance of applying integrated strategies of data and analytic methodology to improving the accuracy of the prediction and implementing the personalized medicine concept.