Application of Machine Learning Techniques for study of drug interactions using clinical parameters for Creutzfeldt-Jakob disease

摘要

Drug discovery is one of the emerging areas of research, in which the association between drugs and diseases are studied to infer the effect of drug induced to patients having similar clinical parameters. As per the regulations of statutory organizations, clinical trials are to be made after the drug proceeds to substantial improvement under variety of guidelines. In some diseases like Creutzfeldt-Jakob disease which is a degenerative brain disorder progresses more rapidly than Alzheimer's disease, it is very much indeed to study the association between drug and its impact over clinical parameter. Even though CJD is a rare occurring disease in young age, but it is more progressing in the elders with asymptotically that leads to dementia and ultimately death with a very short living span of less than a year. Since the clinical parameters to the disease are very less, computational algorithms can be applied on the available clinical parameters with drug labels. Due to the advent of simple, yet powerful machine learning algorithms the modeling of the drug interactions with varying clinical parameters became quite imperative. In this work, the application of various machine learning algorithms with parameter tuning facilities were analyzed to facilitate the study of the drug over clinical parameters. The study ends up in developing a machine learning model, which will suggest the drugs when clinical parameters are inputted. This work analyses different machine learning algorithms such as Logistic Regression, KNN, DTC, RFC, SVC, XG Boost with an ability to tune the parameter with relevance to the available dataset. The designed models are validated against k-fold stratified cross-validation to construe the better classification model. The results show that, Random forest classifier outperform with XG Boost classification algorithm with a mean accuracy precision of 98.39% while the later provides a mean accuracy precision of 97.96%.