Machine learning for genomic profiling and drug discovery in personalised lung cancer therapeutics

Abstract

Lung cancer is a prevalent and lethal malignancy characterised by the uncontrolled growth of abnormal cells in lung tissues, often leading to functional impairment and metastasis. With approximately 2 million new cases and 1.8 million deaths annually, it is a significant contributor to cancer mortality, and projections suggest a substantial increase, with an estimated 3.8 million new cases by 2050 and 3.2 million deaths. So, early diagnosis and rapid drug development strategies are required, while genomics is used to enhance its sequential pattern for better, precise, and personalised medicine. Today, Machine Learning (ML) is transforming modern genomics and drug designing by analysing large genomics datasets to identify genomic sequential patterns and uncover mutations for targeted and better treatments and expedites drug discovery by modelling compound interactions with identified biological targets. The PubMed search was performed to identify relevant publications of the last 10 years and reviewed them critically. ML algorithms like Random Forests, Gradient Boosting, Deep Belief Networks, Autoencoders, Support Vector Machines, Convolutional Neural Networks, and Recurrent Neural Networks are extensively used in modern genomics, while Reinforcement Learning, DNN, GANs and GNNs are used for optimised and personalised drug discovery. ML algorithms face data scarcity and interpretability issues, challenging accuracy and integration with experimental validation. Lung cancer therapeutics are experiencing rapid advancements with the integration of ML, showcasing their potential to solve cancer-related problems with remarkable accuracy, often exceeding 95% in specific applications. However, more optimisation is needed to integrate Artificial Intelligence (AI) efficiently to deal with data heterogeneity and for clinical validation.

Topics & Concepts