From data to drugs: Harnessing machine learning in drug discovery - A review

Abstract

Drug development is a rigorous process essential for improving patient outcomes. However, this complex endeavour requires significant investment and time. The integration of Machine Learning (ML) techniques in drug discovery can revolutionize the field by efficiently processing large amounts of data and accelerating the identification and development of potential drug candidates. This review provides an overview of the impact of ML on the various stages of drug development, from drug design to clinical trials.
Recently, the availability of high-quality databases and the surge in data digitalization has promoted the development of several ML algorithms, which have proved to be effective in classifying outcomes based on multivariate relationships. Deep learning (DL) architectures such as feedforward networks, Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNNs), and long short-term memory (LSTM) neural networks have gained popularity due to their ability to understand complex patterns by mimicking the human brain. In recent times, DL methods have been proven superior when compared to traditional ML algorithms.
ML may play a vital role in virtual screening, de novo drug design and drug repurposing. Virtual screening methods can rapidly screen large chemical libraries and identify promising candidates for further investigation. De-novo drug design involves the use of ML-based generative models to produce new chemical structures with desired properties. Drug repurposing aims to identify new therapeutic uses for existing drugs. Additionally, ML can improve the efficiency of clinical trials by addressing challenges related to patient enrolment, study design, and phase transition.
The integration of ML with high-quality datasets can significantly improve drug development process, thereby increasing efficiency and success rates. However, it is important to address issues related to data quality, preprocessing bias, molecular representation, and interpretation of results. Harnessing the power of AI can accelerate drug development, ultimately benefiting patients and the healthcare industry as a whole.