Applications of artificial intelligence in drug discovery for neurological diseases

Neurological disease encompasses over 1000 disorders, exacts a massive human health and financial toll as well as being a story of extremes. At one end are diseases that are complex and heterogeneous affecting millions, while at the other there are monogenic and rare diseases, with a handful of individuals. What are absent are drugs that can treat or cure the disease. Discovering these is challenging, held back by extreme costs to develop them or in some cases by the limited understanding of the diseases. After decades of drug discovery research there is now considerable data available which can be used to help develop novel compounds more strategically. This includes high throughput screening data with targets, crystal structures of proteins implicated in neurological diseases and adjacent data such as properties of molecules like blood brain barrier permeability as well as an array of in vitro and in vivo toxicity endpoints valuable for any drug targeting the central nervous system. While computational tools have been developing and applied to neurological diseases for decades, we are now in the age of machine learning and artificial intelligence (AI). This promises the potential to expedite the identification and discovery of new molecules. Whether by using individual computational techniques or complex end-to-end approaches, scientists can narrow the molecules they make and test as well as study more targets or diseases which might have been out of reach previously. This review highlights the many different applications of AI potentially enabling new discoveries and treatments for neurological diseases.