The old and the new: combining artificial intelligence with traditional medicine

The history of medicine is a fascinating one – traditional treatments that long precede the modern drugs we are familiar with have been documented across many societies, from the ancient Chinese to the Ayurvedic practices in India, ancient Egyptians, and indigenous tribes across the Americas. Indeed, traditional medicine derived from herbs, minerals, or animal products has been used for millennia to treat illnesses, and some have since been adapted into purer forms and become pillars of modern healthcare. Willow bark, used for pain relief and fever, forms the basis of today’s aspirin. The anti-malarial artemisinin was originally derived from sweet wormwood used in traditional Chinese medicine. It is estimated by the World Health Organization that approximately 40 per cent of pharmaceutical products today have origins in traditional medicine.

While these traditional remedies have stood the test of time, they are not without limitations. There is a lack of standardization and regulation in key aspects of medicine like appropriate prescribing, dosing, safety, and evaluation of their efficacy. However, the toolbox of modern technologies available for drug research is growing, and there is interest in combining the wisdom of traditional medicine with modern techniques to accelerate drug development. In particular, the rapid advancement of artificial intelligence (AI) in recent years has massively boosted the speed at which drug discovery, development, and optimization can occur. As a tool that can be incorporated into machine learning, pharmacology, bioinformatics, and medicine, it can mine vast sums of knowledge passed down and accumulated by those practicing traditional medicine and take a systematic approach to identifying novel therapeutic molecules.

AI could replace people… but not in that way

Humans are complex. One diseased organ can have cascading effects on the rest of the body, which can be confusing when trying to pinpoint druggable targets. Being able to condense anatomical and biochemical knowledge about the human body using AI could dramatically reduce the effort needed to find targets of interest. Context-Oriented Directed Associations (CODA) is one example being developed by researchers at the Bio-Synergy Research Center in South Korea. CODA draws on publicly available literature and data about human biochemical pathways and could be utilized to screen for potential bioactive molecules derived from natural products used in traditional medicine, while also eliminating candidates with potential toxic effects. The intention of the development team is for CODA to generate hypotheses about how specific compounds from plants or animal products could provide therapeutic value in a variety of different diseases and may also reduce the amount of superfluous testing in cells, animals, and humans before reaching an optimized medicine.

Case study: the potential of venomics

One fascinating area where AI is already making strides is in the field of venomics—the study of the bioactive compounds found in venomous animals. Indigenous communities have long understood the medicinal potential of venomous animals. There are multiple examples of venoms being used to treat bites from said venomous animals, pain, and inflammation, and today it is recognized that venoms hold a diverse array of bioactive compounds that could help address some current gaps in medicine. However, much like other traditional remedies, crude venoms cannot be used medicinally today; compounds in venom must be screened systematically to develop libraries of potentially useful drugs.

The emerging availability of artificial intelligence (AI) tools presents new opportunities to expedite previously painstakingly slow protein characterization. For example, machine learning algorithms can now screen thousands of venom proteins, many of which are already documented in public databases, to predict their potential antibiotic activity. By identifying antimicrobial compounds in venom that work through novel mechanisms, AI could help develop the next generation of antibiotics that are effective against resistant bacterial strains. Researchers from the University of Pennsylvania have already published preliminary findings showing that machine learning can dramatically speed up the process of screening venom proteins for potential antimicrobial properties. The deep learning model they developed, antibiotic peptide de-extinction (APEX), predicted almost four hundred proteins or peptides across cone snail, snake, and spider venoms that are distinct from known antimicrobials. This is a particularly timely area of research as antibiotic resistance is becoming an ever-growing global health threat.

A companion to empirical science

Traditional medicine will always have its place in healthcare, particularly in regions of the world where access to modern medicine is limited, but that isn’t to say that it cannot be further improved. The incorporation of AI into research will allow scientists to look back through history through a lens that enhances our understanding of why and how these traditional medicines have worked for hundreds of years.