The need for speed: how drug target discovery can drive change in precision medicine

Employee Perspective: Cyclica’s Marketing and Communications Specialist, Rebecca Woelfle

In today’s drug development landscape, it costs approximately $2.5 billion USD and 10 years to bring a drug from the discovery stage to market. Moreover, only 1 in 10 drugs that enter phase 1 clinical trials go on to receive regulatory approval. These trends underscore the need for a more efficient and sustainable path to drug commercialization. A promising area of drug target discovery that may expedite this (~10-year) process is the application of artificial intelligence (AI) in precision medicine (PM), an approach that aims to personalize treatment options for patients by considering their unique genes and environment. Ultimately, the benefits of applying AI have potential for providing major advances in our understanding of disease mechanisms. So, how is AI being used to develop personalized drug treatments?

AI-based tools that are being leveraged in PM and the drug discovery process include computational modelling and machine-learning platforms to find potential drug targets. A hallmark of disease is heterogeneity, but with computational modelling, scientists can map out complex biological interactions to study this aspect of disease. Specifically, computational modelling may be used to construct a detailed molecular profile for a patient, thereby identifying individuals that have an elevated genetic risk for disease or matching a patient to a particular type of drug therapy. For example, Cyclica is working with PrecisionLife to utilize computational modelling for data-driven drug discovery. PrecisionLife has created insights from patient datasets that reveal the underlying disease architecture and novel drug targets for different diseases, including severe COVID-19 infections. Using Cyclica’s platforms, trained to unlock the entire proteome universe, multiple drug target interactions can be evaluated simultaneously, enabling a more efficient drug design process, all while exploring low-data targets.

Several drug discovery fields are using AI to interpret large datasets and develop tailored drug treatments, from neuroscience to oncology. For example, machine learning algorithms are being used to analyze brain imaging data to find biomarkers for depression, helping to predict a patient’s response to certain antidepressant drugs. Through DNA sequencing, scientists can spot novel genes associated with severe types of epilepsy. In some cases, these genes can be potential drug targets and could help patients living with seizures who are not candidates for brain surgery. Applying AI also aids in the development of personalized cancer treatments including cell therapies and immunotherapies, by revealing cancer subtypes and biomarkers. This molecular ‘fingerprint’ of a patient, generated by AI, can be used to design drugs that activate a patient’s own immune system to specifically kill cancer cells. These targeted drugs may be more efficacious for treating an individual’s specific disease, as well as being safer with fewer adverse side effects. 

While the application of AI has its benefits and is a promising field, limitations do exist. With PM, there are ethical concerns in obtaining informed consent from patients who may not be ‘science literate’ and risks to keeping patient data private. Private patient health data is increasingly valuable and proper security safeguards should be in place to prevent data breaches. These safeguards are of paramount concern for the technologies that consume this sensitive data. Even though the current state of data collection and management in healthcare is complex, these concerns and risks are being acknowledged and substantial progress has been made in ensuring data management is done safely.

As AI data is collected from groups of individuals, analyses often do not consider the vast socioeconomic variables that have historically contributed to disease development and treatment. On the topic of data, another issue that needs addressing is that the data being collected and analyzed for healthcare is largely based on males with European lineage, resulting in skewed predictions and/or ones that are not applicable to entire communities. Cyclica is very committed to addressing the need for change for the historical and current ways in which data is collected, analyzed and used. For more info on this, see our recent Molecule to Medicine Series episode that discusses data bias in AI in healthcare.

Even though it is a nascent technology and applying AI in PM has many challenges, these advancements have exciting potential to expedite the drug development process, reduce the burden of disease and bring therapies faster to patients in need.

Reference

Paul, Debleena, et al. "Artificial intelligence in drug discovery and development." Drug Discovery Today 26.1 (2021): 80.

—

Rebecca is the Marketing and Communications Specialist at Cyclica. With an MSc in neuroendocrinology and a BSc specialization in biology, she developed a keen interest in ligand receptor interactions while investigating her Master's thesis. Having worked for 2 years at a Boston-based biotech and 5 years in science comms at the Ontario Brain Institute’s epilepsy research program, EpLink, she offers a unique perspective of bridging the gap between drug discovery and translatable science to facilitate a stronger understanding of scientific innovation to the broader public.