From Tortoise to Hare, How AI Aims to Speed up Drug Development

Alexander Flemming’s accidental discovery of penicillin exists in stark contrast to the typical processes and timelines associated with drug discovery. Most drugs, on the contrary, are not accidentally discovered and are developed over long timelines ranging from 5 to 15 years, costing many millions of dollars. Broadly speaking, the development process can be divided into 2 phases: the preclinical phase and clinical phase.  While most people are somewhat aware of the processes involved in the clinical phase of drug development (clinical trials) they are often less familiar with the work that is completed in preparation for those studies. In the earliest stage of drug development, it can take evaluating between 200,000 and 1 million different molecules to identify two or three that might have a therapeutic effect for someone with a particular disease. This process has a shockingly low success rate and requires the spending of significant resources and time. When applied to the preclinical drug development process, increased capabilities in data processing and advances in Artificial Intelligence (AI) and machine learning (ML) have the potential to drastically reduce inefficiencies in testing molecules in the initial research and development phase.

AI Is Not Robot Overlords

The phrase “AI” might bring to mind a dystopian future run by robots but, while we’re not quite there yet (hopefully, fingers crossed), we can thank exponential increases in computing power for revolutionizing AI and its applications. AI refers to the ability for computers or machines to execute processes usually attributed to the human mind and is an application that relies on machine learning (ML) and deep learning (DL) techniques, among others. Although the phrases AI, ML, and DL are sometimes used interchangeably, these concepts constitute subsets of each other. ML is a subset of AI that includes algorithms and software that are trained by humans and improve over time. It can be further divided into supervised and unsupervised methods. Supervised methods involve algorithms learning from data that has already been organized into categories and can be used to predict new items that might fit into a category. On the other hand, unsupervised methods involve algorithms trained on uncategorized data and can be used to identify patterns within a data set. Netflix and, by association, supervised ML have become a larger part of most people’s lives recently. Content on Netflix can be defined by various parameters including genre, actors, and length of episode or movie. Machine learning can be leveraged to define a user’s viewing preferences (Comedy, Adam Sandler, 2 hours long) and then recommend new content that best fits within those previously established parameters. DL takes ML learning a step further with multiple algorithms cooperating to review large data sets, identify certain characteristics, and then organize and identify which data points belong to the same category, all without human input. While Netflix leverages machine learning to suggest new original content that you might enjoy, AI and its constituents have the potential to drastically impact how the pharmaceutical industry conducts its business.

Traditional Preclinical Testing is Becoming Outdated

Preclinical testing is used to estimate how humans will end up responding to a potential drug and commonly involves in vitro and in vivo testing, while in silico (Latin for “within the glass” (test tubes and petri dishes), “within the living”, and “in silicon” (the silicon found in computer chips), respectively) testing has also become an increasingly common part of the process. In vitro experiments can be conducted on controlled cell cultures intended to mimic the states of diseased cells within a person and can provide initial insight into the potential toxicity of the compound under investigation, for example. These experiments allow for strict controls of the testing environment, but the responses elicited by the compound often lack real world applicability as isolated cells respond differently than those that function as part of a complete animal. The next stage of experiments is referred to as in vivo experiments and are tests that are often conducted in a combination of animals that may include, but are not limited to, rats, mice, pigs, and non-human primates (ex. chimpanzees). Historically speaking, the negative and positive results of testing drugs in animals have been used as justification to discontinue or proceed with additional experiments, but in general, responses in animals do not generalize to people due to genetic difference between species. Despite the best efforts of scientists to identify and test a molecule, there is a high likelihood that it may be ineffective or harmful and will be abandoned as a potential new drug. A lack of generalizability of animal models to humans and advances in computer technology and data collection have, in combination, contributed to an increased interest in the potential for experiments to be conducted in a virtual environment – in silico.

Using AI/ML to Enhance Traditional Processes

In silico testing leverages computer processing abilities and is a new method that precludes both in vitro and in vivo experiments in preclinical development. Prior to preclinical testing, groups of scientists must collaborate to: identify a target that contributes to a disease, understand its chemical properties and how they may influence interactions with a drug, establish a library of compounds that have the potential to interact with the target, and then choose the compounds that have the greatest likelihood of positively affecting the target. MagicMed is fortunate that research in the areas of mental health and psychedelics have already identified the potential targets that psilocybin and other psychedelics interact with. The identification of these targets is hugely beneficial as it helps to shorten the drug development timeline significantly. While research and experiments serve as the primary source for identifying the causative ages (the targets) that underpin the development of certain diseases, ML has significant applications assisting with the latter steps necessary for identifying a strong drug candidate.

That is not to say that the previous in vitro experiments do not have applications today. In vitro experiments have produced vast amounts of data that sought to investigate the interactions between millions of compounds from various libraries against different disease-causing targets. Any compounds that interacted with the target serve as a starting point for training ML algorithms to recognize certain chemical structures (parameters) that favor interactions with the target and predict somewhat similar, but new molecules. This process is iterative and newly predicted molecules can be fed back into the ML algorithms to assist with additional learning and a greater ability to predict new molecules. ML can help narrow the number of molecules that require further consideration down from tens of millions to thousands. Although the process still relies on the results of in vitro experiments and tens of thousands of molecules is a lot to be tested, ML leverages knowledge of previously successful compounds to help predict new molecules that might have enhanced pharmaceutical effects. A significant reduction in the number of molecules that require testing saves time, money and resources from being spent and wasted investigating molecules that are unlikely to succeed. MagicMed is optimistic that leveraging AI now, and as it continues to develop, will be critical in both shortening drug development timelines and getting new, life-saving drugs to people sooner.

Disclaimer

The above article is sponsored content. CannabisFN.com and CFN Media, have been hired to create awareness. Please follow the link below to view our full disclosure outlining our compensation: http://www.cannabisfn.com/legal-disclaimer/

This article was published by CFN Enterprises Inc. (OTCQB: CNFN), owner and operator of CFN Media, the industry’s leading agency and digital financial media network dedicated to the burgeoning CBD and legal cannabis industries. Call +1 (833) 420-CNFN for more information.

About Liam Harrison