Pharmaceutical companies have the opportunity to capitalize on Artificial Intelligence (AI). AI refers to the ability of computer systems to perform tasks that would normally require human intelligence. Drug companies continuously analyze thousands of compounds, seeking candidates of therapeutic value. The process can be time-consuming and costly: 1–6 years for preclinical development, costing about $1 billion, and 6–12 years for clinical development before FDA approval, costing about $1.4 billion, with a cumulative probability of roughly 8 percent of getting a development candidate to approval. In an effort to speed up the process, improve efficiency, and reduce healthcare costs, several pharmaceutical companies have implemented systems biology, computational modelling, and AI, albeit with mixed success.
AI Saves Half the Time and Money for Drug Discovery:
In February 2017, BERG Health, a 6-year-old startup backed by Silicon Valley billionaire Carl Berg, made headlines when they announced that their AI platform had selected a drug candidate for rare brain cancers that has now entered clinical trials as monotherapy (i.e., stand-alone treatment). The drug candidate, BPM 31510-IV, was guided through early development by the AI-based BERG Interrogative Biology Platform. The platform analyzed patient data from thousands of cancer patients to build a in silico disease model and suggest possible drug treatments.
“We’ve essentially reversed the scientific method,” BERG’s President and Co-Founder, Niven Narain said, “Instead of a preconceived hypothesis that leads us to do experiments and generate a particular type of data, we allowed the biological data from the patients to lead us to the hypotheses.”
Legally required regulatory testing necessitates that all drug candidates pass through animal tests; this requirement is unlikely to change in the near future. However, the potential to select a drug candidate entirely from human data may not only expedite the drug development process but also reduce the attrition rate of drug candidates, thus decreasing overall cost. Narain claims that his AI platform took half the time and half the money of traditional methods.
Drug Target Identification and Toxicity Prediction:
Despite the daunting drug development process, FDA approved drugs are frequently withdrawn from markets. This is primarily due to their side effects or toxicities, which is a fallout of polypharmacology of drugs. Polypharmacology is the interaction of drug molecules with multiple targets that, besides the intended therapeutic effect, can result in side effects. Two companies already leveraging AI and Big Data for the purpose of drug target identification and side effect prediction are Cyclica Inc., a Canadian startup, and One Three Biotech, a spin-off of Weill Cornell Medical College in New York.
Cyclica Inc., founded in 2010, employs a suite of computational algorithms. Their predictive analytics platform, Ligand Express™, combines proteome-docking, ligand effect prediction, and systems biology and drug-protein interactome analysis, to evaluate and compare small molecules, and predict how each drug will interact with the human body (i.e., human proteome). Cyclica’s Ligand Express platform is used and validated through third-party organizations, allowing clients to anticipate a drug candidate’s side effects prior to clinical trials, thereby enabling more informed R&D investment decisions.
Working in the same space as Cyclica is One Three Biotech, founded by Neel Madhukar. The company’s AI-based platform was developed by Madhukar as part of his graduate work with computational biologist Olivier Elemento. When speaking to R&D research company, PreScouter, Elemento and Madhukar explained that their AI platform, BANDIT, helped Oncoceutics Inc. predict the target for ONC201, a first-in-class small molecule that is being evaluated in 5 clinical trials. Their results were later confirmed through in vitro assays, and the physiological relevance of the predicted interaction was established by analyzing clinical samples.
“The predictions are based on patient data. We are teaching AI to connect small molecules with their targets using curated data from public databases and hope to gain access to databases from Big Pharma to improve connection and correlation predictions,” Elemento said. Referring to their closest competitor, Madhukar claims that “While Cyclica may have higher computational capabilities, One Three Biotech’s AI platform has greater accuracy (~90%) than Cyclica’s platform (~70%).”
Cheaper Drug Development, Cheaper Healthcare:
Over the last five years, AI has made headway in various aspects of drug development and is being used by biotech startups and mid-size drug discovery companies. However, AI is yet to be embraced by Big Pharma corporations. As the adage goes, “the proof of the pudding is in the eating”, similarly, the long-term drug safety and clinical efficacy of AI-selected and developed drug candidates remain to be assessed.
AI has already revolutionized other sectors and has the potential to do so in the pharmaceutical industry by increasing drug development efficiency and reducing drug attrition, thereby reducing drug development cost and offering the promise of cheaper healthcare.
