For several years we have been hearing about the potential of Artificial Intelligence (AI) to improve traditional drug discovery and development. In the last two years, clinical trials have begun. The UK’s Exscientia made headlines last April by announcing the start of a Phase 1 clinical trial for a drug it designed using AI for an established protein target. Recursion Pharmaceuticals in Utah uses AI to find new uses for the drugs owned by other companies.
Insilico Medicine has now announced the crucial next step: the start of the world’s first Phase 1 clinical trial of a drug developed from scratch using AI. Its end-to-end platform applies AI to biology for target discovery, and to chemistry for drug design. Intriguingly, the company believes the drug may have anti-aging properties as well as its immediate therapeutic effect.
The disease this new drug is designed to tackle is Idiopathic Pulmonary Fibrosis (IPF). Its causes are unknown (hence “idiopathic”), and it stiffens the lung tissues of older people, and can eventually kill them. It affects five million people each year.
Clinical trial phases
A Phase 1 trial is when a drug is first tested in humans at the dosage level required for therapeutic effect. It tests safety, not clinical effectiveness. Before that, drugs go through preclinical trials on animals, and Phase 0 trials in sub-therapeutic doses. In Phase 2 and Phase 3 trials, a drug is tested for effectiveness, first with a small sample of a few hundred people, and then with a larger sample of thousands.
In the case of this Insilico drug, the Phase 1 trial will involve 80 volunteers, half of whom will be given increasing dosages. The trial will determine how the human body impacts the drug molecule, a process known as pharmacokinetics.
The breakthrough is not only important for sufferers of IPF. It adds to the growing body of evidence that AI is an invaluable tool in developing drugs. The complexity of human biology makes the development of drugs extremely challenging. As long ago as the 1980s, scientists noticed that the cost of pharmaceutical R&D was doubling every decade or so. This was later described as Eroom’s Law, since it was the reverse of the exponential improvement in computing known as Moore’s Law.
The Big Bang in AI
2022 is the tenth anniversary year of the Big Bang in AI, when Geoff Hinton finally got a machine learning algorithm called backpropagation to work, and achieved a famous breakthrough in image recognition. Machine learning, and especially deep learning, is responsible for the explosion of interest in AI this decade, and the minor miracles it has produced, such as interactive maps, near-omniscient search engines, and machines that can write, draw, and create music.
At the time of the Big Bang, Insilico’s founder Alex Zhavoronkov had already worked in a company manufacturing GPUs, a type of chip which has helped enable the rise of deep learning, and he was starting to explore the potential for machine learning within drug development.
The development process
The drug now entering Phase 1 trials, which goes by the somewhat ungainly name ISM001-055, is the result of a three-part, AI-powered process. The first part is called PandaOmics, in which Insilico used natural language processing AI to trawl through massive medical data sets, looking for a protein which could be causing Fibrosis. This protein is called the target.
Fibrosis is a disease closely associated with aging, and Insilico’s deep neural networks were trained on age and different types of fibrosis to identify a range of targets, analysing data from millions of data files, including patents, research publications, grants, and databases of clinical trials.
Once the target was identified, Insilico used Chemistry42, a type of AI called a Generative Adversarial Network (GAN), to identify a molecule which could become an effective drug against the disease. GANs pit two AIs against each other: one generating suggestions, and the other critiquing those suggestions in an evolutionary process.
The third part of the process is Inclinico, in which the company used AI to monitor the effects of the drug in eight volunteers in Australia. This whole three-part process took 30 months and £2.6m, which is a tiny fraction of the traditional approach to drug development, which typically costs billions of dollars, takes a decade or so, and suffers from a 90% failure rate.
In addition to its work with fibrosis, Insilico is researching innovative drugs for cancer, immunity, central nervous system (CNS) diseases and age-related diseases. In oncology, it has a strategic collaboration with Fosun Pharmaceuticals, a major Chinese drug development and manufacturing group.
In the fight against Covid-19, Moderna and BioNTech, which developed the two leading mRNA (messenger RNA) vaccines, were able to get their products into arms much quicker thanks to the efficiency gains from using AI. Insilico is leading the charge to achieve even more dramatic results in drug development.