My doctor is an algorithm

FROM tumors to atopic dermatitis, artificial intelligence (AI) will soon take the helm in the discovery and development of drugs for the most diverse diseases, more effective and safer products, which could come into use much faster than traditional drugs, reducing by about one third the development times. In fact, there are an increasing number of pharmaceutical companies in the world that use IT platforms based on artificial intelligence. Many have already started trials on patients, for example the London startup BenevolentAI has almost concluded a phase II trial for “BEN-2293”, a powerful lotion for atopic dermatitis; the results are expected shortly. The company began its pioneering use of artificial intelligence at the onset of the Covid emergency, using its AI-based computer brains to select the best therapeutic targets of the coronavirus, and choose quickly and completely automatically, among the drugs already in use with other medical indications, the best ones to reposition against Covid.

The advantages of this innovative method of drug development based on artificial intelligence are many, explains Michele Vendruscolo, a chemist and pharmacologist at the University of Cambridge who, with his research group, is using AI to develop new treatments for Parkinson’s and Alzheimer’s with excellent preliminary results. First of all, Vendruscolo points out, AI drastically cuts the development time of molecules: “On average, the use of AI shortens the pre-clinical phase, including the discovery of therapeutic targets, the identification of initial molecules, and optimization processes (dosage, chemical formulation, etc.). So the acceleration could be estimated at around 3-5 years, out of the total 12-15 years needed today to bring a drug to the market”.

Artificial intelligence revolutionizes the very way of discovering new active ingredients. It does this in various ways, explains Vendruscolo: for example, it exploits the data of enormous libraries of molecules to identify in a very short time an active ingredient that is like a “key” that fits perfectly with the “molecular lock” represented by the therapeutic target. The more the molecule fits well to the target without unwanted secondary interactions, the more pharmacologically effective and safe it will be.
But that’s not all, AI can also forge new active ingredients modeling them exactly on the chosen therapeutic target which is then used as a “mould” to shape the new drug. Even more fascinating, AI can create new active ingredients from scratch, taking cues from molecules that exist in nature. Furthermore, underlines Vendruscolo, AI is ultra-fast: for example, if millions of compounds can be tested with traditional methods of screening libraries of molecules, with AI we easily reach billions in less time. Furthermore, AI can quickly optimize other properties required in pharmacology, for example parameters such as toxicity and absorption, aspects that with the traditional way of research and development of drugs are usually discovered in retrospect, often nullifying years of research on initially promising substances, but which end up being rejected (today only one compound out of 10,000 of those studied becomes a drug; with AI, this proportion could rise to one out of 1,000).

The use of artificial intelligence in drug development is already paying off. With the many “AI-native” drugs currently undergoing clinical trials, “the first approvals by the regulatory authorities (Fda, EMA, etc.) could already take place within 3-5 years”, estimates Vendruscolo. For example, the British start-up Exscientia boasts the distinction of having created the first drug made entirely by AI: designed for obsessive-compulsive disorder, the molecule, DSP-1181, was developed and passed preclinical tests in less than a year (traditionally it takes 5-7 years for a new molecule to reach a clinical trial) and is currently in phase I clinical trials. The start-up also has two other molecules in phase I clinical trials: DSP-0038 for Alzheimer’s psychosis and EXS-21546 for advanced solid tumors; moreover, the company has just validated the use of an immuno-oncological test to select patients who will surely respond to its EXS-21546 molecule. And, again, keep an eye on Nimbus Therapeutics, which announced it will begin an advanced (Phase II) clinical trial this year for its psoriasis drug, NDI-034858, a once-daily pill. The Boston-based company is also in trials of an immuno-oncology drug, NDI-101150. New York-based Insilico Medicine just announced that the US Food and Drug Administration (Fda) has granted Orphan Drug Designation to its AI-generated molecule, INS018_055, for the treatment of idiopathic pulmonary fibrosis, a chronic disease that causes progressive decline and irreversible loss of lung function and represents an important unmet medical need worldwide. INS018_055 is a small molecule inhibitor discovered by the Pharma.AI generative AI platform. The platform has been used both to identify a new therapeutic target and to generate new small molecules active against it, thus completely new drugs for this disease. Two phase I clinical trials in New Zealand and China were completed in 2022. The data indicate that the molecule is safe and well tolerated. Insilico plans to launch a phase II global multicenter clinical trial in recent months, says Alex Zhavoronkov, founder and CEO of Insilico Medicine. The FDA’s orphan drug designation, he says, will facilitate further development and commercialization of INS018_055 once clinical trials are completed. Insilico is working to discover new therapeutic targets and develop entirely new drugs for cancer, fibrosis, central nervous system, infectious, autoimmune and age-related diseases.

The new frontiers of pathological anatomy: Artificial Intelligence for ever more precise diagnoses

by Irma D’Aria

05 November 2021

In the future, new antibiotics could also be developed by AI: as stated in the magazine Nature Biotechnology, the “ProGen” system developed by Salesforce Research, was able to generate artificial antimicrobials out of thin air that, while significantly different from any known natural protein, function like those found in nature. They have been successfully tested, for now in test tubes, against the Escherichia coli bacterium.

The list of companies that use a platform based on artificial intelligence is still long and destined to grow exponentially in the coming years: in the meantime, we are looking with hope at various molecules also against rare diseases such as GM2 gangliositosis (a metabolic disease), just to name a few. “Our results with the use of AI for Alzheimer’s and Parkinson’s are also very promising – Vendruscolo anticipates – however it is still too early to say whether these molecules will become drugs. In any case from now on – concludes the expert – AI it will be an essential tool, there will no longer be room for pharmaceutical companies that do not use it because, by speeding up various steps of the drug development process, the “AI-native” companies have a competitive advantage over the others”.