AI Process Discovers, Tests Fibrosis Drug in 18 Months

(NIH.gov)

25 Feb. 2021. A provider of drug discovery services based on artificial intelligence says in 18 months it identified and tested in mice a new drug for a chronic lung disease. Insilico Medicine in Hong Kong says its process to find a new treatment for idiopathic pulmonary fibrosis covered conventional drug discovery stages from hypothesis to initial animal testing that can take years.

Insilico Medicine offers drug discovery software as a service based on deep machine learning and other A.I. techniques. Deep learning is a form of machine learning that finds underlying patterns in relationships, and builds those relationships into knowledge bases applied to a number of disciplines. Insilico says it extends deep learning with a technique called a generative adversarial network using two sets of algorithms that test each other while learning the underlying patterns. Those repeated rounds test the authenticity of new data generated by the model, eventually producing optimized data points.

The company applied these techniques to discover a treatment for idiopathic pulmonary fibrosis, or IPF, a progressive lung disease usually affecting people between the ages of 50 and 70. The disorder results in fibrosis or scar tissue building up in the lungs, limiting the ability of lungs to transfer oxygen to the blood stream. The most common symptoms are shortness of breath and a dry hacking cough, as well as a loss of appetite and weight loss in some cases. The scarring of lung tissue increases over time, often leading to other serious lung conditions, including lung cancer and blood clots in the lungs. Most patients die within five years following diagnosis. Some 100,000 people in the U.S. have the condition, with 30,000 to 40,000 new cases diagnosed each year.

Connecting AI for biology and chemistry

Insilico first identified a targets for IPF therapies using its deep learning target discovery system called PandaOmics that analyzes gene, transcription, protein, and clinical data sets. The company says it amasses these data sets from patent, research publication, grant, and clinical trial databases. In this case, the Insilico team instructed the system to predict responses of prospective fibrosis pathways, as well as the age and sex of patients. The company says the PandaOmics analysis revealed 20 targets for validation, later narrowed to a single intracellular target.

In the next stage, Insilico called on its Chemistry42 system for drug molecule design. The company says Chemistry42 builds molecules from scratch, creating and testing chemical structures that meet the target’s properties and binding characteristics. The system also responds to needs for specified drug properties, such as molecule complexity or shape, metabolic stability, and solubility. For Insilico’s IPF target, the designed molecules needed to create a safe oral drug meeting specifications for availability in the body, selectivity, and stability.

The result, says the company, is an IPF drug candidate code-named INS018_055 that in lab cultures shows the desired chemical activity against fibrosis biomarkers and blocks formation of inflammatory fibrotic cells. In tests with lab mice, the candidate shows no toxic effects for 14 days in concentrations of 60 milligrams per kilogram. In addition, the tests show less fibrosis and improved lung functions in lab mice induced with IPF, equivalent to the current IPF drug nintedanib, but at a lower dose. Tests of drug metabolism and chemical activity in mice suggest INS018_055 can be formulated into a once-a-day oral drug.

Insilico says the company is opening new drug development division with INS018_055 as its first candidate for eventual clinical trials. Alex Zhavoronkov, founder and CEO of Insilico Medicine says in a company statement released through EurekAlert, “To my knowledge this is the first case where A.I. identified a novel target and designed a preclinical candidate for a very broad disease indication.” Zhavoronkov adds, “now we have successfully linked both biology and chemistry and nominated the preclinical candidate for a novel target, with the intention of taking it into human clinical trials.”

More from Science & Enterprise:

• AI Cell Analysis Start-Up Raises $6M Seed Funds
• AI Designs Better Gene Therapy Delivery Viruses
• AI Process Devised for Drug Repurposing
• AI, Discovery Companies Partner on Neuro Imaging Drugs
• Start-Up Creating Digital Therapies for Alzheimer’s

*     *     *