Insilico Medicine
Addressing the Next Age of Drug Discovery

Discovery of medicine has historically been accredited as a ‘serendipitous’ moment—an apt illustration being the discovery of Penicillin. The lion’s share of drug discovery breakthroughs has been left to happy coincidences; today, we have an entirely different story with the protagonist being a Baltimore-based innovator, Insilico Medicine. The company bridges aging research and AI to develop and interpret the highly accurate and biologically-relevant signatures of aging and disease.

These signatures—known as “biomarkers”—are developed for specific age-related diseases using large population-level datasets of healthy patients of various ages and Insilico Medicine trains deep neural networks (DNNs) to predict the chronological age accurately. The company then re-trains the DNNs on the smaller populations with a particular disease, ranks the features by their relative contribution to the accuracy of the system, and identifies the targets that may be addressable with the small molecules. The company then uses its comprehensive drug discovery engine to identify and generate de novo the best molecules to either inhibit or activate the proteins of interest.

Insilico Medicine has identified several thousand promising compounds that have gone through various stages of validation using the AI-powered applications Juvanescence. AI and Pharma.AI. These compounds are then licensed to biotechnology companies for collaborative development. For instance, Insilico partnered with Life Extension, one of the largest nutraceutical companies in the U.S., and launched several nutraceutical products. The customers consuming the nutraceutical products can enter their biomarker data, including pictures and blood tests, into a system called Young.AI, which predicts their biological age and tracks the activity of the nutraceuticals.

“Age is a universal parameter that every living being has, and it is possible to aggregate massive datasets, where patients have age as a label, build predictors of chronological and biological age, extract and interpret the most important features,” explains Alex Zhavoronkov, founder and CEO of Insilico Medicine.

Insilico Medicine is developing the end-to-end pipeline which utilizes AI to form the disease hypothesis for the many age-related diseases

Using a technique called “one-shot learning,” the company multiplexes the features from the many data types and re-trains the predictors of age on the smaller and less abundant datasets of disease. Many of these datasets are managed using a Blockchain-driven data repository intended to be turned into a global marketplace through a partnership with BitFury Group, one of the largest Blockchain companies.

The methodology of maintaining and assessing datasets of diseases came to the rescue of a large pharma company, who approached Insilico Medicine to perform an analysis of one of their failed clinical trial in the respiratory diseases space. They were challenged with the aggregation of many data types to understand the features that are most predictive of the failure of the trial. Not only was the project completed successfully, but they also identified a promising set of molecules and identified new targets for the disease of interest.

Identified among NVIDIA’s top 5 AI companies for social impact, Insilico Medicine has taken the role of a global pharmaceutical innovation hub. Since 2014, the company has collaborated with over 200 academic and industry partners, published over 40 peer-reviewed research papers, and replaced several teams of medicinal chemists and bioinformaticians with the AI-powered drug discovery engine, reducing their workforce to just 36 scientists globally. Insilico Medicine has also announced partnerships with GSK, BioTime, multiple hospital networks in Korea, and the governments of Taiwan and Kazakhstan.

“Unlike other companies in the field, we are developing an end-to-end pipeline which utilizes AI to form the disease hypothesis for the many age-related diseases and discover targets,” Zhavoronkov summarizes. “We will continue to leverage some of the most recent and advanced forms of AI to generate novel molecular structures and address the targets.”