Seed investors are backing a startup that sees machine learning as the key to enabling drugs to be aimed at molecular targets that have long been thought to be out of pharmaceutical companies’ reach.

Unnatural Products Inc. said it has secured more than $6 million in seed financing to fund its mission of solving the problem of “undruggable” proteins—the roughly 80% to 90% of proteins in the body that have eluded drug makers. Some proteins are inaccessible because they lack a binding “pocket” where drugs can dock.

Another challenge, and the focus of Unnatural Products’s initial efforts, is the interactions between large proteins in cells. Disrupting these interactions typically requires compounds with a large surface area, said Unnatural co-founder and Chief Executive Cameron Pye. But compounds with larger surface areas are usually less able to permeate cells, according to Dr. Pye.

Certain molecules found in nature called macrocycles have shown the ability to disrupt protein-protein interactions. Existing macrocycle medicines include the immunosuppressive drug cyclosporine A.

Drug developers historically have needed some serendipity to identify ideal macrocycle drugs. But with the help of machine learning, Santa Cruz, Calif.-based Unnatural seeks to optimize macrocycles to turn them into effective medicines. That includes predicting changes that would need to be made to the drug to improve its potency and ability to hit its intended target with great specificity, according to Dr. Pye.

Unnatural plans to build its own pipeline of compounds and will use the new financing to move its early, discovery-stage research forward, Dr. Pye said. Unnatural seeks to identify cancer treatments initially, but the company says its approach also could apply to inflammation and various other conditions.

Venture investors are also backing several other companies taking different approaches to solving the undruggable-target problem.

One of them, Frontier Medicines Corp., identifies temporary binding pockets that proteins form when they are in motion. Through a library of chemical compounds, Frontier says it can search for drugs that bind to these pockets. The company raised $67 million in June.

Dr. Pye said several strategies will be needed to expand what today is a small population of proteins that drug makers can target successfully.

“It takes a village to solve this problem,” he said.

Artis Ventures led Unnatural’s seed financing. Abstract Ventures, Asset Management Ventures, Better Ventures, Blue Bear Ventures and Rising Tide Fund also participated.

This article was originally posted in the Wall Street Journal. Read it here: https://www.wsj.com/articles/startup-aims-to-unlock-drug-discovery-with-machine-learning-11568823688