(Veteran tech columnist Jon Markman publishes Strategic Advantage, a popular daily guide to the great digital transformation of business and society — and how to invest in it. Click here for a free two-week trial.)
Big stock market moves are always about a change in narrative. Some healthcare stocks are undergoing one of those changes. And it started in a computer lab in London, England.
DeepMind is an artificial intelligence company owned by Alphabet ((GOOGL) -Get Report). Last week the company’s algorithms solved protein folding, a biology riddle that has puzzled researchers for five decades. The solution changes the narrative about risk.
It means many healthcare businesses are dramatically underpriced. Let me explain.
Most modern drug therapies are based on scientists manipulating amino acids to make proteins. As the proteins develop they fold into extremely specific shapes that define their function. Prior to AlphaFold, the DeepMind algorithm, there was no way to reliably predict what 3D shape these proteins would form. This is why investing in biotechnology carries so much risk. Developing new molecular medicines is largely hit or miss.
In theory AlphaFold erases the misses.
It’s like a cheat sheet. With further computation, researchers at pharma and biotech companies should be able to reliably develop proteins for specific functions. It means faster, cheaper drug discovery with much less risk.
The rapid development of AlphaFold is shocking. DeepMind software engineers decided in 2018 to participate in the Critical Assessment of protein Structure Prediction. CASP is an contest run by the biology community to assess the progress of predictive techniques. Since 1994 CASP has been run every 2 years.
In CASP 13 AlphaFold 1 was far and away the best tool in the field. It managed to get the basic shapes correct but the details, such as where the atoms sit, were misplaced. By CASP 14, less than three years later, the second incarnation of AlphaFold was exponentially better, achieving an proficiency rate of 92.4%.
For perspective, this matches the accuracy of expensive, time consuming laboratory techniques such as nuclear magnetic resonance, x-ray crystallography and cryo-electron microscopy, according to a report at MIT Technology Review.
Lab analysis for a single protein can take years and cost hundreds and thousands of dollars. Results from AlphaFold 2 are available in a few days.
And that is the investment story in a nutshell.
Solving protein folding fundamentally changes the narrative about drug discovery. Proven computational AI is a game-changer.
Schrodinger ((SDGR) -Get Report) managers built a cutting edge computational AI and physics-based predictive analytics platform. The promise is to help companies cut the cost of drug discover by using computer simulation.
The New York city-based company was founded in 1990 and boasts 150 Ph.D. scientists. Together they have published 400 peer-reviewed white papers on physics-based computer modelling. In the prospectus filed with the Securities and Exchange Commission in January, Schrodinger managers took the unusual step of claiming a big lead over the nearest competitors. It’s a bold claim backed up facts.
Schrodinger software is in use at 1,250 academic institutions worldwide. The platform also claims as customers all of the top 20 pharmaceutical companies by sales. And its client retention rate its entire 1,150 client list is 96%.
The industry is expected to spend $202 billion on drug discovery by 2022. So far Schrodinger has captured only a tiny sliver of that sum. The company had $85.5 million in sales in 2019, the last full fiscal year.
However, Bristol Myers Squibb ((BMY) -Get Report) announced in November it would pay $55 million upfront to access Schrodinger’s platform. The pair will jointly develop drug treatments for five small molecules designed with the software. If the drugs pan out the deal could be worth an addition $2.7 billion plus a royalty to Schrodinger.
The AlphaFold news means the odds of getting those, and a lot of other drugs, to market is suddenly much better.
More important, AlphaFold shifts the narrative about drug development away extreme costs and long delays. Investors will begin revaluing computational biology facilitators higher. The opening of the R&D floodgates will be a good tailwind, too.
Schrodinger trades at 1,104x forward earnings and 46.7x sales. While those metrics may seem grossly expensive, keep in mind where the computational biology is headed and Schrodinger’s place.
Longer-term investors should buy shares into weakness.