top of page

Xaira Therapeutics: AI Drug Discovery Optimized for Human Biology

The Driver: 

Xaira Therapeutics recently raised $1 billion to apply AI to the discovery and development of new drugs. As the company noted in its press release announcing the funding, Xaira brings together three core elements: advanced machine learning research, expansive data generation to power new models, and robust therapeutic product development. The funding, the 3rd largest biotech funding round since 2010 was led by ARCH Venture Partners and Foresite Capital, joined by F-Prime, NEA, Sequoia Capital, Lux Capital, Lightspeed Venture Partners, Menlo Ventures, Two Sigma Ventures, the Parker Institute for Cancer Immunotherapy (PICI), Byers Capital, Rsquared, and SV Angel, among others.


Key Takeaways: 

  • For every 200 proteins investigated in research, only one in 200 correlates to causation in human disease and shows potential in drug discovery (Nature)

  • The biopharma industry spends over ten times what it did in the 1980’s on development, after adjusting for inflation but yet has shown virtually no gains in productivity since that time (Congressional Budget Office)

  • The use of Machine Learning in drug discovery could save approximately $300-400M per drug (15%-40%)

  • Xaira's recent fundraising was the 3rd largest biotech funding since 2010 behind only Altos Labs in 2022 and Roivant Sciences in 2017 (Endpoints News) 

The Story:

Xaira was founded based on the technology emanating from the University of Washington’s Institute for Protein Design  and the lab of Dr. David Baker. In July of 2021, Dr. Baker’s lab developed “RoseTTAFold, a software tool that uses deep learning to quickly and accurately predict protein structures based on limited information in as little as ten minutes on a single gaming computer. In December 2022, the Institute for Protein Design released RFDiffusion, “a powerful new way of designing proteins that combines structure prediction networks and generative diffusion models.” According to the lab, the model demonstrated “extremely high computational success using the new method and experimentally tested hundreds of A.I.-generated proteins, finding that many may be useful as medications, vaccines, or even new nanomaterials.” Using these technologies as a foundation, Xaira is going to attempt to create and deploy AI models that have been specifically designed, built and optimized based on human biology for drug development and discovery. According to the company, at first it will target protein based antibody drugs but according to Chemical and Engineering News they are “looking for validated targets that are underserved by the current technology”.

The article also noted that, the company came together by a coincidence of timing as both Baker and his co-founder Hetu Kamisetty had been discussing founding a company based upon Baker’s technology while at the same time, two of its backers Bob Nelsen of Arch Ventures and Vik Bajaj of Foresite Capital were looking for investment opportunities in the AI space. In addition to Baker and Kamisetty, who previously worked for Meta, the founding team includes  Dr. Marc Tessier-Lavigne, former Chief Scientific Officer of Genentech and former President of Stanford University as well as Dr. Arvind Rajpal, formerly of Genentech; and Dr. Don Kirkpatrick, formerly of Interline and Genentech. It should be noted that Dr. Tessier-Lavigne resigned from his position at Stanford following an inquiry into research misconduct on papers on which he was the principal author. While the special committee did not find any evidence of “fraud, fabrication, or other intentional wrongdoing” questions were raised about Dr. Tessier-Lavigne’s oversight of the research. Dr. Tessier-Lavigne did subsequently issue or attempt to issue corrections to some of the research cited (two of which were not published due to lack of action on the part of the journal in question). 

The Differentiators:

In addition to the personnel from Baker’s Lab, as noted by Genetic Engineering and Biotechnology News, Xaira is also positioned to benefit from “technologies and personnel from Illumina’s functional genomics R&D effort” which have also been spun out into the company. Moreover, “the proteomics group from Interline Therapeutics, a drug developer whose precision medicine platform is designed to map and modulate protein communities” has also been integrated into the company.

Based on the extensive resources required to create and test AI models for drug development, we think Xaira’s strong funding, deep scientific and regulatory bench and integrated technology platform, position them well for success.

The Big Picture :  

According to a 2019 article in the journal Nature, there is an “overall failure rate in drug development of over 96%, including a 90% failure rate during clinical development.” The article went on to add that their conclusion was based on the assumption that there were “10,000 diseases, 20,000 protein-coding genes, 100 causal genes per disease and 4000 genes encoding druggable targets”...leading to the conclusion that for every 200 proteins investigate in research, only one in 200 correlates to causation in human disease and shows potential in drug discovery.

the discovery of one in 200 protein-to-disease causation in drug discovery.

While these numbers may sound shocking, they have been relatively consistent in drug discovery and development. For example, despite the fact that the biopharma industry spends over ten times what it did in the 1980’s on development, after adjusting for inflation, the overall likelihood for approval of a drug candidate (from phase I through approval) ranges from just 24% for a hematology drug to under 4% for a urology drug. As a result, it appears clear that AI technology could be used to both improve discovery and speed to market of new drug candidates. In fact, according to the U.S. General Accounting Office, the use of Machine Learning in drug discovery could save approximately $300-400M per drug (15%-40%). Moreover, since human disease is often the sum of many simultaneous interactions (and not just an isolated single interaction often targeted in current animal models), AI is likely to be successful both intentionally and unintentionally (ex: identifying drug targets that may not be the focus of the original search).


Search By Tags
Recent Posts
Follow Us
  • Facebook Basic Square
  • Twitter Basic Square
  • Google+ Social Icon
bottom of page