Genesis Therapeutics, an AI drug research and development company focusing on using physical AI for structure-driven drug design, recently announced that it has received additional investment from NVentures, the venture capital arm of NVIDIA, and the cooperation between the two parties has further deepened. The investment will be used to accelerate the development of Genesis AI platform GEMS and enhance its capabilities in structure-driven drug design. The GEMS platform integrates multiple AI methods, including language models, diffusion models, and physical machine learning simulations, to generate and optimize molecules for complex targets. Since its establishment in 2019, Genesis has raised more than US$300 million and has established partnerships with a number of biopharmaceutical giants to promote the development of AI drug research and development.
Originated from Stanford, deeply involved in molecular AI
Genesis Therapeutics was spun out of Dr. Vijay Pande's lab at Stanford University. Co-founder Dr. Evan Feinberg co-invented and wrote several key papers on deep learning technology with Pande during his graduate studies, the most notable of which was the PotentialNet algorithm. The algorithm pioneers the use of novel graph neural networks for molecular property predictions, specifically protein-ligand binding affinities. Feinberg, Pande, and colleagues demonstrated the performance of PotentialNet on potency prediction and further validated its effectiveness through a collaboration between Stanford University and Merck Research Laboratories. Before founding Genesis, Feinberg served as a deep learning consultant for Merck.
Raised over US$300 million in financing and in-depth cooperation with NVIDIA
Genesis was founded in 2019 and raised $52 million in Series A funding a year later. Since then, the company has continued to grow and has raised more than $300 million in funding to date, most of which came from a $200 million Series B round completed in 2023 from investors including NVidia's venture capital arm NVentures.
Through its partnership with NVIDIA, Genesis is working to accelerate the development of its AI platform, GEMS. GEMS is designed to generate and optimize molecules for complex targets by integrating proprietary AI methods including language models, diffusion models and physical machine learning (ML) simulations. The additional financing from NVentures aims to further enhance the capabilities of Genesis' physics AI platform for structure-driven drug design by applying NVIDIA's expertise to improve computational efficiency.
Feinberg said: "Nvidia is a leader in many aspects of the AI stack, both on the hardware side and the lower software layers on top of the hardware. And Genesis has been committed to being a pioneer in molecular AI. So Nvidia's There is a very clear synergy between comparative advantage and Genesis's comparative advantage, making the combination greater than the sum of its parts."
Optimize neural networks to accelerate drug development
The collaboration will cover the optimization of equivariant neural networks, which are valuable for processing 3D geometric data such as protein and small molecule structures. NVIDIA has been working on accelerating computing through neural networks, including training the network and running inference, using the trained model to make predictions on new data or deploying it in real environments.
Feinberg explains: "For the field of molecular AI that Genesis has been pioneering for many years, there are specific types of neural networks that are particularly useful. This is really a continuation of a long-term trend in the field that AI is not a monolith. There are many subfields of artificial intelligence, and these Subfields learn using related but different algorithms."
At Stanford University, Feinberg, Pande and a group of colleagues proposed the PotentialNet family of graph convolutions in a 2018 paper published in ACS Central Science. Two years later, another group of colleagues, together with Feinberg and Pande, showed how by explicitly representing each molecule as a graph, "to our knowledge, unprecedented success" in predicting ADMET (absorption, distribution, metabolism, elimination, and toxicity) properties was achieved. Accuracy," and in a paper published in the Journal of Medicinal Chemistry, shows the significant advantage of the AI algorithm in ADMET predictions over advanced ML used by Merck Research Laboratories.
Close collaboration between founders and mentors
Pande is now a general partner at Andreessen Horowitz (a16z) and a founding partner of the a16z Bio Fund, where he leads the firm’s investments in biology, computer science and engineering. Pande served as Feinberg's doctoral advisor, led a16z's $4.1 million seed investment in Genesis, and co-led the company's $200 million+ Series B with an undisclosed U.S.-based life sciences investor. Financing.
Feinberg said of Pande: "I have been extremely fortunate to work with him for nearly a decade. I think it is rare to be able to work so closely with and learn from such a talented and visionary person."
Continuous innovation leads the development of the industry
Feinberg added: "He (Pande) has always pushed me in a way that has been critical to the success of Genesis. As the field has evolved, he has also continued to evolve. I think that is consistent with us remaining a leader in the field." In a similar way to our status, we continue to innovate and not just be content with imitation, but really push the field forward.”
Feinberg recalled that during his graduate studies at Stanford University, AI mainly had an impact in the fields of computer vision and natural language. "The types of neural networks used for both were actually very different from each other, but neither was very suitable for chemistry. So we developed new types of neural networks," Feinberg recalls. "In the mid-2010s, graph neural networks were better suited for molecules. ”
Feinberg said that from then until now, Genesis has been continuously researching new AI algorithms and "new neural network primitives that are more suitable for molecular AI tasks." "Equivariant neural networks are one of the families we value. It's one of the areas that NVIDIA specifically helps us optimize," Feinberg added.
Pande's lab initially rose to prominence for the distributed computing project it founded, Folding@Home, which was designed to simulate protein dynamics, including the protein folding process.
Feinberg recalls: “Folding@Home leveraged a large number of NVIDIA GPUs around the world for protein folding simulations. After that, NVIDIA GPUs started to be used more for artificial intelligence, especially in vision and natural language. So our company can already be said to be Powerful user of NVIDIA GPUs.”
A “match made in heaven” with NVIDIA
Feinberg said: “When we were introduced to Nvidia and NVentures through our Series B round, it felt like a very natural investor who would not only bring significant capital, but also bring wisdom to the relationship.” The investment really sets the stage for us to work together beyond the client relationship and thereby also enable us to learn from each other, both from our needs and from their lower-level capabilities that we can uniquely leverage with our domain knowledge."
For Nvidia, the partnership with Genesis strengthens its ongoing efforts to apply AI to drug discovery.
Mohamed “Sid” Siddeek, corporate vice president and head of NVentures at NVIDIA, said: “Genesis’ AI platform and related computing advances developed in partnership with NVIDIA will help deliver new generative and predictive AI technologies to Explore untapped chemical pathways and identify drug candidates."
How does GEMS help NVIDIA?
"The goal of GEMS is to be able to efficiently develop very challenging and, in some cases, undruggable targets," Feinberg said. "In order to do that, we need to do several capabilities better than we have done before." "
This involves generating molecules and predicting their potency, selectivity and atomic properties - a combined multi-parameter optimization approach to drug discovery that jointly studies all key properties of a molecule. Feinberg explained that GEMS is composed of two deeply integrated pillars - generative AI and predictive AI - and has used Genesis's own custom language models to generate thousands to millions or even billions of compounds in the cloud.
"But chemistry, synthetic chemistry, is the limiting factor. Only so many molecules can be made in a given time. So it's critical that our predictive AI technology - which predicts potency, selectivity and atomic properties - is as accurate as possible. So, GEMS is really a collective term that describes a combination of deeply integrated technologies," Feinberg said.
GEMS applications in oncology and immunology
Leveraging GEMS, Genesis is developing a pipeline focused on oncology and immunology. In oncology, Genesis is in late-stage lead optimization and is close to nominating a highly potent and selective development candidate for what it calls a pan-mutational allosteric inhibitor of PIK3CA, a common oncogenic driver of breast and colorectal cancer.
Other oncology development efforts focus on small molecules designed to overcome responses to checkpoint inhibitors (lead optimization phase) and prevent cancer cells from evading apoptosis through anti-apoptotic modulators that inhibit the extrinsic cell death pathway. (Discovery phase).
In immunology, Genesis said it has two discovery-stage efforts: one to develop multiple programs to generate small molecules against well-validated autoimmune disease targets; Corrective agents to restore the activity of unspecified damaged proteins to treat "severe inherited autoinflammatory diseases."
Cooperation with biopharmaceutical giants
In addition to internal development work, Genesis is working on announced collaborations with three biopharmaceutical giants, but Feinberg said the company could not comment on those. The most recent collaboration was launched in September with Gilead Sciences, which agreed to use GEMS to assist in the generation and optimization of molecules for Gilead's selected targets, enabling the discovery and development of small molecule therapeutics against multiple targets. .
Gilead agreed to pay $35 million for three targets and has the right to nominate additional targets for an undisclosed predetermined per-target fee. Gilead also agreed to pay additional payments related to the achievement of preclinical, development, regulatory and commercial milestones, as well as tiered royalties on net sales of commercialized products.
Collaborations with two other biopharmaceutical giants:
Eli Lilly - Collaboration worth up to $670 million (of which $20 million is an upfront payment) to discover new treatments in up to five therapeutic areas, launching in 2022.
Genentech, a member of the Roche Group - a collaboration involving multiple targets and multiple diseases, launched in 2020, using Genesis' platform for deep learning and molecular simulation. In 2022, Genentech described the targets it was interested in as "challenging targets that are otherwise inaccessible." The value of the collaboration has not been disclosed.
Genesis is headquartered in Burlingame, California, a suburb of San Francisco, with a fully integrated laboratory in San Diego. The company employs approximately 80 people.
"We do have a lot of expected growth, driven in part by the Series B round, Nvidia's latest investment, and our partnerships," Feinberg said. "I don't have an exact number on where we will be in 12 months." scale, but we do have the headcount to go beyond 80 people.”
The cooperation between Genesis Therapeutics and NVIDIA marks an important step in the field of AI-driven drug research and development. The continued development of its GEMS platform and cooperation with biopharmaceutical giants are expected to accelerate the research and development process of new drugs and bring new treatment options to patients. .