1910, a biotech company focused on AI-driven drug discovery, has made significant strides in developing macrocyclic peptide drugs through its innovative PEGASUS AI model. Founded by Jen Asher, PhD, the company recently distinguished itself by rebranding from 1910 Genetics to simply 1910, reflecting its wider commitment to multi-modality drug discovery. Asher emphasized that the goal is to “make undruggable targets a thing of the past” and to transform every pharmaceutical company into an AI-driven entity.
In a recent publication in the Journal of Medicinal Chemistry, 1910 unveiled PEGASUS, a groundbreaking AI model capable of designing cell-permeable macrocyclic peptides. This advancement holds promise for targeting traditionally difficult intracellular locations, combining the oral convenience associated with small molecules and the high specificity typical of large biologics. Asher described PEGASUS as a “versatile tool” that streamlines the design-make-test cycle, operating in both predictive and generative modes.
The capabilities of PEGASUS include triaging compounds for synthesis, supporting lead optimization campaigns, and creating new starting peptides with desirable characteristics. Notably, the study reported the first cyclic peptides with more than two polar or ionizable fragments that successfully achieved in vitro cell membrane permeability.
Promising Developments in Neurological Therapeutics
PEGASUS is one of two recent flagship publications from 1910, showcasing different modalities emerging from the same platform. In November 2022, the company also released findings in the Journal of Chemical Information and Modeling about CANDID-CNS, another AI model that enhances the potential for oral drug therapies targeting neurological conditions. CANDID-CNS predicts small molecule blood-brain barrier (BBB) penetration within Beyond-Rule-of-5 (bRo5) chemical space, a category that intentionally bypasses Lipinski’s Rule of 5 to address challenging molecular targets.
With only about two percent of small-molecule drugs successfully crossing the BBB, the ability to accurately predict penetration is crucial. CANDID-CNS achieved an impressive 83% success rate in predicting brain penetration and distribution, significantly outperforming the industry standard of 64% established by Pfizer’s CNS Multiparameter Optimization score.
Asher noted that both models are “truly multimodal,” anchored by 1910’s distinctive capabilities in generating wet lab biological data. This data not only fuels the company’s internal pipeline but also supports upcoming pharmaceutical partnerships.
Founded in 2018 and emerging from Y Combinator with a seed round of $4 million led by OpenAI CEO Sam Altman, 1910 has since established a five-year commercial agreement with Microsoft. This partnership allows the company’s platform to be utilized by biotechnology firms, government entities, and research institutions through models of co-discovery, co-engineering, and Platform-as-a-Service (PaaS).
The company’s name, 1910, pays homage to the year when the first patient was diagnosed with sickle cell disease in the United States, marking a milestone in the identification of a molecular basis for disease.
Challenges and Future Potential
In the landscape of drug discovery, naturally permeable macrocyclic peptides are gradually transitioning from theoretical models to clinical application. Last November, Merck announced that its macrocyclic peptide candidate for hypercholesterolemia, enlicitide, demonstrated statistically significant and clinically meaningful reductions in LDL cholesterol during a Phase III trial. If approved, this could lead to the first oral PCSK9 inhibitor, potentially disrupting a market that has primarily relied on injectable treatments.
Despite these advancements, Asher warns that the successful adoption of enlicitide depends on a permeability enhancer, which can compromise cell membrane integrity in the intestine, resulting in increased absorption of non-drug molecules and variability in patient responses. She explained that these challenges, along with the higher costs associated with formulation, make achieving normal physiological cell permeability the preferred approach for ensuring oral bioavailability.
Achieving the permeability of macrocyclic peptides has been a complex task, as desirable traits like low polarity and high lipophilicity often conflict with critical therapeutic features such as high potency and solubility. Addressing this optimization challenge requires expanding current wet lab biological datasets, which are often concentrated in high lipophilicity chemical spaces.
To broaden its therapeutic applications, PEGASUS has been trained on a unique, multi-modal dataset generated from 1910’s proprietary Permeability Proxy Assay (1910 PPA). This assay produces billions of cyclic peptides categorized by permeability-related characteristics and solvent-dependent computational simulations. These data streams complement traditional cell permeability data from established in vitro systems like Caco-2 and Madin–Darby canine kidney (MDCK) cells, which are historically low-throughput and expensive to obtain.
Asher asserts that the integration of wet lab and dry lab methodologies can break through significant data barriers in AI-driven drug discovery. She believes that removing any one of the three data streams would diminish the predictive and generative capabilities of PEGASUS, referring to 1910’s surrogate assays as a “breakthrough for AI model training.”
The biotechnology industry continues to monitor these developments as 1910 takes significant steps toward translating these innovative drugs into tangible clinical results.
