Cellarity, a U.S.-based biotechnology firm, has revealed a groundbreaking manuscript in the journal Science that outlines a framework for integrating advanced transcriptomic data with artificial intelligence models to enhance complex drug discovery.
Based in the United States, Cellarity focuses on developing therapies that correct cellular states through integrated multi-omic models and AI. The company designs innovative treatments for complex diseases by examining the interactions and connections of pathways that define and modulate cellular states. Utilizing a robust drug discovery platform, Cellarity leverages high-dimensional transcriptomics to map these interactions at single-cell resolution.
The AI models created for this platform effectively link chemistry with disease biology to efficiently produce drugs that restore cellular function in diseased tissues. The initial candidate emerging from this platform, CLY-124, is currently under evaluation in a Phase 1 clinical trial aimed at treating sickle cell anemia.
“We believe that a comprehensive view of the cellular state will enable us to develop better therapies that can correct the underlying mechanisms of disease,” said Parul Doshi, the Chief Data Officer at Cellarity. “Our cutting-edge platform allows us to effectively visualize this dynamic and identify the most suitable novel interventions to correct diseases.”
The article published in Science details the evaluations that have shaped the Cellarity platform, emphasizing the rigor and ingenuity involved in successfully integrating advanced transcriptomics with computational tools to facilitate the efficient discovery of new therapeutic candidates. This publication presents a reproducible and generalizable model for incorporating machine learning methods into drug programs, maximizing discovery potential.
The model addresses various limitations associated with conventional phenotypic drug screening through an active, lab-driven deep learning framework powered by high-throughput transcriptomics. By iteratively refining predictions based on experimental outcomes, the framework demonstrated a recovery of phenotypically active compounds that was 13 to 17 times greater than standard industry approaches.
Over recent decades, the drug discovery process has struggled to improve its success rates, partially due to a traditional focus on single targets. Diseases often arise from more intricate interactions than a simple genetic mutation. By examining not only the phenotypic connections driving disease pathophysiology but also the polypharmacological considerations of early candidates, this deep learning platform holds significant promise for accelerating discovery and introducing effective oral therapies for complex diseases, added Jim Collins, Termeer Professor of Engineering and Medical Sciences at MIT, co-founder of Cellarity, and co-author of the publication.
In conjunction with the Science publication, Cellarity is releasing single-cell datasets that span multiple data modalities to foster community engagement, comparative modeling assessments, and a deeper understanding of the nuances of cellular states under chemical perturbation. A perturbational transcriptomic dataset, utilized to evaluate the Cellarity platform in the publication, comprises over 1,700 samples, encompassing 1.26 million individual cells, which can be employed for mapping drug responses among cell types or for a more comprehensive comparative evaluation of perturbation prediction methods.
Cellarity is also launching a multi-omic hematopoiesis atlas of individual cells that combines transcriptomics, surface receptors, and chromatin accessibility to create a multi-layered portrait of this essential and complex biological process. This atlas was used in the publication to develop fine-grained signatures of megakaryopoiesis and erythropoiesis. Additionally, a third dataset captures a timeline of megakaryocyte differentiation under perturbation, which can be analyzed to map the maturation trajectory of megakaryocytes, examine time-resolved drug effects, or support comparative evaluation and model training. Public analyses of these valuable data sets could yield new insights into cellular dynamics and drive innovative methods to accelerate drug discovery across the industry.
