Vevo Therapeutics Plans to Open Source 100M Single-Cell Atlas for Drug Discovery
Vevo Therapeutics has announced plans to open source its Tahoe-100M dataset, described as the world's largest single-cell transcriptomic atlas of drug impacts on patient cells.
Developed in collaboration with NVIDIA's biology foundation model research team and Parse Biosciences' GigaLab Platform, the dataset represents a significant resource for advancing machine learning applications in drug discovery. NVIDIA will contribute expertise in model training and data engineering to make the dataset usable for the broader research community.
Tahoe-100M encompasses data on 100 million cells, drawn from 60,000 experimental conditions, 1,200 drug treatments, and 50 tumor models. It is reportedly 50 times larger than all publicly available drug-perturbed single-cell datasets combined. The dataset was created using Parse’s Evercode split-pool barcoding system, which allows for tagging individual transcripts without requiring specialized equipment. This method high sensitivity in sequencing while reducing common errors, such as ambient RNA contamination. Parse’s high-throughput platform enabled the rapid creation of Tahoe-100M, completing this ambitious project in just one month.
The dataset integrates with Vevo’s Mosaic platform, which is designed to generate high-resolution in vivo data at scale. Mosaic employs innovative techniques for pooling cells from dozens to hundreds of diverse patients into single experiments. This approach allows for the study of drug efficacy and resistance in a way that better represents patient diversity earlier in the drug discovery process. Additionally, Mosaic provides single-cell resolution, capturing both phenotypic and transcriptomic changes to study drug-induced effects with high precision.
Vevo’s focus on in vivo data addresses the limitations of traditional in vitro models, which do not fully account for the complexity of disease in living organisms. By training AI models on its in vivo atlas, Vevo aims to uncover novel drug targets and therapeutic candidates that might remain undetected using conventional methods. By combining large-scale single-cell data with machine learning, Vevo aims to uncover how drugs work and why they might fail, providing researchers with deeper insights into drug effects and resistance.
Topics: AI & Digital
