Abstract LB106: Unleashing the power of AI-driven drug discovery with a 100 million single-cell atlas

Abstract In the past decade, single-cell RNA sequencing has become a ubiquitous and powerful tool for studying a large variety of biological systems across different research applications. One strength of the technology lies in its potential to measure the transcriptome of millions of individual cells within a biological sample or model system. The emergence of AI driven drug discovery requires a vast amount of high-quality input data, covering a large range of experimental conditions, in order to train and refine models to generate accurate or actionable predictions. Recently, single-cell combinatorial barcoding technology has advanced to meet the demands of AI modeling, allowing for simultaneous measurements of whole transcriptomes from 10M to 100M cells in one experiment. Here, we describe the methods for a model system that encompasses over 50 cell lines, ∼300 drugs and over 100M cells that sets a new standard for modeling drug perturbations in diseased cells. In this first-of-a-kind study, mixed proprietary cell lines (Vevo Therapeutics) were co-cultured then treated for 24 hours with hundreds of cancer drugs at 3 concentrations. At the end of the treatment period, cells were dissociated and then fixed to allow for single cell barcoding in batches of more than 10M cells each. Combinatorial barcoding was performed using Parse’s GigaLab platform, scaling the workflow for multi-million cell inputs and higher throughput, while preserving transcript detection quality. The post-barcoding steps for cDNA capture, PCR, library preparation and clean up were automated in 96-well format. The entire experimental workflow from cell treatment through library generation was completed in less than 5 weeks. Sublibraries were converted for processing on the UG 100™ sequencer (Ultima Genomics). Sequencing output data from over 100M cells were processed using the Parse Biosciences pipeline, and then passed through Demuxlet to demultiplex cell lines. This study, encompassing over 100M cells across 60,000 conditions, establishes a scalable and efficient approach to single-cell sequencing, offering an invaluable resource for AI-driven biological research. It is expected to inspire the scientific community to pursue large-scale, multi-million cell single-cell sequencing experiments, advancing our understanding of disease biology and therapeutic responses. Citation Format: Vuong Tran, Efthymia Papalexi, Joey Pangallo, John Thompson, Simone Marujo, Bryan Hariadi, Crina Curca, Oliva Kaplan, Sarah Schroeder, Maria Nigos, Oliver Sanderson, Hoai Nguyen, Alec Salvino, Ryan Koehler, Gokhan Demirkan, Gary Geiss, Johnny Yu, Charles Roco, Alex Rosenberg. Unleashing the power of AI-driven drug discovery with a 100 million single-cell atlas [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_2):Abstract nr LB106.