W212 - Generation of Synthetic Single Cell Transcriptomic Data from High-throughput Cytometry Enables Population-scale, Deep Resolution Immunophenotyping in Health and Disease

Abstract Text: Single cell transcriptomics has driven the rapid progress of systems immunology by revealing the heterogeneity and complexity of the immune system with unprecedented resolution. However, broad clinical impact has been constrained owing to limitations in throughput, excessive cost, and significant sampling error and reproducibility associated with low cell numbers captured per sample. To address this, IMU Biosciences has developed Transcriptomic-Enhanced Cellular High-throughput Phenotyping (TECH-P); a novel machine-learning (ML) approach that uses high-content spectral flow cytometry to faithfully produce synthetic single cell transcriptomic data. Trained on a growing set of samples from healthy donors and patients processed in parallel by the IMU phenotyping platform and single cell RNA sequencing, TECH-P employs modality harmonisation alongside deep learning to accurately predict gene expression profiles for hundreds of thousands of cells per sample. Application of TECH-P to an internal dataset of 1,048 individuals profiled by the IMU platform generated a single cell atlas of the healthy immune system comprising ~120 million cells, surpassing extant single cell atlases. This largescale synthetic immune database faithfully describes known cell subset specific transcriptional programs and replicates known single cell transcriptional signatures associated with aging. Ongoing development of TECH-P aims to bridge single-cell transcriptomics with routine, high-throughput and statistically robust immune profiling of clinical samples. This integration has the potential to transform a wide range of applications, from target discovery and drug development to patient stratification and precision medicine.