Article Text
PDF
Abstract
Background Single-cell technologies have revolutionized the generation of extensive data from human tissues, providing unprecedented resolution. Although numerous cell atlas initiatives have made significant progress in describing and cataloging cellular complexity, fully harnessing the potential of single-cell data for biomarker discovery and clinical applications remains a challenge.
Methods Scailyte has developed an innovative approach that combines multi-omic single-cell analysis, integration of clinical data, and a tailored supervised machine learning platform called ScaiVision (figure 1). This platform enables targeted and sensitive biomarker discovery, transcending specific indications and cell types. By leveraging data augmentation and a cluster-free approach, ScaiVision extracts relevant information while preserving single-cell resolution throughout the analysis. These distinctive features yield remarkably high success rates in discovery of rare cell populations and features from a limited number of samples, providing actionable insights for assay development.
Results We emphasize the unique advantages of our methodology compared to standard single-cell analysis approaches, particularly its successful application in biomarker discovery projects spanning various disease areas. Utilizing scRNAseq data from CAR T cell infusion products, we have trained AI models and identified complex biosignatures that predict the development of severe neurotoxicities and response to CAR T therapy in lymphoma patients.
Conclusions Our approach offers a rapid and efficient end-to-end pipeline for extracting relevant biosignatures from single-cell data, elucidating efficacy, toxicity, and therapeutic mechanisms. These discoveries de-risk drug development programs and serve as the foundation for companion diagnostics and in-vitro diagnostic assay prototypes on widely adopted clinical platforms.
Toxicity prediction: ScaiVision identifies T cell signature predictive of neurotoxicity in CAR-T cell therapy of DLBCL.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.