Current status and opportunities in adaptive data analysis for therapeutic cell manufacturing. (December 2021)

• Record Type: Journal Article
• Authors: Liu, Zhaonan; Chen, Jialei; Wang, Kan; Wang, Ben; Zhang, Chuck
• Is Part Of: Current opinion in biomedical engineering. Volume 20(2021)
• Publication Date: 2021-12
• Subjects: Cell manufacturing -- Data analysis -- Variability; Biomedical engineering -- Periodicals; 610.28
• DOI: 10.1016/j.cobme.2021.100351
• Languages: English
• ISSNs: 2468-4511

Abstract

Abstract: As an emerging technology, therapeutic cell manufacturing faces major challenges in three aspects, knowledge gap, regulations, and case-to-case variability. Among all variability sources, donor-to-donor variability is intrinsic to therapeutic cell manufacturing and can be very large. A few recent articles have addressed this variability, but enormous research opportunities remain. In this opinion article, we focus on the donor-to-donor variability and point out a new sub-field in the data analysis research, adaptive data analysis, as a potential solution to minimize the donor-to-donor variability. An adaptive data analysis framework adapts to each case and provides case-specific instructions on the manufacturing process. We present and discuss three specific adaptive data analysis approaches, including multi-task learning, Bayesian latent variable modeling, and representation learning. These modeling techniques may provide valuable solutions to challenging data analysis problems in cell manufacturing where large variabilities exist.