Preprint Highlight: 7-UP: generating in silico CODEX from a small set of immunofluorescence markers

Abstract

Multiplexed molecular imaging enables the characterization of the spatial organization of different cell types in their native tissue context but is an expensive and labor-intensive technology. This study presents 7-UP, a method that computationally generates 40 protein expression markers, at single-cell resolution, from a panel of seven experimentally measured multiplexed immunofluorescence markers. The method relies on machine learning–optimized selection of seven markers and inclusion of spatial context from the raw image that together enable reconstruction of the full protein panel. 7-UP was validated by demonstrating accurate prediction of 16 cell types and patient outcomes in head and neck cancer data sets and generalized to an unseen cancer type. 7-UP can infer rich single-cell molecular profiles from commonly available multiplexed immunofluorescence assays. This preprint has been assigned the following badges: New Methods, Open Software, Cross-Validation. Read the preprint on bioRxiv ( Wu et al., 2022 ): https://doi.org/10.1101/2022.06.03.494624 .