RESTORE: Robust intEnSiTy nORmalization mEthod for Multiplexed Imaging

Abstract Recent advances in multiplexed staining and imaging technologies promise to significantly improve the understanding of the functional states of individual cells and the interactions between the cells that comprise complex tissues. This often requires compilation of results from multiple samples. Quantitative integration of information between samples is complicated by variations in staining intensity and background fluorescence due to differences in fixation and staining between samples, and other technical artefacts that obscure biological variations. Failure to remove these unwanted artefacts will complicate downstream analysis and diminish the value of multiplexed imaging for clinical applications. Here, we describe a Robust intEnSiTy nORmalization mEthod (RESTORE) to compensate for unwanted variations in multiplexed imaging. To do this, first, we automatically identify negative control cells for each marker within the same tissue and then use their expression levels to infer background signal level. Second, the intensity profile is normalized by the inferred background level of the negative control cells to remove between-sample variation. We evaluated the performance with theoretical simulations and tested the method using two different datasets; one comprised of multiplex images of three adjacent sections cut from a tissue microarray (TMA) carrying 59 cores from separate breast cancers and stained on different days, and one comprised of a pair of longitudinal biopsies obtained from one patient where tissue biopsy, fixation, processing and multiplexed staining/imaging were done at different times. We demonstrate that RESTORE can remove unwanted variations effectively and shows robust performance. Our code and the demonstration datasets are freely available for download at https://gitlab.com/Chang_Lab/cycif_int_norm.