ABSTRACT
Multiplexed imaging technologies enable the study of biological tissues at single-cell resolution while preserving spatial information. Currently, high-dimension imaging data analysis is technology-specific and requires multiple tools, restricting analytical scalability and result reproducibility. Here we present SIMPLI (Single-cell Identification from MultiPlexed Images), a novel, flexible and technology-agnostic software that unifies all steps of multiplexed imaging data analysis. After raw image processing, SIMPLI performs a spatially resolved, single-cell analysis of the tissue slide as wells as cell-independent quantifications of marker expression to investigate features undetectable at the cell level. SIMPLI is highly customisable and can run on desktop computers as well as high-performance computing environments, enabling workflow parallelisation for large datasets. SIMPLI produces multiple tabular and graphical outputs at each step of the analysis. Its containerised implementation and minimum configuration requirements make SIMPLI a portable and reproducible solution for multiplexed imaging data analysis.
SIMPLI is available at: https://github.com/ciccalab/SIMPLI.
Competing Interest Statement
The authors have declared no competing interest.
Footnotes
To further clarify the extent of methodological advance and the advantages of SIMPLI compared to existing counterparts, we revised Introduction, Table 1, Fig.1, Supplementary Fig.1. We compared the analysis of normalised and raw images, p.11-12, Supplementary figure 2. We reassigned cells in the lamina propria varying the mask overlap, p.12, Supplementary figure 2. We added another segmentation method and compared the results with the one already implemented in SIMPLI, p.17, Supplementary figure 3. We compared the cell phenotypes from unsupervised clustering at various resolution as well as with those from thresholding, p.17-18, Supplementary figure 3. We repeated the spatial analysis between PD1+CD8+ T cells and PDL1+CD68+ macrophages after re-identifying the latter with the thresholding approach, p.21-22. We revised the heterotypic spatial analysis using more restrictive cut-offs and adding a permutation test to strengthen the results, p.27. We expanded and clarified the modularity in the choice of analysis methods in terms of: Cell segmentation (conventional vs deep learning), Phenotyping (unsupervised vs thresholding), Spatial analysis (homotypic vs heterotypic). These are now described in the text (p.7-9) as well as in the revised Figures 1 and S1. We added further recommendations on the parameter choice in the software documentation and in the Methods and discussed the method's limitations (p.31). Finally, we added the Data Availability and Code Availability sections (p. 42), deposited all new data in Zenodo (Access Codes provided in the text).