199P - Scalable multiplexed image analysis across cancer types as part of the IMMUcan consortium

Background

Highly multiplexed tissue imaging (HMTI) technologies have enabled the in-depth characterization of the tumor immune microenvironment (TIME) and how it relates to patient prognosis and treatment success. To analyze the TIME across breast, head and neck, colorectal, lung, and kidney cancer, the Integrated iMMUnoprofiling of large adaptive CANcer patient cohorts (IMMUcan) consortium performs broad molecular and cellular profiling of more than 2500 cancer patients, associated with longitudinal clinical data. As part of IMMUcan, multiplexed immunofluorescence (mIF) imaging detects major cell phenotypes across whole tissue slides while imaging mass cytometry (IMC) provides a zoomed in view on local cell phenotype interactions.

Methods

We developed two computational pipelines to process and analyze mIF and IMC data. The analysis of mIF data is supported by IFQuant, a web-based tool for image analysis that facilitates user-guided cell phenotyping. IMC data is analyzed using a workflow encompassing image analysis and machine learning-based cell phenotyping. Various computational approaches have been developed for the spatial analysis of HMTI data. First, the tysserand package allows the construction of spatial networks from which the mosna package computes cell phenotype assortativity scores. Second, the Spacelet and Cellohood models were developed to analyze higher-order interactions between cell phenotypes in tumor tissues.

Results

Based on a selected set of ten samples from five cancer indications, we validate that all major cell phenotypes as detected by the two computational pipelines show high correlation between mIF and IMC. The computed assortativity measures can be used for the definition of cellular niches and patient stratification. The Spacelet and Cellohood models allow the association of gradients of immune infiltration patterns and cellular neighborhoods to clinical data.

Conclusions

We have developed scalable computational tools for the analysis of mIF and IMC data within the IMMUcan consortium. These form a crucial foundation for the robust extraction of molecular features across five cancer indications facilitating patient stratification and clinical association studies.

Legal entity responsible for the study

EORTC.

Funding

IMI2 JU grant agreement 821558, supported by EU’s Horizon 2020 and EFPIA.

Disclosure

All authors have declared no conflicts of interest.