Immune cell gene signatures for profiling the microenvironment of solid tumours

The immune composition of the tumour microenvironment has been shown to regulate processes including angiogenesis, metastasis and the response to drugs or immunotherapy. To facilitate the characterisation of the immune component of tumours from transcriptomics data, a number of immune cell transcriptome signatures have been reported, i.e. lists of marker genes that together are indicative of the presence a given immune cell population. The majority of these gene signatures have been defined through analysis of isolated blood cells. However, blood cells have been shown not to reflect the differentiation or activation state of similar cells within tissues, including tumours, and consequently perform poorly on tissue data. To address this issue, we generated a set of immune gene signatures derived directly from tissue transcriptomics data using a network-based deconvolution approach. We define markers for seven immune cell types, collectively named ImSig, and demonstrate how they can be used for the quantitative estimation of the immune content of
tumour and non-tumour tissue samples. The utility of ImSig is demonstrated through the stratification of melanoma patients into immuno-subgroups of prognostic significance and the identification of immune cells from single-cell RNA-Seq data of derived from tumours. ImSig is available as an R package (‘imsig’).