Abstract

The tumour microenvironment is a mixture of heterogeneous cell types including cancer, immune, and stromal cells. Recent studies have shown that the interaction between tumour and non-tumour cells influences the fate of the tumour. For example, we have previously shown significant variability in tumour-immune microenvironments among different tumours within the same patient leading to differing outcomes (Jiménez-Sánchez, et al. Cell 2017). However, the immune cell type characterisation of tumours using bulk gene expression data is still challenging. Thus we propose the following well-defined project: Can we reliably estimate the relative proportion of different immune cells present in a tumour using gene expression data? Multiple groups have tried to address this question, however the current state-of-the-art methods present at least one of the following issues: a) inaccuracy, b) cell bias, c) not comprehensive enough. We have already derived consensus gene signatures that outperform previous methods when benchmarked using a small set of tumour tissue samples. However, we want to make a more thoroughly benchmark and refine the consensus signatures/method (optional). Therefore, the specific aims in sequential order for the project are: 1) Compile benchmark data used by the other methods, 2) benchmark our consensus method against each of the other methods using their own test data, 3) (optional) refine/improve consensus method and signatures, 4) (optional) search and compile independent test data fro additional benchmarking. If aims 1 and 2 are completed and there is time, then aims 3 and 4 could be explored, however completing aims 1 and 2 would be enough to submit for publication.