A tumour comprises not only cancer cells but other populations, including immune and stroma, collectively termed the tumour microenvironment (TME). Single-cell RNA sequencing has offered unparalleled insights in different and rare cell populations. However, bulk RNA-seq remains a cost-effective and widely-used approach to analyse large clinically annotated cohorts. Given the ever growing volume of transcriptomics data, computational methods for estimating immune cell proportions and studying tumour immune dynamics have gained prominence.

This thesis presents novel bioinformatic approaches for studying the TME, yielding insights into how the complex cellular landscape can be altered by cancer and therapeutic intervention. Central to this work is the development and refinement of Consensus^TME, a novel ensemble computational method that estimates immune and stromal cell proportions in the TME from bulk transcriptomics data. This method outperforms existing approaches, ranking in the top three for all cancer-related benchmarks.

A key contribution of this research is the comprehensive comparison of various computational tools used in TME analysis, establishing Consensus^TME as a robust method. Enhancements to this tool include refined human gene sets and the ability to estimate immune infiltration in mouse models, broadening its applicability. The dedicated web resource, www.ConsensusTME.org, and R package, “ConsensusTME”, facilitate its use. The web resource further provides additional deconvolution resources to aid development and vigorous benchmarking of new methods.

The biological impact of this work is showcased through case studies, particularly in the context of pancreatic and ovarian tumours. The findings reveal how treatments, such as chemotherapy or hedgehog inhibition, dynamically alter the TME. Additionally, the application of the 10x Visium platform provides new spatial perspectives on TME features, underlining both the utility and limitations of current methodologies.

In summary, this thesis advances our understanding of the TME through bioinformatic innovation but also offers practical tools and insights for future cancer research. Potential therapeutic implications of these findings underscore the importance of understanding the TME for continued development cancer treatment strategies.