Background Recent advances in computational pathology and spatial transcriptomics have opened new opportunities to link tumour morphology with underlying molecular and cellular processes. However, these domains are often studied in isolation: deep learning models predict gene expression from histological whole slide images (WSIs), while spatial transcriptomic approaches characterise the tumour microenvironment (TME) without direct integration with histological features. A unified framework connecting tissue morphology, gene expression, and spatial organisation remains lacking. Aim This project aims to develop and apply integrative computational methods to jointly analyse histological images, gene expression, and spatial tumour microenvironment features in breast cancer. The central hypothesis is that tumour-intrinsic molecular phenotypes such as chromosomal instability and cell cycle activity, and tumour-extrinsic factors such as TME composition and spatial organization, can be inferred from tissue morphology and jointly contribute to clinically relevant outcomes. Methods Two complementary approaches will be combined. First, deep learning models (e.g. SEQUOIA, HE2RNA, and related architectures) will be applied to WSIs to predict gene expression profiles, focusing on biologically meaningful gene signatures such as chromosomal instability and cell cycle-related markers. These predictions will enable benchmarking of model performance using interpretable, clinically grounded metrics and assessment of associations with patient outcomes. Second, spatial transcriptomics data will be analysed to characterise tumour microenvironment composition and spatial relationships between cell populations. Tools such as METI and related spatial analysis frameworks will be used to quantify how immune and stromal cells are distributed relative to tumour cells and how these spatial patterns influence tumour proliferation and molecular phenotypes. Reasoning for Computational Requirements The project will leverage large-scale datasets (e.g. TCGA and spatial transcriptomics cohorts), requiring substantial computational resources for high-resolution image processing, patch-level model inference, and spatial modelling. High-performance computing infrastructure is therefore essential to efficiently execute these workflows. Expected Outcomes and Impact By integrating histology, gene expression prediction, and spatial analysis, this project will investigate how tissue morphology encodes both molecular tumour states and microenvironmental structure. The work aims to establish a unified computational framework for analysing tumour biology, improve biologically meaningful benchmarking of deep learning models, and contribute to the development of clinically relevant biomarkers in cancer.