Machine Learning for Spatial Transcriptomics Analysis in Human Cancers: Tackling Limited Data Challenges

This project addresses the pressing challenge of data scarcity in spatial transcriptomics (ST), crucial for advancing cancer research.

ST data, revealing spatial and gene expression information in cancer cells, holds transformative potential for understanding cancer biology.

However, its advancement is hindered by data scarcity, which restricts the application of advanced machine-learning techniques in ST studies.

Our approach introduces three innovations: developing sparse Bayesian learning algorithms for efficient small dataset analysis, designing a simulator for generating synthetic data and developing multimodal foundation models that integrate spatial, histological, and gene expression data.

By addressing data scarcity, this initiative aims to promote significant impacts on both research and clinical oncology.

This research aligns with EPSRC's "Healthcare Technologies" theme and strategic priorities in "artificial intelligence, digitisation and data: Driving Value and security" and "transforming health and healthcare"