Machine learning to decipher temporal & spatial determinants of cancer growth

Cancer follows an evolutionary process shaped by the environment of the tumour. Deciphering which genes and processes enable tumour cells to adapt to changing environmental conditions and cancer therapies remains challenging. Barcoded single-cell RNA-sequencing (scRNA-seq) links gene expression to competition between clonal populations of tumour cells to reveal how tumours adapt and evolve. Extension of barcoding to spatial transcriptomics (ST) enables study of how microenvironmental factors impact clonal dominance and transcriptional regulation. Machine learning (ML) techniques are ideal for barcoded scRNA-seq and ST applications because they integrate extensive, diverse data and link them to specified outcomes or properties of interest. This project develops ML methods to detect genetic and transcriptional features that enhance tumour cell survival from barcoded single-cell and spatial transcriptomics data. The new methods will be applied to identify drivers of disease progression and therapy resistance in leukaemia, melanoma and brain cancer.