Development of a deep learning-based approach to characterise the immune landscape in liver cancer
Development of a deep learning-based approach to characterize the immune landscape in liver cancer
Liver cancer, worldwide, is the third leading cause of cancer mortality. Most of the primary liver cancers are hepatocellular carcinoma (HCC). In the past 10 years, increased prevalence of obesity and type II diabetes led to an acceleration of HCC in the UK which is projected to rise by 38% in the next 20 years1. Advanced stage HCC are aggressive with a poor long-term prognosis - up to 70% of patients relapse within 5-year post surgery mainly due to persistent inflammation. While recent studies have provided an in-depth characterization of the immune landscape in lung and colorectal cancer, HCC is understudied and the role of human natural killer (NK) cells – particularly abundant in the liver tissue where they represent 30-50% of resident lymphocytes- is not fully understood.
NK cells are innate immune cells increasingly recognized for their potent anti-tumour function and multiple advantages as immunotherapeutic tool. However, several preclinical studies have evidenced that bona fide anti-tumour effectors such as NK cells and also NKT and CD8+T cells can promote tumour growth in causing tissue damage that drives tumorigenesis in this type of cancer. This paradoxical role highlights our lack of mechanistic understanding of this typical model of cancer driven by chronic inflammation, delaying the testing of a panel of IO approaches currently trialled against other types of cancer.
Thus far, we are limited by current approaches to study the relationship between inflammation, tumour molecular alterations and progression in human tumours in general and HCC in particular. It is becoming increasingly evident that we need to develop novel, spatially explicit computational and experimental approaches to study the complex cancer-immune interactions.
The aim of this project is to apply deep learning methods to analyse digital pathological images, map the spatial distribution of key effector cells and infer spatial patterns in both the tumour microenvironment (TME) and NTME; and integrate such spatial phenotypic data with transcriptomics data in HCC. This will be done using datasets from both human and translational mouse models of HCC in the following 3 steps:
1. Characterise the spatial organisation of NK and T cells in preclinical models of HCC on the immune landscape.
2. Relate the lymphocyte spatial patterns from mouse model to those in human HCC
3. Test how image-based lymphocyte spatial patterns are associated with the expression of key receptor/ligands pathways with HCC patient’s overall survival
This is a highly multidisciplinary project linking cancer immunology with bioinformatics through the integration of digital pathology, genomics and machine learning.
1. Tumour Immunology
2. Deep learning
3. Mouse models
4. Nk cells and T effector cells
6. Liver cancer
Convergence Science PhDs cover tuition fees for UK/EU students only.
Applicants with a Biology or Biochemistry degree should have theoretical and practical knowledge in bioinformatics.