We have shown that a family of genes called killer immunoglobulin-like receptors (KIRs) significantly increases immune cell survival in vivo¹ and impacts human health during chronic virus infection and diabetes^2-4. This PhD project will test the hypothesis that tumors hijack the KIR-survival mechanism to enhance their own fitness. This hypothesis is supported by our striking observation that every KIR gene is under considerable positive selection in tumours⁵ and that KIR expression by tumors is associated with poor prognosis⁶. Knowledge of genes underlying cancer is critical for diagnostics and rational therapies. This proposal will contribute to our knowledge of cancer driver genes and the evolutionary processes underlying cancer development. 

This project will combine data science (including machine learning and survival analysis), bioinformatics and cancer genomics to address our questions. We will test our hypothesis in a large clinically annotated cohort of 3700 metastatic patients (from Hartwig Medical Foundation) as well as in other data sets. We will analyse sequence data, map every sequence to multiple polymorphism of the KIR gene family, detect homology and differentiate between true somatic mutations and paralogous sequence variants using machine learning methods. 

Skill learnt include: 

Handling large datasets 

Working on the high performance cluster 

Scripting in R 

Working with DNA sequence data and RNA expression data 

Learn the tools of cancer genomics and evolution 

Use of epitope prediction software 

Cox regression to analyse mortality 

Machine learning (likely to include random forest, naïve Bayes, elasticnet) 

1. Zhang, Y. et al. KIR-HLA interactions extend human CD8+ T cell lifespan in vivo. J. Clin. Invest. 133 (2023).  

2. Mora Bitria et al. Inhibitory KIRs decrease HLA class II-mediated protection in Type 1 Diabetes. bioRxiv, 2024.2003.2027.586933 (2024).  

3. Boelen, L. et al. Inhibitory killer cell immunoglobulin-like receptors strengthen CD8(+) T cell-mediated control of HIV-1, HCV, and HTLV-1. Sci Immunol 3 (2018).  

4. Seich Al Basatena, N.K. et al. KIR2DL2 enhances protective and detrimental HLA class I-mediated immunity in chronic viral infection. PLOS Pathog. 7, e1002270 (2011).  

5. Zapata, L. et al. Immune selection determines tumor antigenicity and influences response to checkpoint inhibitors. Nat. Genet. 55, 451-460 (2023).  

6. He, Y., Bunn, P.A., Zhou, C. & Chan, D. KIR 2D (L1, L3, L4, S4) and KIR 3DL1 protein expression in non-small cell lung cancer. Oncotarget 7, 82104-82111 (2016).  

Appearance of metastatic lesions, which is usually associated with therapy-resistance, is the main reason for failing to cure patients with cancer. Circulating tumour cells (CTCs) are the main path of cancer dissemination to distant sites, but specific treatment modalities that decrease their survival or seeding efficiency are not currently available. To develop such therapeutic strategies, one must first understand how cancer cells change their molecular make-up to survive and adapt to the stress of circulation. Here, we will build a microfluidics device that reproduces the variable pressure profile that cancer cells are exposed to in vivo. Cancer cells will be subjected to circulation in this device for variable periods of time and changes to their proteomics and transcriptomics profiles as compared to uncirculated cells determined using mass-spectrometry and RNA-Sequencing, respectively. Advanced data analysis techniques, including pathway inference, building of static functional interaction networks, and gene ontology analysis will be used to identify molecular changes that could be therapeutically targeted to overcome the stress-induced aggressiveness of CTCs. These inferred therapeutic strategies will then be tested in 2D and 3D in vitro assays for cell proliferation, invasiveness and drug response, as well as in circulation in our device, using genetic and drug based intervention. This research will identify novel targeted therapies to decrease the metastatic potential of cancer cells. Because of (1) the direct correlation between lung metastasis and the failure to cure patients and (2) the relative uniformity of genetic background in this disease which will simplify choosing representative cell lines for the proposed work, we choose osteosarcoma as a disease model for this project. The student recruited on this PhD will be trained in the building and operation of microfluidics devices, proteomics and transcriptomics profiling techniques, advanced integrative data analysis using the R and Python programming languages as well as functional Bayesian networks, and biological phenotypic assays such as growth, drug response and invasion assays using advanced 2D and 3D models. This will provide the recruited student with a solid grounding in translational multidisciplinary research which will prepare them for a career in convergence biomedical research. 

Our goal is to create an effective and personalised immunotherapy against pancreatic cancer, which is known as one of the most difficult cancer types to cure and is a cancer type with highest mortality rate. Pancreatic cancer is unresponsive to cancer checkpoint inhibitor immunotherapy due to few number of tumour-infiltrating immune cells and high collagen expression. Ishihara previously found that collagen binding domain (CBD) protein specifically accumulates within the tumour vasculature (but not within healthy vasculature) after intravenous injection. Ishihara created collagen binding cytokine IL-15 super agonist which can specifically proliferate anti-tumour T cells and recrute immune cells into tumour. Sadanandam lab has defined different cancer, fibroblast or immune subtypes using patient pancreatic cancer samples, and developed 18+ syngeneic mouse models to study immunotherapy in this cancer types. Our hypothesis is that collagen binding domain can target specific subtypes of pancreatic cancer well, and pancreatic cancer can be induced to become immune “hot” using the collagen binding cytokines or chemotherapy such that checkpoint inhibitor therapy will be effective. The student will analyse immune cells and molecular mechanisms after collagen binding cancer therapy to identify combination therapy. Then the pancreatic cancer models from Sadanandam lab’s repository to be classified to the potential therapies. We believe that this approach could be applicable for other cancer types to create tumour targeted and personalised therapy and can be an innovative way for effective and safe cancer medicine.  

The student will learn an interdisciplinary opportunity of bioengineering, immunology, in vivo modelling and bioinformatics. This team has successful trained multiple students with multidisciplinary skill sets before.   

Prostate cancer affects 1in 8 British men. Whilst many are realising a cure for their cancer as result of earlier detection, metastatic disease nevertheless remains a significant burden and challenge to treat. With multiple lines of treatment now available, there is a need to develop novel biomarkers that can predict an individual response to treatment and survival in the quest for a personalised approach to patient care. Prostate cancer has some of the highest circulating free DNA and Circulating Tumour Cell levels of all solid tumours allowing serial tumour genomic analyses during treatment. 

The androgen receptor and its signalling is a key driver of prostate cancer progression. The detection of AR variants and their associated modulatory microRNAs can potentially be exploited as a non-invasive biomarker in the selection and monitoring of treatment, elucidating drug resistance and clonal evolution. 

This PhD research will build already on the concept of micro-RNA and circulating free DNA arrays that Professor Bevan’s team has already initiated and will incorporate clinical assessments on men having treatment with chemotherapy and androgen receptor targeted therapy under the care of Dr Mangar bringing together clinical and translational research.  

The study will utilise a novel RTLAMP assay for measuring the expression profiles of two mRNA biomarkers: the androgen receptor splice variant 7 (AR-V7) and the Yes associated protein1(YAP-1). The assay will be incorporated onto a novel lab-on-chip (LOC) device utilising ion-sensitive field-effect transistors (ISFETs), to allow for the rapid detection of the biomarkers as an alternative to reverse transcriptase polymerase chain reaction assays which are time consuming and costly. Biomarker data analysed using the LOC platform will be collected from various prostate cancer cohorts, from those undergoing active surveillance to patients receiving chemotherapy and androgen receptor targeted agents. 

The research will involve transcriptomic data analysis evaluating the correlation of AR-V7 and YAP-1 mRNA profiles with clinical risk parameters including PSA values at diagnosis, clinical staging, and histological parameters.  The ratio of AR-V7 to the AR-FL transcript and YAP-1 mRNA profile will be correlated to clinical outcomes in patients with metastatic prostate cancer.  As such it will encourage learning opportunities spanning clinical, laboratory and bio-informatic disciplines. 

Inflammatory bowel disease (IBD) is a chronic condition that increases lifetime risk of colorectal cancer. To mitigate this risk, people with IBD are enrolled into surveillance programs whereby they undergo regular colonoscopy examinations to detect and remove cancer precursors (i.e.dysplasia). Yet there are challenges in current surveillance protocols and many IBD cancers are not prevented. Our team are interested in finding new biomarkers to risk-stratify IBD patients, so that those who are at high risk of cancer can be followed more closely, and those at low risk can be spared unnecessary and costly intervention. 

We have previously performed genomic analysis of IBD biopsies taken during routine colonoscopy, and shown that burden of copy number alterations is predictive of future cancer risk. This project proposes to build upon this knowledge, by examining routine histological images of these biopsies and using artificial intelligence (AI) methodology to search for predictive image-based features. We aim to generate a new, rapidly translatable tool for predicting future cancer risk from H&E images of the IBD bowel. 

This student will be based in the lab of Prof Trevor Graham at the Institute of Cancer Research in Sutton. The project will provide the student with training in cutting-edge computational analysis of histological images, as well as providing opportunities to generate new data in our wet-lab. Clinical supervision will be provided by co-supervisor Prof Ailsa Hart, who will provide access to patient material and associated clinical data. 

In this project we propose to develop high-resolution image-based profiling to develop early readout assays for CTC response to drugs. Image based profiling is prevalent using clinical imaging modalities such as MRI, CT and PET. Sophisticated AI techniques to aid in diagnostics have arisen owing to the rich high-dimensional data present within images that is difficult or not possible to detect with human perception. It is a less developed and explored area with regards to morphologically profiling primary cells using microscopy. Advanced image analysis can be used to extract cellular features that reflect cell response to treatments such as drug or genetic perturbations.  

Motivating this is the need to rapidly establish CTC response to anticancer therapeutics for clinical use. A significant challenge is the limited number of cells extracted from patients and the number required for traditional drug assays. We have developed novel microfluidic technology to measure on-chip drug response profiles of cells isolated from blood. For kinetically fast-acting agents such as cytotoxics, readouts can be made within 24-72 hours. For novel inhibitor class agents or cytostatics such as CDKi’s and PARPi’s these require assays that span 7-10 days or more. Readouts for inhibitor/cytostatic drug assays compare the growth/proliferation of cells under test to a control set hence why inhibition assays require the time they do to reliably determine whether inhibition has, or hasn’t, occurred. The objective then becomes how to reliably maintain CTCs ex vivo over these timescales. Poor probability of culturing CTCs aside, the time required to expand primary cultures does not support rapid assays that might help inform treatment decisions. We believe new readout methods that circumvent these issues are required for on-chip drug screening. 

As with its more established genomic/transcriptomic/proteomic counterparts, morphological profiling aims to generate bioactivity profiles (e.g. heat map of transcripts from a PCR array) that describe a perturbed cellular state or condition. We seek here to measure how morphological profiles arise over the length of a drug assay and determine if, and when, morphological profiling can be used in lieu of proliferation assays when testing cytostatic/inhibitor class drugs. By doing so, there is potential to screen patient-derived CTCs more rapidly and against clinically relevant compounds that are not currently possible with standard techniques. 

Brain tumours cause a range of difficulties for patients, including cognitive, neurological, 

and behavioural problems. In some patients, changes in these features over time may relate to tumour recurrence. Others make be linked to side effects of treatments. Importantly, the ability to automatically detect subtle changes within individual patients – even before a patient or caregiver has detected them – could improve long-term outcomes if early detection leads to earlier, more effective, treatment interventions. 

Recently, we have begun using long-term multi-modal wearable sensing and mobile app 

technology to gather longitudinal data in a cohort of more than 100 people living with brain tumours. These data include symptoms scores, quality of life measures, and detailed 

recordings of movement, speech, and cognition. It may be that, when combined with clinical and outcome data, we can develop highly sensitive measures of tumour recurrence. 

A potentially important missing piece of the puzzle is a more direct readout of brain function. Clinical electroencephalography (EEG) safely records the brain’s electrical activity from the scalp. Whilst traditional clinical EEG uses ~20 “wet” electrodes attached with conductive gel by a healthcare scientist, with advances in electronics there are now simpler ~2-8 “dry” electrode EEG headsets that can be easily attached by anyone. For our patient cohort, this development means we can add in regular (i.e., daily) self-initiated EEG recordings into our multi-modal data collection. The challenge, then, is to derive from these EEG signals objective metrics that can provide the most clinically informative information, e.g., that may be able to detect tumour recurrence. 

In this project, we will introduce wearable EEG headsets into a subset of our cohort, and 

use the methods of signal processing and machine learning to: (1) evaluate the signal:noise and recording characteristics of the wearable EEG devices; (2) examine the relationships between EEG signals and other data collected, e.g., measures of cognition and behaviour; (3) develop machine learning classifiers that can predict tumour recurrence, exploring which combination of data modalities (e.g., EEG + movement + cognition, EEG + movement, EEG alone, movement alone) yields the most accurate and practicable solution. 

Supported by a multi-disciplinary team of experts, the student will receive training in signal processing, neurophysiology, cognitive neuroscience, and machine learning, and gain practical experience in clinical trials of novel technologies. 

This is an exciting opportunity to use a convergence science approach to find the best ways to use new technology to improve outcomes for people living with brain tumours.

Brain tumours cause a range of difficulties – including cognitive, neurological, and psychiatric problems – and these change over time, linked to tumour progression and treatment interventions. It is challenging to account for these problems based on available data (e.g., brain scans), and to predict what will happen. 

We have pioneered the use of wearable sensing and mobile technology to gather multimodal data on symptoms and physical features in people with brain tumours. This includes data on speech and cognition, alongside scans and clinical data. However, what is missing is a unifying framework to integrate these data at the individual patient level to produce a clinically-informative tool. 

Recently, the advent of whole-brain mesoscopic generative simulations of brain activity has created a paradigm for integrating biological, neurophysiological, cognitive, and clinical data in individual patients, via the creation of “digital twins”, biophysically-inspired models of brain function, derived from differential equations underlying the biophysics of regional neural activity (“neural mass models”). This – importantly – incorporates individual-level data (e.g., from MRI, diffusion tensor imaging) and the model is fitted to a patient’s own neurophysiological data (e.g., EEG). This approach has recently been demonstrated in epilepsy surgery. 

Creation of digital twins unlocks several potential capabilities: Firstly, by modelling the generating process, rather than analysing the resulting observations, we can examine mechanistically how a tumour produces a patient’s problems. Only rarely is this mapping trivial (e.g., one tumour in left motor cortex producing exclusively right-sided weakness). Rather, a digital twin enables a complex systems approach to link tumour-related disruption to the brain’s structural connectome to abnormalities in the dynamics of large-scale functional brain networks important for different cognitive/neurological functions. Secondly, by incorporating biological information about spatiotemporal tumour spread, we can predict what problems are likely to develop. Thirdly, we can test treatment interventions in silico, addressing what would happen if the generative model were perturbed by, e.g., tumour surgery, thus providing a means to guide individualised therapies in a principled way. 

In this project, we will develop and validate the digital twin approach using our multi-modal data in people with brain tumours, acquiring high-density EEG to provide detailed neurophysiological signals. We will test whether digital twins can (1) explain the range of observed problems (2) predict the effects of tumour progression, and (3) predict the effects of treatment. Supported by a multi-disciplinary team of experts, the student will receive training in programming, computational and cognitive neuroscience, and machine learning.