Academic research community (GECIP)

We would like to make use of variant data across all tumours sequenced by Genomics England in order to learn more about the patterns of gene alterations across tumour types. This can hopefully increase our understanding in how different cancers develop and highlight the potential value of utilising therapies originally envisaged for one tumour type across additional cancers.

A fusion gene is a gene made from joining parts of two different genes. Fusion proteins that are made from the 'instructions' given by these fusion genes can lead to development of some cancers. We are investigating a type of fusion gene that is shown to be involved in cancer activity across a variety of cancer types. At this time there is no approved therapy for treatment of patients with this specific type of gene fusion. Through the Genomics England database, we are interested in understanding the proportion of patients whose cancer contains this specific fusion and to learn more about the characteristics of patients that have this fusion (such as sex and cancer type). The study will also identify how many patients have one of two other well-known types of gene fusions that cause cancer, and use this as a “positive control” in order to understand how accurately fusions can be identified in the Genomics England database. This is a descriptive study, a study that describes the characteristics of the population but does not make comparisons and does not look at the causal questions such as risk factor of the disease. The results may help to design future clinical trials of new therapies in cancers that have this particular gene fusion, such as which patients to recruit to the trials and how many patients to recruit.

As we get older our blood cells acquire genetic variants. The presence of such blood cells may influence cancer risk and survival. This project aims to find out more about the presence of these abnormal blood cells on cancer risk and survival.

The diagnosis and treatment of certain cancers requires accurate assessment of DNA throughout the whole tumour. New DNA sequencing technology offers the opportunity to improve upon existing technologies to determine the overall DNA structure of the tumour. This project will compare such technologies with the aim of improving the diagnosis and treatment of certain types of cancer including leukaemia.

Early detection of newly diagnosed cancer and relapse leads to improved patient outcome. A promising technology to detect cancer early, is to identify cancer associated DNA (so called circulating tumor DNA) in a patient’s blood sample. Circulating tumor DNA (ctDNA) is a component of cell-free DNA (DNA no longer contained within the cell but floating about) that is shed by malignant tumours into the bloodstream and other bodily fluids. In prior work, the UCL/ ArcherDx team demonstrated the utility of ctDNA for minimal residual disease detection in non-small cell lung cancer patients (Abbosh, et al., Nature. 2017; 545(7655):446-451 and Abbosh, et al., AACR presentation Apr. 2020, Manuscript in preparation). Compared to standard radiological scan, disease relapse can be detected up to a three years earlier via ctDNA profiling. This approach creates a window to treat the disease earlier and obtain improved survival. This project aims to identify a panel of genes which can be applied across a broad population (eg. a standardized, off-the-shelf type solution) to give an accurate diagnostic signal of early / relapsed cancer, as can be detected via ctDNA, picked up in a simple blood test.. The manufacture and production of an off-the-shelf cancer monitoring solution can be standardized, drastically reducing the cost of such a technology, as a personalized panel would not be required for each patient and the patient’s tumor would not need to be sequenced up front, so a standardized cancer monitoring system would allow patients to receive a cancer monitoring readout more quickly. Variant calling files (VCFs) from the National Genomics Research Library, summarising variation in the genomes of participants in the 100,000 Genomes Project, will be analysed to identify unique genomic variants in low noise genomic positions (short stretches of highly stable/rarely varied genome). The list of genomic variants will be optimised by enrichment for genetic sequences that are commonly associated with cancers and depletion of commonly detected variants associated with the aging of the immune system. Data will then be separated into training and test sets, so that the sensitivity and specificity of a diagnostic tool can be tested to assess its ability to detect cancer across various cancer types (starting with non-small cell lung carcinoma, NSCLC).The project will be a partnership between a medical genetics diagnostic company called ArcherDX located at 2477 55th Street, Suite 202, Boulder, CO, 80301 USA, and University College London (UCL).

Patients with suspect solid tumours undergo biopsies to retrieve part of the tissue for analysis. Our growing understanding of the variants underpinning cancer has led to a more personalised approach to treatment planning and the use of therapies targeted to a patient's specific variants. As the patient's DNA is extracted from the biopsy for sequencing, it is important that this tissue is preserved in the highest quality. In the clinic, most biopsies are stored as formalin-fixed paraffin-embedded (FFPE) samples, which are cheaper, easier to use and more accessible. However, this process damages the DNA, making it much more difficult to sequence and interpret. Freezing the biopsy and storing it as a flash-frozen sample preserves the DNA without damage but this is much more expensive and impractical for clinics in the long run. In the 100,000 Genomes Project (UK100kGP), whole genome sequencing (WGS) is performed on flash-frozen tumour samples from NHS patients to study cancer driver variants and mutational signatures. In this project, we will explore whether the same information can be found in the UK100kGP FFPE samples. We will aim to answer the question: if we only have access to FFPE samples, what information can we extract reliably? We will also investigate how to distinguish between true variants and ones introduced by FFPE-related DNA damage. Through this project, we will learn how best to use FFPE samples for reliable mutational analysis with the aim of making targeted therapies more accessible for all NHS patients.

This research project is approved, but is not approved for publication. The KRAS gene is mutated in approx 15% of all human cancers, making it one of the most frequently mutated oncogenes. In pancreatic cancer and Non-Small Cell Lung Cancer (NSCLC), approx. 90% and 30% of cases in Western populations, respectively, have an aberration in the KRAS gene. In NSCLC, the most common form of lung cancer, there are several subgroups within the cohort of KRAS mutants defined by additional variants in other genes, for example TP53 and STK11. We would like to determine whether non-coding variants (ie variants not changing the amino acids directed by genes) are enriched in KRAS mutant cancers. The initial focus of the proposal is on colorectal and pancreatic cancers in addition to NSCLC. Based on preliminary findings, we may however extend the analysis to additional indications. We plan to annotate non-coding variants with state-of-the-art algorithms to differentiate driver from passenger variants (changes that are present but do not act as drivers), therefore potentially discovering additional subgroups within KRAS mutant cancers that we might link to additional readouts in subsequent steps.

This research project is approved, but is not approved for publication. We propose to combine calculation of genetic risk metrics (measuring each participant’s genetic predisposition for a set of obvious, observable, and measurable traits ) with a very large set of patient medical record data to identify characteristics of each disease and the mechanisms, outcomes, and clinical progression associated with each. We propose to investigate in the following areas. Method details are listed in the technical summary section. 1. Compute genome-wide predictions for a panel of relevant traits: Genome-wide Polygenic Risk Scores (PRS) reflect a quantitative measurement of the tendency to develop certain phenotypic traits, e.g., certain blood biomarkers, by aggregating across usually a large number of genetic loci that are known to be related to a specific phenotype in a weighted manner. PRS study can shed light on the disease causal mechanism from the commonly occurring genetic variants. Normally these kinds of genetic loci will have very little impact on the phenotype individually. However, combined together, these loci can help us disentangle the molecular mechanisms behind certain diseases. 2. Analyze the genetic variants/variants that cause a potential loss of function or a gain of function of the protein products within a given disease cohort. While certain genetic variants have minimal functional impact individually (see bullet 1), other variants significantly impact the function of the protein products of disease-relevant genes. These types of variants are predicted Loss of Function variants (pLoF) or Gain of Function variants (pGoF). Individual LoF/GoF variants repeatedly occurring on the same gene or gene pathways across patients with similar disease phenotypes provide evidence of the involvement of those molecular processes in the etiology of disease. PRS analyses (bullet 1) coupled with analysis of pLoF/pGoF variants will give us a robust picture of likely molecular causes of the disease of interest. 3. Focus on somatic tumour mutation profiling results provided by GEL whole genome sequencing data and investigate known and novel patient subgroups based on the small genetic variants, copy number variants and large structural genetic variants currently provided by GEL bioinformatics pipeline. Leveraging the rich patient phenotypic data within GEL, identify the association between patient subgroups discovered in GEL and phenotypic characteristics, to better understand the disease mechanisms and provide insights on actionable drug discovery plans. 4. Take the insights from the 100,000 Genomes Project participant data and combine with insights from other datasets we work with (for instance adding detailed medication information or rich biomarker data, to supplement the genetic data in Genomics England). Pooling all of this together helps us run our advanced algorithms to produce more detailed biological and therapeutic insights related to diseases of interest. Ultimately, the goal of this work is to match the right patients to the right therapies at the right time, and to identify unmet needs and associated potential points of intervention. 5. Engage with patients directly. To bring this to fruition we have engaged and intend to engage with clinical experts and patient groups in critical areas of unmet need. We believe that we must not only identify disease-modifying interventions targeting the mechanisms underlying each patient's specific disease, but that we must also understand the manifestation and impact of these mechanisms on each patient's journey, for instance via devices and patient-reported outcomes, in order to better measure, model, and treat the outcomes that matter most to each person in the context of their lives and their disease.

Specific differences in cancer across ethnic groups have been reported. For example, survival can differ in different ethnic groups. We would like to further understand this by looking for differences at the molecular level (for example, in DNA) and in other factors such as gender, age at diagnosis and cancer stage. We are particularly interested in the types and patterns of mutation present in inherited DNA and the DNA present in tumours. We will use genetic information and self-reported ethnicity to determine ethnic origins and then examine whether cancer features (clinical, pathological, molecular, etc.) differ by ethnicity and, if so, determine any implications for cancer management.

One of the main considerations when conducting any studies using databases (such as Genomics England) is to understand how the patient population being studied in Genomics England may over represent or underrepresent certain characteristics (such as cancer stage at diagnosis) in comparison to the true, underlying population. For example, Genomics England may contain more patients diagnosed with earlier staged cancers than late stage advanced cancers when compared to the prevalence of early stage vs. late stage cancers in the UK. Cancer stage at diagnosis is an important characteristic, since treatment decisions are heavily dictated by whether a patient has early stage vs late stage cancers. Understanding how the spread of stages in the cancer population from Genomics England compare to the spread of stages in cancer from the larger UK population will help researchers interpret and discuss results and findings for studies conducted using the Genomics England cancer data in the proper context. We propose to conduct a study where we look into patients enrolled in the cancer program in Genomics England and understand how many patients are diagnosed with specific stages (stage I, II, III, IV) among the different cancer types.

One of the reasons why cancer develops is because of changes in a cell's DNA, caused by things like tobacco smoking or errors in the body's normal repair system. Particular patterns of variants can give clues towards the factors that have contributed towards a cancer's development, such as smoking, and these patterns are called "mutational signatures". Several of these mutational signatures have been described, including signature 11 (SBS11) which is thought to result from the use of a particular group of chemotherapy drugs, in particular one called temozolomide (TMZ). It is not entirely clear why some patients who have temozolomide develop SBS11 and some don't. This may be partly due to problems with the body's DNA repair system, but we are not sure if this is necessary for SBS11 development, or whether people can get SBS11 without this repair deficiency. Furthermore, it's not completely clear whether SBS11 only occurs after temozolomide use, or whether similar chemotherapy drugs can cause the same pattern. We would like to study this further by using information from patients who have SBS11 and were treated in Manchester. We would like to look at what chemotherapy they had received, and how long this was before they had a sample of their tumour taken. We would also like to look to see if they have any variants in their body's repair system. We hope that this would give us more information about which chemotherapy drugs and which repair deficiencies can cause SBS11. We think that by understanding this better, we may be able to use this information to inform treatment decisions in the future. For example, patients who show SBS11 may not benefit from further treatment with temozolomide.

This research project is approved, but is not approved for publication. The NGRL includes a rich set of data of those who have had treatment for cancer. We would like to look in more depth at the data relating to germline genetic variation, tumour genetic variation, and response to treatment, for a series of de-identified individuals across a range of cancers, including colon cancer and lung cancer. Our first step is to evaluate the NGRL data for variety of cancer types, including what treatments were received and patient outcomes after treatment. Next, we will proceed with research study design and potentially genetic association evaluation based on the results of summarizing NGRL data. The clinical data may be incomplete for some individuals, thus data characterization as the first step is critical for determining our future research study design and what analyses we ultimately perform. Through better understanding of the relationship between germline genetic variation, tumour genetic variation, and response to treatment, we have the opportunity to identify new drug targets as well as genetic variations that identify individuals more likely to have successful treatment outcomes and a smaller risk of adverse effects with specific drugs.

The data of nearly 20,000 cancer patients is included in the NGRL. In this project, we will use sequence data to evaluate the performance of the pipelines used for analysis of these genomes and return of results to clinical specialists. We expect that this evaluation will also inform scientists who conduct research using cancer data in the NGRL on the accuracy of the genomic data in it.

This research project is approved, but is not approved for publication. The MET inhibitor capmatinib is approved in the US and Japan for treatment of lung cancer with MET mutation exon 14 skipping variants. These variants result in an altered and over active form of the MET kinase, which drives cancer growth. We are interested in gaining a comprehensive overview of other genomic alterations of the MET gene that occur in cancer in order to identify additional cancer patients who might benefit from treatment with capmatinib.

Many aspects of a person’s lifestyle affect their risk of developing cancer. Some of these risk factors are well understood, such as the links between smoking and lung cancer, and UV light exposure and melanoma. Other potential risk factors are less well characterized, partly because it is often not possible to accurately measure a risk factor across a person’s life. For example, obesity can be quantified at the current time by calculating BMI, but many people do not have accurate records of their BMI spanning back decades. This information can be crucial as some cancers start developing years before diagnosis. Inaccurate risk factor measurements can lead to bias, which may result in factors being incorrectly linked, or incorrectly not linked, to cancer risk. Germline genetic variants are those variants we inherit from our parents, and have been associated with many cancer risk factors. These germline variants provide an alternative way of measuring a person’s exposure to a risk factor across their life, and can therefore be used to test whether a factor affects cancer risk. The 100,000 Genomes Project has sequenced both germline variants and variants that occur in tumours, allowing us to investigate whether and how potential risk factors influence cancer development.

Many cancer therapies damage DNA, leading to variants in both cancerous and healthy cells. These variants may result in long-term side effects, including an increased risk of other cancers. The contribution of variants caused by chemotherapy and radiotherapy to the risk of a patient developing another cancer is not yet fully understood. Analyzing whole genome sequences of cancer patients who have previously undergone cancer therapy will allow us to better characterize the impact of these treatments on a person’s genome and better assess their long-term risk.

Chromosomal rearrangements can lead to formation of oncogenic gene fusions. Identifying the fusions capable of producing a chimeric protein, a hybrid protein with a new function, especially those involving genes encoding various tyrosine kinases (e.g. ALK, FGFR, NTRK gene families), can be important in cancers with high genome instability such as sarcomas as they offer opportunities for targeted therapy applications. In this project we attempt to predict pathogenic potential of chromosomal rearrangements derived from WGS data

Immunotherapy is a ground breaking treatment for cancer. Yet despite many patients seeing their cancers cured or in remittance, many do not. In this project we will focus on how the immune environment differs across cancer types. By examining these differences we aim to explain some of this variety in response. We will focus on the the number of variants present that the immune system recognises as well as examining known mechanisms that cancer can evade and escape the immune system. We are able to use the sequencing data within the NGRL to quantify the type and number of immune cells. We will combine this analysis with images showing the immune state of the cancer. This will confirm the results genetic analysis. Finally we will examine clinical data showing response to immunotherapy. We will confirm these results on separate data sets for many different cancer types.

Cancer variants run the gamut from small alterations involving few DNA bases to complex variants involving large genomic regions. Small alterations are easy to detect using current DNA sequencing technologies (i.e. short-read or Illumina sequencing). However, complex DNA alterations cannot be fully characterized using mainstream DNA sequencing techniques. Therefore, the functional consequences of these variants and their clinical implications remain largely uncharacterized, although increasing evidence suggests that they play a crucial role in cancer evolution and drug resistance. In this project, we will develop computational tools to analyse data from sequencing machines whose unique technology permits to resolve these complex alterations by continuously reading long stretches of DNA; hence these methods are collectively known as "long-read sequencing". These emerging technologies have enormous potential to resolve complex DNA alterations, and thus, help us better understand their clinical implications, and more broadly, the genomic architecture of human tumours. We will compare short and long-read sequencing data generated for the same tumour specimens to assess and compare the diagnostic power and potential to resolve DNA alterations of both technologies. Although we will primarily focus on the analysis of brain tumours, sarcomas, and hypermutated tumours, the tools we will develop will be applicable across diverse cancer types, and will lay out the foundation for the application of long-read sequencing methods in clinical settings, as well as the development and validation of additional algorithms.

Understanding cancer heritability has a potentially enormous impact for disease prevention, early detection and tailored treatment. GWAS based on single nucleotide polymorphisms only explain a modest fraction (2-15%) of heritable variation in cancer risk. This 'missing' heritability could reside on understudied non-coding variability, structural variations (SVs) including copy number variations (CNVs). So far, analysis of these features has been limited by the lack of large whole genome sequencing dataset and poor availability of methods to detect and investigate structural variation for association purposes. We propose to utilize pan-cancer data from the NGRL in order to firstly investigate the effect of non-coding variants in the heritability of cancer and then to develop computational methods to analyse the impact of structural variation and copy number variation in cancer heritability.