Join us for this seminar by Simon Pearce (CRUK Manchester Institute) as part of the North West Seminar Series in Mathematical Biology and Data Sciences. Details of the full series can be found here https://www.cms.livjm.ac.uk/APMSeminar/
The talk will be hosted in person in the Simon Building, Room 3.62. For those who cannot attend in person the talk will also be streamed via zoom, please contact carl.whitfield@manchester.ac.uk or igor.chernyavsky@manchester.ac.uk for the zoom link, or sign up to the mailing list.
Abstract: Profiling of DNA methylation in circulating free DNA (cfDNA) from plasma for detection, identification and monitoring of cancer is extremely promising. This DNA occurs from natural cell turnover, but in patients with cancer some of the DNA comes from tumour cells, and therefore contains an epigenetic signature representing both the process of tumourigenesis and the cell-of-origin. We have established a novel method (T7-MBD-seq) for multiplexing DNA, followed by a Methyl-Binding Domain (MBD) protein enrichment to investigate genome-wide methylation patterns. We couple this with a data augmentation technique that involves in silico mixtures of tumour samples with non-cancer control (NCC) cfDNA to mimic the low proportion of circulating tumour DNA, as well as a novel method for converting publicly available methylation arrays to be comparable with T7-MBD-Seq data.
I will discuss two particular applications of this; Small Cell Lung Cancer (SCLC) and Cancer of Unknown Primary (CUP).
SCLC is a high-grade neuroendocrine carcinoma characterized by high proliferation and early metastatic spread. Although SCLC is treated as a homogenous disease, recent studies describe distinct molecular subtypes based on transcription factors, with evidence of differing therapeutic vulnerabilities. Liquid biopsies are particularly relevant to SCLC as clinical biopsies are scarce. We applied T7-MBD-seq to both SCLC patient-derived preclinical models and cfDNA from patients with SCLC, and generate a classifier able to discriminate patient cfDNA from NCCs, including 6/6 stage IA patients. Furthermore, we build a subtype classifier based on SCLC cell lines that correctly assigns all 33/33 of our preclinical models, as well as 10/11 of cfDNA samples where the subtype was known. Our data show potential clinical utilities of cfDNA T7-MBD-seq methylation profiling as a non-invasive, universally-applicable approach for sensitive detection and subtyping of SCLC.
Around 2% of all new cancer diagnoses in the UK are CUP; these patients present with metastatic tumours, but the primary origin of the cancer cannot be determined. Determining the primary tumour site may enable access to type-specific therapies, potentially leading to therapeutic benefit. We generate a tumour type classifier across 29 tumour classes, utilising array samples from The Cancer Genome Atlas (TCGA). On an independent test set of 170 cfDNA samples, comprising 143 patients with metastatic cancer across thirteen cancer classes and 27 NCCs, we find an overall sensitivity of 87.1% and specificity of 97.9%.
When applied to 41 patients with CUP, we predict a primary tumour type in 32/41 cases (78%). Of the patients where predictions are made, 75% were supported by the clinical evidence with subsequent primary or suspected/differential diagnoses. We show that we are able to accurately predict tissue-of-origin from a single liquid biopsy for many patients, opening stratified treatment options leading to potentially better clinical outcomes.
This seminar will discuss these two applications, as well as some of the open questions in how best to determine differentially methylated regions in samples with variable tumour content. I'm particularly keen to talk to people who have experience in the negative binomial generalised linear models (as used in DESeq2 for instance).
To subscribe to the mailing list for this event series, please send an e-mail with the phrase “subscribe math-lifesci-seminar” in the message body to listserv@listserv.manchester.ac.uk
