UKINETS2018 Oral Communications (1) (3 abstracts)
PUnNETS (Prediction of Unknown Neuroendocrine Tumour Site) – a DNA methylation-based classifier
Alison M Berner 1 , Christodoulos Pipinikas 1 , Anna Karpathakis 1 , Harpreet K Dibra 2 , Ismail Moghul 1 , Amy Webster 1 , Tu Vinh Luong 3 & Christina Thirlwell 1,
1UCL Cancer Institute, London, UK; 2University of Birmingham, Birmingham, UK; 3Royal Free Hospital, London, UK.
Neuroendocrine tumours (NETs) of unknown primary (UP-NETs) represent up to 22% of NETs. Primary site identification enables patients to access appropriate treatment but is not always possible by immunohistochemistry or imaging. Given the epigenetic dysregulation of NETs, we aimed to use methylation array data to determine UP-NET tissue-of-origin. DNA from formalin-fixed paraffin embedded tissue from 76 pancreatic NET (PanNETs) (54 training and 22 validation samples) and 53 small intestinal NET (SINETs) (43 training and 10 validation samples), were run on the Illumina 450K methylation array. Probe filtering and normalisation were performed, including removal of probes not on the Illumina EPIC array, to allow use with EPIC array samples. Differentially methylated probe (DMP) analysis was performed for PanNET and SINET training samples. Methylation values for these DMPs were used as input for an ensemble learning support vector machine (SVM) training algorithm with 70 ensembles and 250 bootstrap reanalyses. The resulting PUnNETS (Prediction of Unknown NET Site) classifier was then used to predict the tissue of origin for the validation samples. 594 DMPs with differential methylation >40%, P<0.001, enabled separation of training samples by tissue of origin using hierarchical clustering and were used for classifier input. The PUnNETS classifier had an average accuracy of 100% in the training set and 98.75% in the test set (derived from multiple divisions of the training set). PUnNETS accurately predicted the tissue of origin for all 32 validation samples with confidences scores 68.6–100%. PUnNETS uses methylation array data to predict the tissue-of-origin UP-NETs with a high level of accuracy and following further validation has the potential to be used in clinical practice.
Volume 60
Article tools
My recent searches
My recently viewed abstracts
Authors
Endocrine Abstracts
ISSN 1470-3947 (print) | ISSN 1479-6848 (online)
© Bioscientifica 2024 |
Privacy policy |
Cookie settings
BiosciAbstracts
Bioscientifica Abstracts is the gateway to a series of products that provide a permanent, citable record of abstracts for biomedical and life science conferences.
Find out more