Weighted learning for long-read DNA sequencing consensus methylation detection

Authors

Emma Wade

Document Type

Article

Publication Date

Summer 2022

Keywords

JGM

JAX Location

In: Student Reports, Summer 2022, The Jackson Laboratory

Abstract

DNA methylaion is an important epigenetic marker. With the release of Oxford Nanopore sequencing, novel methylation detection methods based on machine learning have been developed. However, those models are trained on fully concordant data or imbalanced data and are inefficient in methylation detection at regions with discordant non-singleton DNA methylation patterns, an area of special interest in cancer research (Figure S1). Here, a new model is built using weight to improve methylation predictions of discordant regions, a crucial step forward in cancer research. After prediction and evaluation, the addition of weight improved accuracy, correlation with BS-Seq, and coverage of ONT-methylation calling tools. The new models are a step forward to a more affordable alternative to bisulfite sequencing.

Please contact the Joan Staats Library for information regarding this document.

COinS