Statistical Modeling and Analysis of Human DNA Methylation Data to Detect Differential Methylation in Osteosarcoma

• Activity Number: 209 - Statistical methods for genomic and epigenetic data analysis
• Type: Contributed
• Date/Time: Tuesday, August 10, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Sujay Datta*
• Companies: University of Akron
• Keywords: DNA methylation; Illumina; bisulfite; osteosarcoma; CpG island; statistical modeling

Abstract

This project is about statistical modeling and analysis of methylation data generated from human osteosarcoma samples by the Illumina Infinium Human-Methylation 450 DNA Analysis BeadChip genome-wide assay for the identification of differential methylation. It involves quantitation of the prevalence of methylated cytosine at 485578 individual CpG island sites in bisulfite-modified DNA samples from 16 primary human osteosarcoma tissue-samples and a non-neoplastic control sample that combines DNA from non-malignant bone (SCFE growth plate and supernumerary toe), cultured hMSC samples, adult skeletal muscle, fetal liver, adult adipose tissue and fetal brain. Data generated at the Akron Children’s Hospital in Ohio and extracted/preprocessed by Illumina’s proprietary software (GenomeStudio Methylation Module v1.8) were reported as difference scores D at the site in question for each OS sample compared to the average of all controls. We adopt 3 approaches for differential methylation discovery: (1) a predictive modeling approach to find out which CpG sites’ methylation status are significant, (2) a Hotelling’s T-squared test-based approach and (3) a principal components-based approach