Title: "DPT Approach: A New Strategy for Differential Analysis of Multiple Enrichment-based High-throughput Sequencing Profiles"
Speaker: Yaomin Xu, Department of Quantitative Health Sciences, The Cleveland Clinic, Cleveland, OH
Place: PHYS 223
Date: March 27, 2012, Tuesday, 4:30pm

Abstract

Comparing next-generation sequencing-derived enrichment profiles among multiple groups containing multiple samples remains an analytical challenge. Here, we propose a new statistical analysis strategy for multi-group comparisons of epigenomic profiles and provide an in-depth demonstration of this approach in identifying differentially methylated sites at a genomic scale. Our method combines machine learning and statistical modelling techniques with three integrated parts: signal Deconvolution, whole genome Pattern recognition, and differential Testing (DPT approach). In our demonstration data comparing normal colon and two subtypes of colon cancer tissue samples, the DPT analyses yielded a list of 20,246 differentially DNA-methylated sites (DMSs) with statistical significances. The DPT approach identified DMSs at high-spatial resolution and without the limitation on the positional shape whether it is a peak or a broad range. Additionally, under a regression framework, the DPT approach provided the flexibility to adjust for either genomic or phenotypic co-factors as covariates. Furthermore, our showcase analyses performed with high reproducibility between the demonstration dataset and an independent dataset containing normal and cancer colons. Finally, we provide validation of a set of randomly selected colon cancer DMSs using the gold standard capillary electrophoresis-based bisulfite sequencing, and these results confirmed the accuracy, consistency, and spatial resolution of the DPT approach. The DPT approach provides a breakthrough analysis strategy to facilitate the typical biological experiments containing relatively small sample sizes as well as the imminent large-scale clinical studies to understand the dynamic epigenomic variation and their regulatory functions to human diseases.

Associated Reading:
Yaomin Xu, Bo Hu, Ae-Jin Choi, et al. 2011. Unique DNA methylome profiles in CpG island methylator phenotype colon cancers. Genome Research. doi:10.1101/gr.122788.111



Click here www.stat.purdue.edu/~doerge/BIOINFORM.D/FALL11/sem.html for a full schedule of BIOINFORMATICS SEMINARS, past and present.