Package: crossarrayEWAS 0.99.1

crossarrayEWAS: Design-Aware Region and Block Detection for Cross-Array Longitudinal EWAS

Estimation and region-calling routines for epigenome-wide association studies that combine Illumina HumanMethylation450 and MethylationEPIC samples and follow subjects over more than one visit. Probe-level effects are fitted within subject, so subject identity, array type and chip are absorbed rather than modelled; adjacent probes are grouped by within-subject co-methylation; regions are obtained by total-variation denoising with a distance-decaying penalty, which places boundaries where the effect track jumps instead of thresholding a smooth curve; and long-range blocks are called from a three-state hidden semi-Markov model whose transition probabilities decay with genomic distance. Selection stability and permutation nulls resample subjects, not samples, so a subject's visit series is never split.

Authors:Katarzyna Kamieniecka [aut, cre], Samantha McLean [aut], Krzysztof Poterlowicz [aut]