DNA Methylation Assays

High Throughput Data AnalysisBIOS 691-803, VCU

Winter 2010Mark Reimers, PhD

DNA Methylation

• Cytosine bases sometimes methylated

• Shuts down transposons

• In vertebrates:– Condenses chromatin– Renders genes inaccessible– Heritable in cell lineages– Developmental fate decisions

DNA Methylation

Adding a Methyl to Cytosine

Cytosine methylation is passed on to daughter cells

How Does Methylation Happen?

Distribution of Methylation

Distribution of CpG sites

DNA Methylation and Transcription

• Methyl groups block access to some transcription factors

• Me-C attracts MBD proteins that further suppress transcription

• Heavy methylation predisposes chromatin to condense

Methylation in Development

Methylation in Cancer

Assaying Methylation

• MeDIP (Methylated DNA immuno-precipitation)– Antibody to Me-C => ChIP – chip– Doesn’t distinguish among nearby sites

• Multiple restriction enzyme assays• Isoschizomer (HpaII/MspI) assays:

– MIAMI (Microarray-based Integrated Analysis of Methylation by Isoschizomers)

– HELP

• Bisulphite conversion of meC -> T, then hybridize to SNP style array

MeDIP

• Genomic DNA is randomly sheared by sonication

• Immunoprecipitate with an antibody that specifically recognizes 5-methylcytidine (5mC)

• Hybridize against control (no antibody) on array

MeDIP Data

EIF2C4

Copyright restrictions may apply.

Ordway, J.M. et al. Carcinogenesis 2006 27:2409-2423; doi:10.1093/carcin/bgl161

McrBCA schematic of three array probes (X, Y and Z) arranged along a chromosome is shown

Short fragments with methylated CpG’s have been removed

The HELP Assay

• MSPI cuts at 5’-CCGG-3’ – methylated or not

• HPAII cuts at 5’-CCGG-3’ only if unmethylated (useful restriction enzyme)

Sample

MSPI

HPAII

LabelPCR amplify

LabelPCR amplify

Co-hybridize

HELP Data

HELP log ratios

Methylation Data Analysis

• Regional QA

• Normalizing Bias in ratios– Probe sequence– CpG density– Intensity– Fragment length (for HELP & similar)

• Estimation– Are methylations similar at neighbors?

Normalizing MeDIP – CpG Bias

• Direct approach

• Compare to a standard: – fully methylated – Tanay: M.SssI treatment

• Indirect estimate– Regress M (ratio) on CpG density (assuming

all neighbors are equal)– Down et al: BATMAN

Normalizing Intensity Bias

• Strong intensity dependent bias in each chip

• Different intensity dependence in each chip

• Correlated with CpG density

• Ignored by BATMAN!

Normalizing Sequence Bias

• Significant dependence of intensity on CG

• Dependence differs among chips!

HELP ratios and fragment length

Normalizing Fragment Length - HELP

• Distribution of intensity varies by L

• Fit density curve and line up

More HELP technical biases

CHARM

Critique of CHARM• Improves reliability of mcrBC by assuming

smoothness

• Doesn’t incorporate probe effects

• No pre-processing of Illumina data used as reference

• Detailed data shows ‘block’ structure– With single CpG sites deviating from most– Difficult to detect using CHARM

Block Structure of Methylation

General Principles of Complex Assay Normalization

• Many reactions: each influenced by differences in processing conditions

• Differences in technique induce similar biases in probes with similar technical characteristics

• Aggregating probes by technical character (IF independent of biology) is an effective way to estimate bias on each chip individually