Statistics In Pairwise Methylation Chip Data Analysis

Question

[Vova Naumov]

Hi! We have data fom Illumina HumanMethylation450 chip. Samples are 22 pairs of heath and cancer tissue of colon. Data look like:

        NAME NormX69p1  PatX69p2 NormX71p1  PatX71p2
1 cg00000029 0.3132485 0.2388572 0.3576819 0.3425549
2 cg00000108 0.8783011 0.8674102 0.9334134 0.9128920
3 cg00000109 0.6492147 0.7035905 0.8155046 0.8260411
4 cg00000165 0.2349398 0.4401709 0.3531564 0.3631415
5 cg00000236 0.6231979 0.7310550 0.8357330 0.7973528
6 cg00000289 0.4467668 0.3847913 0.6090226 0.6029724

Question is what statistical methods I need to use to estimate, if difference between individuals is more than difference between different tissue types. Now we use R for statistics. First what comes intro my mind is ANOVA, but I can't melt my data well because it's too large. We have done hierarchical clustering and it works like this:

And tis is pairwise correlation of beta-values between samples:

• So I'd like to get some answers like:
  • what statistical appoach is it beter to use to find intra- and interindividual differences
  • maybe you have seen some papers that can help ma with statstcs
  • what software do you rcommend to use

Thank you!

Answer 1

[balaji]

have you done pca or heatmap on this data ?

[Sukhi Singh]

use comments for this query, like I am using as its not an answer :)