Quality control of Affymetrix arrays

1
Quality control of Affymetrix arrays
2
What can go wrong?

• RNA degradation (before hyb)
• 3/5
• Dirty samples
• background, present calls
• Uneven hybridizations (during hyb)
• RLE, NUSE
• Machine problems

3
If gt10, Affymetrix may provide with new array
4
MAS5 report

• Report Type Expression Report
• Date 1100AM 07/01/2003
• __________________________________________________
  ____________________
• Filename 030701RE_U133A_PS1.CHP
• Probe Array Type HG-U133A
• Algorithm Statistical
• Probe Pair Thr 8
• Controls Antisense
• __________________________________________________
  ____________________
• Alpha1 0.05
• Alpha2 0.065
• Tau 0.015
• Noise (RawQ) 1.970
• Scale Factor (SF) 0.943
• TGT Value 200
• Norm Factor (NF) 1.000
• __________________________________________________
  ____________________
• Background
• Avg 44.89 Std 0.81 Min 43.10 Max 47.30

5
RNA degradation

• 3/5 ratios of GAPDH
• AffyRNAdeg in package affy
• 3/5 ratios for all genes on the array

6
RNA degradation
7
RNA degradation

• Results show that RNA degradation is reproducible
  when making technical duplicates

8
affyPLM

• Library to work with probes
• fitPLM()
• fits a linear model with probe effect

9
affyPLM
10
affyPLM

• image(cel.f)
• plots weights from fit

11
Relative Log Expression

• RLE(affybatch)
• the expression value of one array compared to
  median of all

12
Normalized Unscaled Standard Errors

• NUSE(affybatch)

13
NUSE
RLE
14
(No Transcript)
15
Solution

• Uneven hybridizations normalization takes care
  of
• RNA degradation outliers problematic
• High background the results will not be as
  clear, but trends can still be seen
• Bioinformatics and Computational Biology
  Soultions Using R and Bioconductor. Gentleman et
  al. Chapter 3 Quality assessment of Affymetrix
  GeneChip Data, Bolstad et al.

16
Machine problems

• With time, settings change which reflects on
  the data
• Roos Jahangir Tafrechi
• Arnolda de Nooij-van Dalen

old vs old
new vs old