Hotel Estoril Eden, Monte Estoril,
Portugal
5-8 October 2005

 

 
 

Previous
Main
Abstracts
NextText Box: Participants
Text Box: Programme


A Statistical Package for Quality Control and Low-Level Analysis of Illumina BeadArrays

Mark Dunning1, Isabelle Camilier1, Natalie Thorne1 and Simon Tavare1,2
1Department of Oncology, University of Cambridge, UK
2
University of Southern California, USA

Gene expression levels are often monitored using microarrays, typically either two-colour spotted arrays or oligo arrays such as those from Affymetrix. Illumina have created an alternative microarray technology (BeadArray) based on randomly arranged beads, each of which carries copies of a gene specific probe. The method of attaching probes to beads results in a more favourable hybridisation surface. A random sampling from an initial pool of beads produces an array with, on average, 30 randomly positioned replicates of each probe type on an array. This degree of replication and the high density of BeadArrays make them suitable for high-throughput experiments, such as experiments associated with the HapMap project.

With such high-throughput experiments being performed on BeadArrays, there is clearly the need for a statistical tool to be widely available to confirm the quality of such arrays and investigate any problems that may occur. We therefore developed an R library implementing quality control checks and low-level analysis for BeadArrays. 

We used TIF image files obtained by scanning BeadArray experiments to assess the procedure used by Illumina to calculate foreground and background intensities for each bead and hence the amount of hybridisation to the attached probes. We first used the methods described by Illumina to re-create the values exactly and then considered if all the stages are necessary and if some stages could be modified.  

We will present the results of our investigation into the positioning and numbers of each bead type on a particular array and whether this agrees with the expected random distribution. We will also present our findings on the process for determining which beads of a particular type are classed as outliers on an array and what impact this has on the eventual bead summaries produced. Methods will then be presented which use these bead summaries to compare the variability between arrays within in the same experiment.  

We will then demonstrate functions enabling the user to view the original TIF image scanned from an array and be able to see information about intensities of beads in a particular region and also highlight beads determined to be outliers. It will be shown that these functions provide a useful tool for understanding any systematic problems that can occur on BeadArrays.