Skip to main content
Premium Trial:

Request an Annual Quote

2002 ABRF Microarray Research Group Study Explores Sources of Experiment Variability


  Researchers are still trying to come to grips with variations between microarray experiments that don’t reflect real biological differences. That’s why in its 2002 study, the Microarray Research Group (MARG) of the Association of Biomolecular Resource Facilities (ABRF) chose to get to the root of the problem, assessing possible factors contributing to variability in experiments that use Affymetrix arrays or cDNA spotted arrays.

At the ABRF meeting in Austin last week, Kevin Knudtson of the University of Iowa and Andrew Brooks of the University of Rochester Medical Center presented the group’s preliminary results.

“The most interesting thing that we found was that most of the variability was lab-to-lab, and not chip-to-chip” said Brooks, referring to the Affymetrix part of the study.

Researchers from ten universities participated in the retrospective study. Seven microarray core facilities provided data from almost 900 Affymetrix murine U74A and human U95A arrays, and Albert Einstein College of Medicine provided data from a smaller number of cDNA spotted arrays.

For the Affymetrix arrays, which were used to study 20 different types of tissues or cell lines, the researchers collected 3’/5’ ratio values for b-actin and GAPDH, as well as average difference values for three hybridization controls and RawQ, background, and scaling metrics.

To study the contribution of variability between array lots, at least five arrays with the same lot number were used for each sample type. To assess the influence of tissue type, array lot, sample processing, and scanner settings, the data were analyzed by non-parametric testing followed by a transformation analysis, “which allowed us to assess which variables are more important,” said Brooks.

Lab-to-lab variation, meaning differences in sample preparation and processing, accounted for the greatest source of error, followed by the photomultiplier tube (PMT) setting, using a “combined” tissue and cell line comparative analysis.

Originally the scientists had also planned to distinguish between individual tissue types and cell lines, but the number of samples was not sufficient to reach statistical significance.

The cDNA array study looked at an approach for “overall global normalization and correction” of the arrays, Brooks said, and found some sources of process errors, such as probe labeling, sequence-dependent dye biases, slide treatment, hybridization and scanning.

“The main solution to problems across both platforms is going to be a combination of an appropriate array and experimental design, and also in global normalization of the data regardless of the platform,” said Brooks.

In order to complete the study within the next few months, MARG is currently inviting researchers from academia as well as private companies to contribute their microarray data. Intellectual property is not an issue, Brooks reassured, since only the tissue type, general data about the array, and results for a few genes are required.

One “industry leader” has already committed to participate, he added. The group aims to publish the results of the study in a journal “that is going to have the largest reader base to help Affymetrix and cDNA [array] users,” Brooks said.

— JK

The Scan

Study Links Genetic Risk for ADHD With Alzheimer's Disease

A higher polygenic risk score for attention-deficit/hyperactivity disorder is also linked to cognitive decline and Alzheimer's disease, a new study in Molecular Psychiatry finds.

Study Offers Insights Into Role of Structural Variants in Cancer

A new study in Nature using cell lines shows that structural variants can enable oncogene activation.

Computer Model Uses Genetics, Health Data to Predict Mental Disorders

A new model in JAMA Psychiatry finds combining genetic and health record data can predict a mental disorder diagnosis before one is made clinically.

Study Tracks Off-Target Gene Edits Linked to Epigenetic Features

Using machine learning, researchers characterize in BMC Genomics the potential off-target effects of 19 computed or experimentally determined epigenetic features during CRISPR-Cas9 editing.