Reproducibility of oligonucleotide arrays using small samples

BMC Genomics. 2003 Jan 30;4(1):4. doi: 10.1186/1471-2164-4-4.


Background: Low RNA yields from small tissue samples can limit the use of oligonucleotide microarrays (Affymetrix GeneChips). Methods using less cRNA for hybridization or amplifying the cRNA have been reported to reduce the number of transcripts detected, but the effect on realistic experiments designed to detect biological differences has not been analyzed. We systematically explore the effects of using different starting amounts of RNA on the ability to detect differential gene expression.

Results: The standard Affymetrix protocol can be used starting with only 2 micrograms of total RNA, with results equivalent to the recommended 10 micrograms. Biological variability is much greater than the technical variability introduced by this change. A simple amplification protocol described here can be used for samples as small as 0.1 micrograms of total RNA. This amplification protocol allows detection of a substantial fraction of the significant differences found using the standard protocol, despite an increase in variability and the 5' truncation of the transcripts, which prevents detection of a subset of genes.

Conclusions: Biological differences in a typical experiment are much greater than differences resulting from technical manipulations in labeling and hybridization. The standard protocol works well with 2 micrograms of RNA, and with minor modifications could allow the use of samples as small as 1 micrograms. For smaller amounts of starting material, down to 0.1 micrograms RNA, differential gene expression can still be detected using the single cycle amplification protocol. Comparisons of groups of four arrays detect many more significant differences than comparisons of three arrays.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Nucleic Acid Hybridization
  • Oligonucleotide Array Sequence Analysis / methods*
  • RNA / genetics
  • RNA / isolation & purification
  • Reproducibility of Results
  • Sensitivity and Specificity


  • RNA