International Mammalian Genome Society

The 16th International Mouse Genome Conference (2002)

Plenary Presentations * Oral Presentations *
Poster Presentations: Complex Genetics and Disease Modifiers * Developmental Genetics * Functional Genomics * Gene Discovery * Genetic Manipulations to Alter Gene Function * Mouse Models: Human Disease and Pharmacogenetics * Sequence Annotation and Comparative Analysis of Genomes *
Attendees * Sponsors * Table of Contents * Photographs * Awards

POSTER 83 - SOME CHARACTERISTICS OF AFFYMETRIX PROBESETS

L Schalkwyk
Institute of Psychiatry

The arrays of in-situ synthesized oligonucleotides produced by Affymetrix currently offer the best-standardised method of large-scale measurement of transcript levels.  Data can be compared relatively well between experiments, including those in different laboratories.  The oligonucleotides (probes) are 25 bases long and for each a single mismatch probe is used as a specificity control.  One recurring problem with processing data from these chips is that many mismatch probes give a higher signal intensity than their perfect match counterparts. Several methods have been proposed for background removal and normalisation, but none so far use of prediction of duplex affinity from probe sequences.A duplex melting temperature (Tm) criterion is used by Affymetrix in their (proprietary) probe selection method, but not in analysis of hybridisation data.  A reasonable prediction of binding affinities can be made using the nearest-neighbour model, although some of the probes may not show simple two-state melting. Applying this method to the Affymetrix U74Av2  probe list gives a Tm range of over 40º. Most hybridisations show a modest effect of predicted Tm on signal intensity.  The fact that the effect is modest indicates that the methods have been successfully optimised to not discriminate between a wide range of Tms. There is thus not likely to discrimination between perfect match and mismatch, especially for higher-melting probes.  The observed ratio of perfect  to mismatch does indeed fall strikinglywith increasing predicted Tm indicating that the predicted Tms do offer a way to improve background subtraction and adjust signal intensities.