Gene chip analysis methods

Microarray Suite5.0 (MAS)

The Affymetrix Microarray Suite assesses gene expression in two ways: absolute and comparative. For absolute analysis, it calculates signal intensities and makes detection calls then computes a p-values for each call. Signal intensity denotes the difference in brightness between the Perfect Match (PM) and mismatch (MM) probes in a probe set. A probe set is comprised of a series of probe pairs each containing one PM probe and one MM probe. The PM probe perfectly matches the target cRNA sequence. The sequence for PM and MM are the same, except for a change to the Watson-Crick complement in the middle of the MM probe sequence. The reason for including a MM probe is to provide a reference intensity that is caused by background cross-hybridization, other non-specific binding, and stray signals. A failure to discount those signals may impair the accurate assessment of the gene expression.

Presence calls are based on the p-values computed by a non-parametric Wilcoxon’s rank test comparing the population of PM probes and MM probes for each gene on the chip. Using the default settings, a probe set with a detection p-value less than 0.4 is considered present (represented by a letter ‘P’ in the detection column of any given sample), a p-value between 0.6 and 0.4 is considered marginal (represented by a letter ‘M’), and a p-value greater than 0.6 is considered absent (represented by a letter ‘A’). A weight of reliability can be based on the specific p-value, i.e., a p-value closer to 0 indicates a gene more likely to be present. On the other hand, a p-value closer to 1 indicates a gene more likely to be absent.

For comparative analysis between two chips, the signal log ratio, change and change p-value are calculated. The signal log ratio is related to the fold change by the following formula:

Fold change=2signal Log ratio for signal log ratio≥0

Fold change=(-1) x 2- signal log ratio for signal log ratio<0

The change of a gene expression between the experimental and baseline conditions will be one of the following: Increase, Marginal Increase, No Change, Marginal Decrease or Decrease depending on the p-value (calculated using the Wilcoxon signed rank test and Tukey Byweight ). The p-value ranges for each of these call are 0 – 0.0025, 0.0025 – 0.003, 0.003 – 0.997, 0.997 – 0.9975, and 0.9975 – 1, respectively, for those calls. A document created by Affymetrix describing the algorithms used in all statistical calculations can be provided upon request. All analysis results can be exported as Excel files, and all files generated during the experiment are available for download from the GeneX website (http://genes.med.virginia.edu)   

 

Affymetrix Data Mining Tool 3.0 (DMT)

The DMT is used for conducting more high level analyses. As opposed to the MAS which can only perform comparisons between two chips, the DMT can perform statistical comparisons between two groups of chips. It can also calculate averages, and fold changes associated with two group comparisons. We also use the DMT to perform cluster analyses and queries. For more information on any Affymetrix product, see their website (www.affymetrix.com

 dChip

dChip is an open source software developed to provide an alternative algorithm for analyzing GeneChip data. Measurement of gene expression level  in dchip is by way of a model base approach. For high end analysis, it provides two main functions: hierarchical clustering and sample comparisons. The cluster analysis algorithm produces a cluster tree that visually represents the behavior of selected genes by grouping together genes that behave similarly to each other across all the samples in the experiment. Samples are similarly clustered based on their similarity to each other. The “compare samples” function compares individual samples or groups of samples based on a number of different criteria that can be modified by the user. The results are stored in Excel files and include fold changes, confidence intervals, and p-values. More information concerning the dChip software and the actual application can be found on their website (http://biosun1.harvard.edu/~cli/dchip_request.htm)