Comparison Analysis (Experimental vs Baseline arrays)

# Compare the difference values (PM-MM) of each probe pair in the baseline array to its matching probe pair on the experimental array.

# Before comparing two arrays, variations between the two experiments caused by technical and biological factors must be corrected by scaling, normalization or a Robust normalization.

# Change p-value
Using the difference between PM and MM as well as PM and background intensities, the Change p-value is calculated by the Wilcoxon's signed rank test.

# Change Call
Increase (I): p-value < gamma1
Marginal Increase (MI): gamma1 < p-value < gamma2
No Change (NC): gamm2 < p-value < 1-gamma2
Marginal Decrease (MD): 1-gamma2 < p-value < 1-gamma1
Decrease (D): p-value > 1-gamma1

# Signal Log Ratio Algorithm
One-step Tukey's Biweight method

Single Array Analysis (Oilgonucleotide expression arrays)

# Single stranded DNA, 25 bp

# 14~20 probe pairs for each gene

# Each probe pair has a Perfect Match (PM) and a Miss Match (MM)

# MAS 4.0 (Average difference) = average of PM-MM difference

# Low-level anaysis: feature extraction, normalization, computation of expression indexes

# High-level analysis: t-test, ANOVA

# Discrimination Score
R = (PM - MM) / (PM + MM)

# Detection p-value by One-sided Wilcoxon's Signed Rank test
H0: E(R) = tau (default = 0.015)
Ha: E(R) > tau

# Detection Call
Present: p-value <= alpha1
Marginal: alpha1 < p-value <= alpha2
Absent: p-value > alpha2

defaults: alpha1=0.04, alpha2 = 0.06

# Signal Algorithm
One-Step Turkey's Biweight Estiimate

If PM > MM, informative
if PM < MM, uninformative and use an imputed value called Idealized Mismatch (IM)