Fig. 2.

Filtering and statistical test algorithm for identifying aging- and cognition-related genes (ACRGs). The initial set of 8799 gene probe sets contained on the HG-U34A gene chip was reduced according to a priori filters before statistical testing to decrease multiple comparisons and expected false positives. Gene probe sets were removed if they were called absent (1a), if they were ESTs (1b), or if the difference between the young and aged groups did not comprise at least 75% of the maximal normalized age differences (1c). Each of the remaining 1985 (gene) probe sets was then tested by ANOVA across the three age groups (n = 9–10 per group) to determine whether it changed significantly with aging (2). Each of the 233 genes that changed significantly with age (p ≤ 0.025) was then tested across all animals (n = 29) for significant behavioral correlation with the OMT and SWM 24 hr retention values (Pearson's;p ≤ 0.025). Age-dependent genes that correlated with either or both tasks were identified as primary ACRGs. Additionally, 11 genes that were not correlated behaviorally were included as ACRGs (3b) because their age-dependent alterations were significant at a much higher confidence level (ANOVA;p ≤ 0.001).