Figure 2

EST cluster member number correlation with hybridization intensity (Avg Diff) on the MuscleChip. Panel A. Shown is a log-linear scatter plot of Avg Diff (expression level of each RNA given by normalized hybridization intensity on the MuscleChip) versus the relative percentage of EST cluster members in a non-normalized normal muscle EST sequencing project (n-EST/cluster in 28,074 sequenced clones). The number of unique EST clusters shown is 2,052. The correlation coefficient of the log-linear regression is 0.67 indicating that, on average, there is a relationship between EST cluster member number and hybridization intensity. A small number of "outliers" from this analysis are indicated and numbered as described in the Results. The large majority of these outliers correspond to non-(_at) probe sets, where probe set design rules were altered or dropped. Thus, most outliers are likely due to cross-hybridization to other mRNA species. Panel B. Shown is a log-linear scatter plot of the average absolute intensity of six individual profiles (y-axis: average Avg Diff) versus the relative percentage of EST cluster members (x-axis: relative percentage of 28,074 on log₁₀ scale) in a non-normalized normal muscle EST sequencing project. The number of unique EST clusters shown is 2,052. The correlation coefficient of the log-linear regression is 0.6 showing a correlation between number of sequences in EST clone and the averaged absolute intensity of individual samples hybridized to the MuscleChip. High outliers are non _at only probe sets. Panel C. Shown is a log-linear scatter plot of the average absolute intensity of six individual profiles (y-axis: average Avg Diff) versus the relative percentage of EST cluster members (x-axis: relative percentage of 28,074 on log₁₀ scale) in a non-normalized normal muscle EST sequencing project. The EST clusters shown are reduced to ideal (_at only) probe sets. The correlation coefficient of the log-linear regression is 0.6. The difference in sensitivity between the two techniques may contribute to lower linear correlation in the intermediate and high abundant transcripts giving a correlation coefficient of 0.6 identical to the correlation when the non-ideal probe sets are included.