Table 1.
Number of genes with redundant probe sets before and after filtering out the absent probe sets.
Figure 1.
Scheme for genome-based redundant probe sets grouping.
Genomic alignments for probe set consensus sequences are tested for instances of probe sequences within the genomic portion of the alignment. For all probe sets whose consensus sequence alignments contain all eleven probes, a new trimmed target region genomic alignment is generated bounded by the five and three-prime-most probe sequences. In parallel, genomic alignments for mRNA are combined into groups based on genomic overlap patterns. The trimmed target region alignments are then compared with the exon overlap groups from the mRNA branch of the pipeline. All probe set target sequence alignments that overlap with at least one member of the mRNA exon overlap groups are then annotated as targeting the given group. Probe sets interrogating members of the same exon overlap group are designated redundant probe sets.
Figure 2.
The distribution of genes with various significance consistency index values.
A consistency index was calculated for each gene with multiple probe sets and at least one probe set significantly different across compared groups. The y-axis indicates the proportion of genes with the consistency index indicated on the x-axis. AG and GG refer to Affymetrix groupings and genome-based groupings, respectively. AGP and GGP refer to AG and GG groupings in which only probe sets called as “present” were included in the calculation. Significance level is FDR 0.005.
Figure 3.
Significance consistency among the redundant probe sets changes depends on FDR threshold.
A consistency index was computed for each gene with redundant probe sets at different FDR level. AG and GG indicate the original Affymetrix grouping and proposed genome-based groupings, respectively. AGP and GGP represent the AG and GG groupings. AGP and GGP refer to AG and GG groupings in which only probe sets called as “present” were included in the calculation.
Table 2.
Relationship between P/A calls and significance for differential expression for genes with two probe sets.
Table 3.
Consistency of fold changes between redundant probe set pairs.
Figure 4.
Fold change comparison across redundant probe sets.
We first tested each probe set on the array for the brain region effect and identified a list of significant probe sets using FDR 0.005 significance level. We then compared this significant probe set list with the list of genes with two present redundant probe sets according to genome based grouping. For genes with at least one significant probe set, we plotted the two fold changes (on log scale) against each other with random naming of probe set1 (prs1) and probe set2 (prs2) in a gene. Genes with one significant probe set are presented by points. Genes with two significant probe sets are represented by circles. We also tested the two fold changes obtained from each probe set for genes with two redundant probe sets using FDR 0.05 significance level as described in Methods. Genes that show significant difference in their fold changes are represented by red V.
Table 4.
Comparison of consistency across sub data sets and consistency across redundant probe sets based on re-sampling.
Table 5.
Summary of redundant probe sets consistency evaluation.