Figure 1.
The main resource for identifying alternatively spliced exons was compiled using two data sources produced by high-throughput genome-wide techniques: sequence-derived computed data and expression-derived experimental data. (A) Sequence-derived computed data: putative alternatively spliced exons were predicted using a sequence data-driven alternative splicing model and the genomic locations of ISEs which were identified by scanning the intronic regions near intronic SNPs for putative ISE motifs (6-mer in length, see Materials and Methods). (B) Expression-derived experimental data: exon-level microarrays and cis-expression quantitative trait loci data (cis-eQTL, p<0.01, Materials and Methods) were obtained from our previous work with HapMap LCLs (87 Utah residents with Northern and Western European ancestry (CEU) and 89 Yoruba from Ibadan, Nigeria (YRI)). (C) Experimental data (panel B) was integrated in order to not only verify predicted exon skipping events from AS modeling data (panel A) but also to identify regulatory relationships between ISE SNPs and putative exon skipping events. (D) We investigated SNP associations with SI using an additive model and the allele associated with an increase in exon skipping. (E) Potential functionality of genetic variations was computationally evaluated by investigating the protein domains encoded by skipped exons and thus the impact on the protein function.
Figure 2.
Enrichment of Exon Skipping–Associated ISE SNPs among Complex Human Trait-Associated SNPs.
(A) The distribution of the number of predicted exon-skipping ISE SNPs observed for each of the 1,000 random draws of the 3,353 SNPs from bins matched for minor allele frequency (MAF, CEU) and matched on the distance to the nearest exon (of the 3,353 human-trait associated SNPs downloaded from the NHGRI catalog) is shown in the bar graphs. The actual number of 111 predicted exon-skipping ISE SNPs observed in the 3,353 SNPs from the NHGRI catalog is shown as a solid asterisk. The distance to the nearest exon of intronic SNPs was calculated using Ensembl Gene predictions (ensGene.txt.gz) and the SNP annotation file (snp129.txt.gz) downloaded from the UCSC genome browser, http://hgdownload.cse.ucsc.edu/goldenPath/hg18/database/. For the distance to the skipped exon of ISE SNPs, we used exon skipping events from Ensembl Gene predictions. (B) The distribution of the number of predicted exon-skipping ISE SNPs observed for each of 1,000 draws of 49 SNPs from bins matched for MAF to the 49 SNPs associated with human tissue specific exons identified by Heinzen et al. [15] (bins include all SNPs in HapMap, CEU) is shown in the bar graphs, with the actual number of 6 predicted exon-skipping ISE SNPs observed in the 49 from Heinzen et al. shown as a asterisk.
Figure 3.
Enrichment of Exon Skipping–Associated ISE SNPs among cis-eQTL SNPs.
Predicted exon-skipping ISE SNPs are enriched among cis-eQTL SNPs (p<0.001). The distribution of the number of predicted exon-skipping ISE SNPs observed for each of 1,000 draws of 68 SNPs from bins matched for minor allele frequency (MAF, CEU) to the cis-eQTL SNPs is shown in the bar graphs, with the actual number of 68 predicted exon-skipping ISE SNPs observed in the cis-eQTL SNPs shown as a solid asterisk.
Figure 4.
Candidate ISE SNPs Associated with an Increase in Exon Skipping.
The y-axis is the absolute value of 1 minus SI (|1-SI|); here, a higher level of |1-SI| corresponds to a higher level of exon skipping. The allele with an asterisk showed an association with increased exon skipping. In the case of the SNP rs339408 (with alleles A/G), the G allele is located within three ISE motifs, namely at the second, third, and fourth site of AGGGAT, CAGGGA, and TCAGGG, respectively, but the A allele is the exon skipping associated ISE allele (A). In the case of the SNP rs6436071 (A/T), the T allele participates in two ISE motifs, namely at the second and third site of CTTGGC and GCTTGG, respectively (B). In the case of rs1265112 (A/G), the A allele is located at the fourth site in the CACACT ISE motif but the G allele is the exon skipping associated ISE allele (C). Finally, for rs12599391 (C/T), the T allele is located at the 3rd site of the AATTGT ISE motif but the C allele is the exon skipping associated ISE allele (D).
Table 1.
Protein Domains for Skipped Exons.
Figure 5.
Structural Alignment of Pairs of Transcript Isoforms.
(A) SLC25A15 (B) RNF8. SLC25A15 and RNF8 show differences in the 3D structures for the protein with and without the skipped exon. The structural difference is quantified by RMSD and TM-score.
Figure 6.
No Difference in Enrichment of Human Complex Trait-Associated SNPs among ISE SNPs Classified According to Physical Distance from the Associated Alternatively Skipped Exon's Splice Site Junction.
The distribution of the number of predicted exon-skipping ISE SNPs observed for each of the 1,000 random draws of the 3,358 SNPs from bins matched for minor allele frequency (MAF, CEU) to the 3,358 SNPs downloaded from the NHGRI catalog (bins include all SNPs in HapMap, CEU) is shown in the bar graphs. A solid asterisk represent the actual number of 4, 11, 21, and 49 predicted exon-skipping ISE SNPs away from the associated skipped exon within the tested distances as shown in panel A, B, C, and D, respectively.
Figure 7.
ISE SNPs in NHGRI Catalog Are Likely to Be Farther from Skipped Exon than Intronic SNPs in NHGRI Catalog Are from Nearest Exon.
P-value was calculated by using a two-sample Kolmogorov-Smirnov Test.
Table 2.
Distance to the Nearest Exon of Intronic SNPs Tends to Be Smaller Than the Distance to the Skipped Exon of ISE SNPs.