Table 3.
Colorectal cancer risk associations identified by a colorectal mucosa-specific transcript isoform-wide association study (TIsWAS).
| # | ENSEMBL identifier | Gene | Chr | Start (bp, GRCh37) | End (bp, GRCh37) | P S-MultiXcan | Mean z score | Effect size | n models | n indep | Top GWAS SNP at <1Mb | SNP location | P GWAS |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | ENST00000609196 | ACP6 | 1 | 147,101,453 | 147,131,116 | 6.43E-11 | −1.264 | −0.048 | 4 | 3 | rs1541187 | 147,051,493 | 1.44E-04 |
| ENST00000493129 | ACP6 | 1 | 147,127,341 | 147,142,574 | 1.65E-23 | −5.781 | −0.482 | 2 | 2 | rs1541187 | 147,051,493 | 1.44E-04 | |
| 2 | ENST00000273153 | CSRNP1 | 3 | 39,183,346 | 39,195,066 | 9.99E-07 | 4.891 | 0.099 | 1 | 1 | rs4676609 | 39,214,256 | 4.63E-06 |
| 3 | ENST00000274695 | CDKAL1 | 6 | 20,534,688 | 21,232,635 | 1.29E-06 | −4.841 | −0.046 | 1 | 1 | rs9295474 | 20,652,717 | 7.61E-08 |
| 4 | ENST00000481601 | CCDC183 | 9 | 139,694,767 | 139,702,192 | 9.60E-07 | −4.490 | −0.048 | 2 | 2 | rs2811736 | 139,651,954 | 3.12E-05 |
| ENST00000464157 | ABCA2 | 9 | 139,902,688 | 139,903,240 | 7.39E-07 | −4.951 | −0.235 | 1 | 1 | rs2811736 | 139,651,954 | 3.12E-05 | |
| 5 * | ENST00000543000 | PLEKHG6 | 12 | 6,426,733 | 6,427,529 | 3.30E-09 | 6.003 | 0.076 | 3 | 2 | rs10849433 | 6,406,904 | 6.73E-17 |
| 6 | ENST00000448790 | TOX4 | 14 | 21,945,335 | 21,967,315 | 1.22E-07 | 5.290 | 0.498 | 1 | 1 | rs3811252 | 22,855,779 | 2.11E-05 |
| 7 | ENST00000478981 | BNIP2 | 15 | 59,955,092 | 59,961,148 | 9.91E-07 | −4.893 | −0.326 | 1 | 1 | rs7182962 | 59,945,783 | 6.04E-08 |
| 8 | ENST00000310144 | PSMC5 | 17 | 61,904,543 | 61,909,379 | 4.18E-10 | 6.247 | 0.553 | 1 | 1 | rs12449782 | 61,576,249 | 2.18E-05 |
As per Table 2, SMultiXcan uses a two-sided F-test to quantify the significance of the joint fit of the linear regression of the phenotype on predicted expression from multiple tissue models jointly. All associations shown were transcriptome-wide significant after Bonferroni correction for 27,941 transcripts with an S-MultiXcan model (i.e. P = 0.05/27,941 = 1.79 × 10−6 for the PS-MultiXcan). Novel associations were called when >1Mb from both a GWAS-significant SNP and a TWAS locus. As expected, all these loci showed evidence of a risk association in the full TWAS (FDR < 0.05, P < 2.86 × 10−3). Transcripts with boundaries < 1 Mb apart were considered to be in the same cluster. This resulted in seven CRC associations. One further association (*) was identified based on conditional TIsWAS analysis (Supplementary Table 8). Other annotations are as per Table 2.