Differential expression of transcript isoforms in schizophrenia
Transcriptome sequencing (RNA-seq) enables investigation of differential expression of transcript
isoforms and genes, with expression level of the former likely more meaningful than of the latter since
transcript isoforms often encode different proteins with different biological function.
We profiled the transcriptome (>9M single-end RNA-seq reads per subject) in lymphoblastoid cell lines
of 1,278 European ancestry subjects (550 cases, 728 controls) from the MGS collection. We used programs
Tophat and Cufflinks2 for read alignment, transcript assembly, and quantification (FPKM) of each transcript
isoform and gene. We restricted our analyses to the 108,346 transcripts and 21,215 genes from Gencode v.20
annotation that showed detectable expression (FPKM>0) in at least 80% (1,022) of subjects.
We used square root as the variance stabilizing transformation for the FPKM, and then adjusted for 12
covariates. After inverse normalization of residuals, we evaluated the association of expression levels with
schizophrenia status by linear regression.
We found 1,265 transcripts and 1,252 genes to be differentially expressed by affection status
(Bonferroni corrected p<0.05). Of these transcripts, 1,015 were in genes also showing significant differential
expression in the gene-level analysis, while 250 were in genes not significantly differentially expressed. 447
differentially expressed genes did not show differential expression at the transcript isoform level. We found a
number of significant transcripts and genes in HUGO Gene Nomenclature Committee gene families, such as
histocompatibility complex, histones, interferons, and interleukins.