De novo transcriptomes built from hundreds of human cornea, retina, and RPE RNA-seq samples identifies thousands of differentially expressed ocular specific gene transcripts and novel eye disease relevant exons

Abstract

Purpose : Better understanding of the role of genetic variation in eye disease requires a complete set of eye tissue specific gene transcripts and the ability for the scientific community to easily leverage this data. We use hundreds of cornea, retina, and retinal pigmented epithelium (RPE) tissue RNA-seq samples to construct de novo transcriptomes. Our results are made available via an interactive web app at eyeIntegration.nei.nih.gov (eyeIntegration).

Methods : All healthy publicly available human cornea, retina, and RPE RNA-seq samples were found by searching the Sequence Read Archive for ‘cornea|retina|RPE|ocular|eye’. The raw sequence for each sample was downloaded and processed in a reproducible Snakemake pipeline that identifies poor quality and mislabeled samples, builds de novo transcriptomes (StringTie), quantifies differential expression (limma) against 22 non-ocular Genotype-Tissue Expression (GTEx) tissues, and identifies eye specific exon usage (rMATs). To re-classify variants of unknown significance (VOUS), we extracted eye disease related variants from ClinVar.

Results : 65 cornea, 156 retina, and 156 RPE samples across 42 projects were found. Thousands of novel transcripts were found in each tissue and many are significantly differentially expressed relative to the GTEx dataset (Table 1). We built a plotting tool to efficiently visualize our novel transcript splicing patterns (Figure 1). Finally, we find seven non-coding VOUS in patients with ocular eye diseases within our novel ocular exons.

Conclusions : We have leveraged the wealth of publicly available human RNA-seq data to build the most comprehensive human ocular transcriptomes to date. We identify thousands of novel eye tissue specific transcripts and make our data available at eyeIntegration. Finally, we demonstrate how these ocular specific transcriptomes can be used to better classify genetic variants in eye disease.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

Counts refer to the number of RNA-seq samples pulled from the SRA that passed quality control. Differentially Expressed (DE) novel transcripts are q < 0.01 in differential expression (limma) between the ocular and GTEx tissues.

 

Gene structure and splice graphs given for transcript ENST00000621736 (A) and a novel transcript (B) of NR2E3. Red exons are not seen in the reference transcriptome.