Page 114 - Read Online

Ratnapriya. J Transl Genet Genom 2022;6:240-256 https://dx.doi.org/10.20517/jtgg.2021.54 Page 248

[88]
based on the disease association of markers in or near genes . A recent analysis of AMD-GWAS data using
this approach reported that eight genes (C2, C3, LIPC, MICA, NOTCH4, PLCG2, PPARA, and RAD51B)
[89]
strongly contributed to significant pathways driving AMD association .

Differential expression analyses
A second approach to gain insights into the disease pathways is to compare the gene expression between
normal and disease tissues. These differential expression (DE) analyses facilitate the identification of novel
biological processes and genes involved in disease [90,91] . This approach overcomes the limitation of analyzing
genes driven by genetic association and could reveal candidates that can offer an explanation for “missing
heritability”. In AMD, RPE/choroid and the photoreceptors of the neural retina have been the primary
[92]
focus of gene expression studies . Analysis of these tissues from both macular and extramacular regions
can offer key insights. However, the limited availability of these tissues, along with the technical and logistic
difficulty of obtaining high-quality RNA from the postmortem tissues, have been rate-limiting steps in
large-scale comparisons between normal and AMD donors. Multiple studies have looked into the gene
expression changes during AMD and have reported several differentially expressed genes [80,92-94] . However,
these findings have been difficult to replicate across studies, which could be attributed to several factors.
First, variations in gene expression exist widely. Thus, using only small samples can lead to gene discovery
that is reflective of interindividual as well as disease-associated changes. DE analyses also fail to capture
small changes in gene expression because of other confounding factors such as age, gender and allele
frequencies of GWAS risk variants between cases and controls. Cellular heterogeneity in the bulk RNA-seq
adds another layer of complication in the interpretation of DE genes in the context of the disease.
Incorporation of deconvolution methods and scRNA-seq in future comparative studies of normal and
disease conditions can help in characterizing complex cellular changes in response to pathology [72,73] .

Co-expression network analysis
Cellular processes are driven by multiple interacting genes and one might expect that genes whose
expression is highly correlated will also have a functional correlation. Generation of such co-expression
[95]
networks from the transcriptome data have been used to associate genes of unknown function with
biological processes, to prioritize candidate disease genes at the GWAS loci or to pinpoint transcriptional
[96]
regulatory modules in disease [97,98] . Weighted gene co-expression network analysis (WGCNA) is a widely
used method to identify co-expressed gene modules . Analysis of co-expression WGCNA modules built
[99]
using human retina transcriptome data was shown to be enriched for genes within known AMD loci, and
pathways involved in AMD (complement, extracellular matrix, and angiogenesis pathways) were closely
connected . These analyses can be further refined to dissect regulatory networks that are altered in disease
[80]
and reveal genes that are likely to be regulators of disease processes. To further refine the modules and
assign causality, genes within the module can be subjected to regulatory network construction. While these
approaches may still leave a considerable number of candidates that may not be feasible for follow-up
studies, integration of this approach with additional GWAS, eQTL and epigenomics datasets will aid in
identifying disease AMD genes that remain unidentified through traditional genetic approaches.

Transcriptome-wide association studies
Gene discovery in GWAS is reaching saturation as the bulk of the remaining heritability is likely to be
attributed to large numbers of common variants of small effects, rare variants, epigenetic changes, or
environmental cues. Identification of novel associations with small effects requires large GWAS cohort sizes
and lowering the genome-wide significance can lead to a higher false-positive rate. There has been a lot of
progress on the methodologic approaches that integrate multiple data types, as they can offer unique
insights and advantages in the interpretability of GWAS findings. Integrating gene expression with the
GWAS can help in increasing the power of association as well as provide functional interpretation of GWAS