2017). Nevertheless, the large data available from GWAS is indeed valuable and

contains great potential for future therapy. The gap between GWAS and DR may be

bridged closer using additional methods to identify the most relevant genes and

deduce drug candidates using drug-drug, drug-disease, and biological pathway

mapping and analysis (Lau and So 2020).

Most current studies use GWAS-based prioritization methods to create lists of top

potential causal genes for a disease of interest. A recent study (Zheng et al. 2020)

used a series of computational methods to identify plausible candidate genes for

coronary artery disease (CAD). These methods included computational programs,

such as Sherlock, NetWAS, SMR, GWAB, TWAS, Prix fixe, DAPPLE, and

DEPICT. GWAS summary data that has been narrowed down using variables,

such as co-functionality, gene-functionality, and tissue-specificity factors, is subse-

quently fed into all or some of the eight computational programs. The resulting hits

are then further analysed for biological functions using GO/KEGG enrichment

analysis, tissue-specific gene expression, and interactions with other cellular

pathways and crosstalk analysis (Zheng et al. 2020). The significance of such an

analysis ultimately provides mechanistic insights into the disease pathogenesis and

narrows down the pool of drug targets that can be repositioned for the studied

disease.

The general flow of GWAS begins with identifying the basic unit of genetic

variation, SNPs, as markers of a particular genomic region of interest. Disease

indication SNPs are distinguishable from common SNPs in the human genome,

which makes a strong basis for study. Linkage analysis is a technique which aids in

the identification of gene mutations such as the cystic fibrosis transmembrane

conductance regulator (CFTR) gene known to cause CF. This technique, however,

does not work well for diseases where the genetic mechanisms are influenced by

external factors such as heart disease (Bush and Moore 2012). GWAS obtains

genomic data from diverse diseases according to the type of disease through

customizing certain parameters, tools, and analysis criteria (Bush and Moore

2012). The array of GWAS data contains a large potential pool of novel targets

that can be used for DR (Pritchard et al. 2017), albeit GWAS alone is not the final

answer. Deeper analysis of GWAS is required to identify key target genes for DR,

including functional genomic techniques, identification of existing drugs, and pre-

clinical validation of drug targets.

5.3.2.2 Network-Based Approaches and the Support of Artificial

Intelligence (AI)

Network-based models of DR extract information from diverse databases and

provide results of key connections between the information. In the form of coded

nodes and edges, the data output is commonly seen in the form of a network of

connections. The nodes commonly represent either a drug, disease, or gene, while

the edges represent the interactions between them (Yella et al. 2018). These

approaches use GWAS data, cellular pathway mapping, and drug database informa-

tion to output networks by either entirely knowledge-based inputs (information from

databases) or computationally inferred information from existing inputs (aided by

5

Genomic Approaches for Drug Repositioning

59