Our knowledge driven HDN gene and molecular database systems approach consists of the following steps: 1) Determination of sibling diseases for IBD based on drug information, 2) Collection of IBD and sibling disease genes from multiple databases, 3) Scoring the disease genes by evidence-based ranking weighted by "the frequency in databases" and "the frequency in PubMed", 4) Evaluation of the disease genes for Protein-Protein Interaction relations, and 5) Investigation of GO-based functional similarity of drug targets to the putative IBD genes. We summarized the criteria for our selection of genes specific to IBDmild and IBDsevere in Table . Our results may lead to an elucidation for IBD pathogenesis that remains largely unknown.
Criteria for selection of genes specific to IBDmild and IBDsevere
Sibling diseases, closely aligned to a complex disease such as IBD, provide a novel opportunity to use comprehensive omics data to identify a core biochemical or treatment pathway not previously identified in medical biology. If correctly defined, a collection of sibling diseases can 'cover' the entire pathophysiological process of a target disease more completely than any one disease. Wall et al.
determined autism sibling diseases based on commonly involved genes [18
]. We defined IBD sibling diseases based on drug indications. Drugs act on changing pathogenic states of a disease. Accordingly, it is highly likely that the sibling diseases share a common molecular mechanism. This approach may be generalizable if drugs acting on sibling disease states, are available across the developmental progression of a given central disease.
Figure indicates that the HDNs for IBD change according to the state of the disease progression. Our results show that JAK2
, which are known to be associated with IBD [41
], are specific to the HDN of IBDmild
(Figure ) [10
]. These genes form the central part of the HDN of IBDmild
and have links to pro-inflammatory genes (IL1A/1B
, and IL23A/R
), which are common in both HDNs. These suggest that JAK2
are key factors in the early stage of IBD pathology. On the other hand, the HDN of IBDsevere
(Figure ) indicates a relation of IBD to more malignant diseases like cancer. Patients with long-standing IBD have an increased risk of developing colorectal cancer [42
]. Such a transition in functional classes of genes was also observed with differential expressed genes between IBDmild
obtained by GSE6731 (nodes with arrows in Figure ).
The HDNs of IBDmild
include 13 genes obtained by GWAS reported in [8
]. The 13 genes are highlighted by colored borders in orange in Figure and listed in boldfaced in Additional file 3
: Table S2. Most of the genes are interconnected with each other except ICAM1
. All interconnected genes belong to the early response of inflammation, which includes cytokines, chemokines, receptors, and cellular signaling molecules. The other two genes, ICAM1
, belong to the late response of inflammation, i.e., enhancement of immune response. Our HDNs illustrate what molecules intervene between the two sibling disease pathways. Both in the HNDs of IBDmild
are common to both disease states, indicating that the transcriptional regulation intervenes between the early (IBDmild
) and late (IBDsevere
) responses. In this way, our HDN analysis and approach helps to clarify the molecular and therefore disease implications of GWAS candidate genes.
A recent comprehensive review of molecular pathways for IBD pathogenesis [10
] supports characteristic genes in our HDNs of IBDmild
. The genes characteristic of the HDN of IBDmild
, and IL23
) belong to Th17-cell differentiation in [10
], and the genes characteristic of the HDN of IBDsevere
) belong to Th1-cell driven responses in [10
]. The physiological balance between Th1 and Th17 may be deteriorated by environmental factors such as intestinal bacteria stress, which eventually leads to autoimmune responses composing IBD ("hygiene hypothesis" introduced by Strachan [43
]). Our HDNs suggest a transition from Th17 to Th1 dominancy along with progression of malignancy [44
]. Our approach stratifying the disease-related genes into IBDmild
enables us to infer a clinically significant transition of a state of a disease such as the Th1/Th17 transition.