Factor graph-aggregated heterogeneous network embedding for disease-gene association prediction

BMC Bioinformatics

Table 2 Case study results

CUI	Disease name	CUI	Disease
C0575081	Gait abnormality	CUI:C0678230	Congenital Epicanthus

Gene	Original DB	Gene	Original DB
FGFR3		FBN1
POU1F1		IKBKG
ROR2		LBR
GRM1		KCNJ2
OFD1	DisGeNet*	ROR2	DisGeNet
BMPR1B		SOX9
FGFR2		FLNA	DisGeNet*
RPS29		GDF5
IL6		LMNA
SLC9A6		TBX3
MYH6		FOXG1	DisGeNet
RAF1		OFD1	DisGeNet
COL6A2		HDAC6
MKKS		GRM1
MAP3K7		TGDS
PTCH2		WDR60	DisGeNet
KCNJ2		GJA1	DisGeNet
RAG1		OAT
LMNA	DisGeNet	ZMPSTE24
FAS	DisGeNet	TREX1

In the prediction results of the above table, the candidate genes with known association were labeled in the original DB, and candidate genes marked with "*" indicate newly discovered associated genes, that is, there are not exist in dataset but records in the latest online database. The results show that our model has the ability to mine new disease gene associations, such as OFD1-C057508 and FLNA-C0678230. Our model does not remember the existing associations in the original dataset, but predicts new candidate genes by mining the hidden patterns. This is very important, because it is difficult to mine new genes only by making a high score for the known associations. Therefore, our model can help to decipher the relationship between diseases and genes, which has certain biomedical significance

ISSN: 1471-2105