Using genetic markers to orient the edges in quantitative trait networks: The NEO software

BMC Systems Biology

Table 2 Using NEO to identify genes that are reactive to Insig1.

Edge orientation Insig1↓	LEO. NB. OCA female	Model prob	Path coef	Wald test pval	df	χ ²	Known literature/novel (a.k.a.)	LEO. NB. OCA male	Male mice val†
Fdft1	1.4	0.75	0.64	<e-20	5	2.7	+	0.2
Dhcr7	1.2	0.24	0.79	<e-20	5	6.8	+	1.9	*
Scd1	1.2	0.58	0.63	<e-20	5	3.8	+	0.4	*
Sc4mol	1.1	0.35	0.68	<e-20	5	5.6	+	0.5	*
0610030G03Rik	1.1	0.82	0.67	<e-20	5	2.2	novel (Tlcd1)	2.4	*
Fads2	1.0	0.64	0.61	<e-20	5	3.4	+	0.02
Adipor2	0.97	0.98	0.61	<e-20	5	0.73	+	-2.0
Fasn	0.96	0.72	0.77	<e-20	5	2.9	+	1.0	*
Eaf2	0.89	0.16	0.54	6e-16	5	8	novel, Eaf2	-0.5
Stard4	0.87	0.46	0.59	<e-20	5	4.6	+	0.7	*
Fads1	0.86	0.82	0.73	<e-20	5	2.2	+	-0.5
Dlat	0.84	0.81	0.58	<e-20	5	2.3	+	0.7	*
Rdh11	0.82	0.80	0.73	<e-20	5	2.3	novel, Rdh11	-0.7
B430110G05Rik	0.81	0.87	0.52	2e-13	5	1.8	novel (Slc25a44)	0.7	*
Aqp8	0.72	0.61	0.59	<e-20	5	3.6	+	-1.6
Slc23a1	0.63	0.58	0.49	1e-11	5	3.8	novel, Slc23a1	0.1
Slc25a1	0.63	0.37	0.65	<e-20	5	5.4	novel, Slc25a1	-0.3
Acac	0.59	0.64	0.73	<e-20	5	3.4	+	-0.4
Acas2	0.58	0.19	0.64	<e-20	5	7.4	+	-2.5
Gale	0.53	0.65	0.58	<e-20	5	3.3	novel, Gale	-0.3
Mod1	0.38	0.60	0.59	<e-20	5	3.7	+	0.6	*
Qdpr	0.37	0.74	0.59	<e-20	5	2.7	novel, Qdpr	0.4	*
6030440G05Rik	0.35	0.70	0.54	8e-15	5	3	novel (Frmd4b)	-0.7

Here we used the female mouse liver BxH data to illustrate that NEO can be used to identify genes that are reactive to a given trait (here Insig1). The table reports the 23 genes with highest LEO.NB.OCA(Insig1 → B) scores. Since 14 of the 23 genes are already known to be reactive to Insig1, these results represent a highly significant validation success of NEO; using the 23388 genes on the array and assuming that there are 200 known genes downstream of Insig1 (a conservative estimate), the Fisher exact p-value of validation success is p = 1.0 × 10^-13. The NEO analysis in female mice also implicates 9 novel genes. PubMed searches on these genes did not turn up any information about a role of these genes in liver or sterol homeostasis. The table also reports the analysis results using the male mice of the BxH cross. The validation (val†) column shows a star (*) if the initial finding in female liver was replicated in the independent test set of 129 F2 male mice; we defined validation success as LEO.NB.OCA score above 0.3 using the default settings of automatic SNP marker selection. The male liver analysis confirms three of the nine novel genes suggested from the female analysis: Tlcd1, Slc25a44, and Qdpr. The fact that not all genes can be replicated in the male data may reflect known differences between female and male mouse liver tissue expression profiles [48].

ISSN: 1752-0509