Identification of a 5-Gene Signature Predicting Progression and Prognosis of Clear Cell Renal Cell Carcinoma.

BACKGROUND Although the mortality rates of clear cell renal cell carcinoma (ccRCC) have decreased in recent years, the clinical outcome remains highly dependent on the individual patient. Therefore, identifying novel biomarkers for ccRCC patients is crucial. MATERIAL AND METHODS In this study, we obtained RNA sequencing data and clinical information from the TCGA database. Subsequently, we performed integrated bioinformatic analysis that includes differently expressed genes analysis, gene ontology and KEGG pathway analysis, protein-protein interaction analysis, and survival analysis. Moreover, univariate and multivariate Cox proportional hazards regression models were constructed. RESULTS As a result, we identified a total of 263 dysregulated genes that may participate in the metastasis of ccRCC, and established a predictive signature relying on the expression of OTX1, MATN4, PI3, ERVV-2, and NFE4, which could serve as significant progressive and prognostic biomarkers for ccRCC. CONCLUSIONS We identified differentially expressed genes that may be involved in the metastasis of ccRCC. Moreover, a predictive signature based on the expression of OTX1, MATN4, PI3, ERVV-2, and NFE4 could be an independent prognostic factor for ccRCC.

Background

Clear cell renal cell carcinoma (ccRCC) is the most common malignancy in the kidneys, which has increasing incidence and mortality rates worldwide [1]. Treatments for localized ccRCC can vary from radio-frequency ablation to partial or radical nephrectomy; however, once RCC progresses to distant metastasis, the curative effect of current targeted drug therapies is limited [2]. Additionally, when first diagnosed, approximately 30% patients already have metastasis [1]. Therefore, it is urgent to understand the underlying mechanism of metastasis and to identify novel biomarkers with greater prognostic values.

The TNM staging system has been used for over 80 years and is important for estimating the outcome of various cancers; however, it provides an incomplete prognostic value [3-5]. Clinical outcomes can differ significantly among patients with the same tumor stage [6]. Despite surgical removal of the tumor, a subgroup of patients experience recurrence, indicating that at the time of curative surgery, the metastasis was already present [7]. However, no consensus was reached regarding the surveillance protocols of RCC, and no available tumor-associated biomarkers can predict recurrence in patients who may have benefited from earlier therapy [8]. Previous studies in colorectal cancer proposed several gene signatures and proved to be useful in predicting prognosis [9-11]. In this study, we divided patients from the Cancer Genome Atlas database into a non-metastasis group and a metastasis group in order to screen the differently expressed genes. Furthermore, we constructed a risk scoring system based on upregulated genes involved in metastasis to identify a multi-gene signature for use as an independent predictor for ccRCC.

Material and Methods

Data collection

The TCGA database contains large cohorts of genomic abnormalities and clinical information across the world, and is publicly available. RNA sequencing counts data from the ccRCC cohort, which consists of 539 tumor samples and 72 normal tissues, were obtained from the TCGA data portal (). Clinical data pertaining to patients’ age, gender, grade, stage, survival and recurred/progressed outcome were also acquired from the TCGA data portal. We divided patients based on N stage and M stage into 2 groups. Patients with both M0 and N0 stage were assigned to the non-metastasis group, whereas M1 and/or N1 patients were assigned to the metastasis group.

Identification of differentially expressed genes (DEGs)

We identified the DEGs using the edgeR package, with a cutoff of adj.p-value <0.05 and a |logFC| >2 [12]. DEGs were visualized with volcano plot through the gplots package in R (version 3.5.2).

Enrichment analysis of DEGs

We performed a functional enrichment analysis of the DEGs using DAVID (Database for Annotation, Visualization, and Integrated Discovery) to determine the gene ontology (GO) categories by using cellular component (CC), molecular function (MF), or biological processes (BP), as well as KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway [13]. P<0.05 was defined as significant enrichment. An online web tool was used to visualized these processes ().

Construction of PPI network

We used the STRING database to retrieve the protein-protein interaction (PPI) network of DEGs, and we used Cytoscape software to reconstruct and visualize the network [14,15]. Individual network modules with 10 or more nodes were shown.

Univariate and multivariate Cox analysis to screen the candidate genes

CcRCC samples were separated into 2 groups according to the median gene expression. Then, age (≤60/>60), sex (male/female), grade (G1–G2/G3–G4), stage (I–II/III–IV), T stage (T1–T2/T3–T4), N stage (N0/N1), M stage (M0/M1), specific gene, and survival data (time and state) were all included into the Cox regression model to preform univariate and multivariate Cox analysis using SPSS 22.0 (IBM Corporation, Armonk, NY, USA).

Establishment of a prognostic signature based on candidate genes

The stepwise multivariate Cox regression analysis model was constructed based on the candidate genes to extract the mRNA-based model with the best predictive ability. The criteria for inclusion and exclusion was set as P<0.05. Subsequently, the risk score for each patient was computed using the mRNA-based prognostic model as follows: Risk score=expRNA1*βRNA1+ expRNA2*βRNA2+expRNA3*βRNA3+…expRNAn*βRNAn, where expRNA was the mRNA expression level and βRNA referred to the regression coefficient derived from the multivariate Cox hazards regression analysis. Based on the risk score for each patient, patients from the TCGA database were separated into 2 groups: a low-risk group and a high-risk group. Kaplan-Meier survival analysis was performed to assess differences in overall survival and disease-free time of patients using a log-rank test in GraphPad Prism 7.0. In addition, the receiver operating characteristic (ROC) curve was utilized to evaluate the specificity and sensitivity of the survival and disease-free prediction by the area under the curve using the R package “survivalROC” [16]. Heatmaps and clustering were generated based on the ClustVis open web tool [17].

Predictive value assessment

To evaluate the clinical value of our risk scoring system, we analyzed the clinical characteristics and risk scores in univariate Cox regression. We included factors with P<0.05 into the multivariate Cox regression analysis model. Then, a P<0.05 was treated as an independent prognostic factor. Moreover, to assess the relationship between risk level and clinical characteristics, we regrouped the patients based on age, sex, grade, stage, T stage, M stage, N stage, vital status, and risk level. A P<0.05 was considered as statistically significant using the chi-square test.

Results

Differentially expressed genes related to the metastasis of ccRCC

In this study, we defined M0 and N0 patients as the non-metastasis group (198 cases), while M1 and/or N1 patients (89 cases) were defined as the metastasis group. Altogether, 263 genes were found to be dysregulated according to the cutoff criteria, among which, 101 genes were upregulated and 162 gene were downregulated (Figure 1, Supplementary Table 1). Functional enrichment analysis of gene ontology revealed that dysregulated genes were mainly enriched in sequence-specific DNA binding, receptor binding, the extracellular region, the integral component of the plasma membrane, ion transmembrane transport, and insulin receptor signaling pathway. KEGG pathway analysis indicated that genes were primarily enriched in neuroactive ligand-receptor interaction and synaptic vesicle cycle (Figure 2). Moreover, the PPI network consisted of 10 modules, which included 255 nodes and 316 edges. The most significant module is shown in Figure 3.

Figure 1

The volcano plot for DEGs related to metastasis. The x-axis is -log10(FDR) and the y-axis is logFC. The red dots represent upregulated genes and green dots represent downregulated genes.

Figure 2

Go term and KEGG pathway analysis for DEGs. (A) Top 10 molecular function (MF) processes. (B) Cellular component (CC). (C) Top 10 biological processes (BP). (D) KEGG pathway analysis.

Figure 3

The most significant module. The color and the size of a node indicates the number of proteins interacting with the designated protein.

Survival-related genes by Cox regression analysis

To identify key genes that may affect overall survival of patients, we performed Cox proportional hazard regressions analysis on upregulated genes. Twenty key genes were demonstrated to influence overall survival: OTX1, FOXE1, FAM83A, HMGA2, KRT6A, DPYSL5, ANXA8, MATN4, ROS1, CSMD3, MAGEC3, AMER2, CPLX2, PI3, KRT13, ERVV-2, ANKFN1, VTN, NFE4, and ZNF114 (Figure 4).

Figure 4

(A–T) Survival-related upregulated genes. Kaplan-Meier survival curves were generated for genes with P<0.05 in multivariate Cox regression analysis.

Construction of a risk scoring system based on candidate genes

For the purpose of extracting a signature that possesses the best predictive efficacy, 20 key genes were subjected to the stepwise multivariate Cox regression model. Results from the model revealed a total of 5 genes that proved to be significant survival predictors. The related information of these 5 genes is shown in Table 1. Subsequently, the risk score for each patient was computed as follows: expOTX1*0.725+expMATN4*0.473+expPI3*0.548+expERVV-2*0.458+expNFE4*0.410. According to the median risk score, we assigned these scores to the low- or high-risk group. Overall survival analysis showed that the low-risk group had better prognoses compared with the high-risk group (Figure 5A). The prognostic ability of the 5-gene signature was assessed by the AUC value of the ROC curve. The AUC was 0.687 for 3-year and 0.695 for 5-year overall survival, indicating a good performance of the 5-gene signature (Figure 5C, 5E). Risk scores in the low-risk group ranged from 0 to 0.215232506445 and ranged from 0.215490360701 to 360.615372760823 in the high-risk group (Figure 5G). Disease-free survival analysis revealed a significant difference between the 2 groups, with the low-risk group having a longer disease-free time (Figure 5B). In the ROC curve, the AUC for 3-year disease-free survival was 0.674 and 0.681 for 5-year disease-free survival (Figure 5D, 5F). Risk scores in the low-risk group ranged from 0 to 0.187870897657 and from 0.194574172497 to 360.615372760823 in the high-risk group (Figure 5H). Figure 6 shows the expression patterns of all the 5 genes in the 2 groups. The expression of OTX1, MATN4, and PI3 were significantly higher in the high-risk group in the 2 cohorts (Figure 7).

Table 1

Overall information of 5 genes constructing the prognostic signature.

Gene	Gene name	Gene type	Hazard ratio	Coefficient	P value
OTX1	Orthodenticle Homeobox 1	Protein-coding	2.064	0.725	<0.0001
MATN4	Matrilin 4	Protein-coding	1.605	0.473	0.007
PI3	Peptidase Inhibitor 3	Protein-coding	1.73	0.548	0.002
ERVV-2	Endogenous Retrovirus Group V Member 2	Protein-coding	1.581	0.458	0.009
NFE4	Nuclear Factor, Erythroid 4	Protein-coding	1.506	0.41	0.015

Figure 5

The 5-gene predictive signature in ccRCC. (A) Kaplan-Meier curve of OS in the low- and high-risk groups. (B) Kaplan-Meier curve of DFS in the low- and high-risk groups. (C) ROC curve for the 3-year survival prediction by the 5-gene signature. (D) ROC curve for the 3-year disease-free survival prediction by the 5-gene signature. (E) ROC curve for the 5-year survival prediction. (F) ROC curve for the 5-year disease-free survival prediction. (G) Risk scores distribution among OS cohort. (H) Risk scores distribution among DFS cohort.

Figure 6

Expression pattern of the 5-gene signature in OS and DFS cohort. (A–C). In the OS cohort, the expression levels of OTX1, MATN4, and PI3 were significantly higher in the high-risk group. (D–F). In the DFS cohort, the expression levels of OTX1, MATN4, and PI3 were significantly higher in the high-risk group.

Figure 7

Heatmap of the 5 genes. (A) Heatmap of the OS cohort. (B) Heatmap of the DFS cohort. Red indicates the high-risk group, while blue indicates the low-risk group.

Assessment of gene signature prognostic value

Univariate Cox regression analysis of the prognostic power of our risk scoring system showed that age, grade, stage, T stage, N stage, M stage, and risk level were all indicators of poor outcome. Then, these 7 indexes were entered into the multivariate Cox regression model, showing that risk level could be treated as an independent prognostic factor (Table 2). Furthermore, as is shown in Table 3, based on the chi-square test, risk level was significantly correlated with sex, grade, tumor stage, T stage, N stage, M stage, and vital status. Collectively, our results demonstrate that our 5-gene signature is a robust tool for use in predicting prognosis and recurrence.

Table 2

Univariate and multivariate analysis of risk level and patient survival.

Variables	Univariate analysis			Multivariate analysis
	HR*	95% CI	P value	HR	95% CI	P value
Overall survival
Age (years)
≤60 (257)	1.683	1.228–2.306	0.001	1.540	1.122–2.114	0.007
>60 (246)
Sex
Male (325)			0.791
Female (178)
Stage
I+II (303)	4.313	3.092–6.015	<0.0001	2.511	1.256–5.019	0.009
III+IV (200)
T stage
T1–T2 (321)	3.482	2.534–4.785	<0.0001			0.681
T3–T4 (182)
N stage
N0 (487)	3.925	2.124–7.255	<0.0001			0.093
N1 (16)
M stage
M0 (425)	4.572	3.21–6.294	<0.0001	2.202	1.500–3.232	0.0001
M1 (78)
Grade
G1–G2 (232)	2.644	1.860–3.759	<0.0001			0.051
G3–G4 (271)
Risk level
Low risk (251)	2.592	1.859–3.612	<0.0001	1.779	1.251–2.530	0.001
High risk (252)

HR estimated from Cox proportional hazard regression model; multivariate models were adjusted for age, grade, T, N, M, and stage.

HR – hazard ratio; CI – confidence interval.

Table 3

Relationship between clinical parameters and risk level.

Subgroup	High risk	Low risk	Total	P value*
Age				0.503
≤60	125 (24.85%)	132 (26.24%)	257
>60	127 (25.25%)	119 (23.66%)	246
Sex				0.008
Male	180 (35.79%)	151 (30.01%)	331
Female	72 (14.31%)	100 (19.88%)	172
Grade				<0.0001
G1–G2	86 (17.10%)	146 (29.03%)	232
G3–G4	166 (33.00%)	105 (20.87%)	271
Stage				<0.0001
I+II	119 (23.66%)	184 (36.58%)	303
III+IV	133 (26.44%)	67 (13.32%)	200
T stage				<0.0001
T1–T2	132 (26.24%)	189 (37.57%)	321
T3–T4	120 (23.86%)	62 (12.33%)	182
N stage				<0.0001
N0	111 (45.87%)	114 (47.11%)	225
N1	16 (6.61%)	1 (0.41%)	17
M stage				<0.0001
M0	178 (37.47%)	219 (46.11%)	397
M1	57 (12.00%)	21 (4.42%)	78
Vital status				<0.0001
Alive	141 (28.03%)	200 (39.76%)	341
Dead	111 (22.07%)	51 (10.14%)	162

Chi-square test was used.

Discussion

CcRCC has been shown to display distinct variability in clinical outcome, possibly due to the intrinsic molecular heterogeneity, which remains unclear, especially with regard to the mechanism of distant metastasis [18]. Moreover, the clinically available parameters, such as TNM stage and Fuhrman grade, are indispensable for prognostic prediction [19]. Nevertheless, there remains an urgent need to detect prognostic biomarkers due to the high heterogeneity in ccRCC.

In the current study, we performed bioinformatic analysis between the non-metastasis and metastasis ccRCC group to identify genes involved in metastasis. As a result, we found that 263 genes were dysregulated; functional enrichment analysis of these genes revealed that dysregulated genes were primarily enriched in sequence-specific DNA binding, receptor binding, extracellular region, integral component of plasma membrane, ion transmembrane transport, insulin receptor signaling pathway, neuroactive ligand-receptor interaction, and synaptic vesicle cycle. Most importantly, we identified a 5-gene panel signature (OTX1, MATN4, PI3, ERVV-2, and NFE4) after the Cox proportional hazards regression analysis. Then, a risk score was acquired by combing the 5 genes. Recently, Wei et al. also identified key genes involved in the metastasis of ccRCC using similar bioinformatics methods [20]. However, in our study, we make our inclusion criteria clear with regard to the metastasis and non-metastasis groups. Moreover, we calculated each patient’s risk score based on the 5-gene signature. The 5-gene signature could independently predict overall survival for ccRCC patients, demonstrating that this signature might be useful in clinical practice.

OTX1 encodes a member of the Bicoid sub-family of homeodomain-containing transcription factor, which may play a role in sensory and brain organ development. It has been described as a vital molecule for axon refinement [21]. Terrinoni et al. demonstrated that the p53 protein can directly induce OTX1 expression by acting on its promoter in breast cancer, and Figueira-Muoio et al. revealed that the OTX pathway is important in medulloblastomas development [22,23]. OTX1 was also found to promote colorectal cancer progression in vitro through epithelial-mesenchymal transition and hepatocellular carcinoma progression by regulation of the ERK/MAPK pathway [24,25]. In bladder cancer, OTX1 combined with FGFR3 and TERT can function as a surveillance biomarker [26]. However, the role of OTX1 in ccRCC is still unknown. PI3, also called elafin, encodes an elastase-specific inhibitor that functions as an antimicrobial peptide [27,28]. Caruso et al. demonstrated that elafin predicts poor outcome in ovarian and breast cancer patients, and it may play a role in tumor dormancy; moreover, it has been shown that elafin is an important therapeutic target for breast and ovarian carcinoma [29-31]. MATN4, a member of the von Willebrand factor A domain-containing protein family, has not been widely studied in cancer to date [32]. A study showed that under acute stress, CXCR4 and MATN4 are involved in the regulation of hematopoietic stem cells proliferation and expansion [33]. ERVV-2 is functionally important in reproduction, and NFE4 is involved in preferential expression of the gamma-globin genes in fetal erythroid cells [34,35]. These 2 genes have not been well defined in cancer biology, particularly in ccRCC.

In summary, our study used an integrated analysis to identify differentially expressed genes that participate in metastasis of ccRCC. Furthermore, we constructed a 5-gene signature with a quantitative index that exhibited an independent prognostic value. In the future, this 5-gene signature may be used to identify patients who need regional lymph node dissection during radical nephrectomy [36]. Since these 5 genes are correlated with poor outcome, they might be therapeutic targets for ccRCC. However, in vivo and in vitro studies are still needed to reveal the biological functions of these predictive mRNAs in ccRCC.

Conclusions

We identified differentially expressed genes that may participate in the metastasis of ccRCC. More importantly, we established a predictive signature based on the expression of OTX1, MATN4, PI3, ERVV-2, and NFE4, which could serve as significant progressive and prognostic biomarkers for ccRCC.

Supplementary Table 1.

Differentially expressed genes involved in metastasis in ccRCC.

Genes	Log FC	Genes	Log FC	Genes	Log FC
PRSS38	7.293012973	PASD1	5.849055467	BAAT	4.972016934
KCNE5	4.734843904	NFE4	4.607079054	ALPG	4.569998517
FDCSP	4.526032273	CABP2	4.453865694	OLFM4	4.268762733
GAGE1	4.26812475	LHX3	4.065474738	KRT13	4.019547465
CRABP1	3.807319872	SOHLH1	3.801178392	CACNG6	3.763439672
VSTM2B	3.632937049	ANXA8	3.591407826	H2BFM	3.555425223
AMER3	3.524447368	MAGEC2	3.503241713	ERVV–2	3.464200342
CPLX2	3.401097143	GABRA3	3.388167297	RORB	3.361591792
MUC16	3.298663286	MARCOL	3.250302515	ZDHHC22	3.239809076
IGFL3	3.196619228	MTRNR2L6	3.179688152	C1orf94	3.13666645
PI3	3.126993299	CSMD3	3.047313412	ISL1	2.981373363
SP8	2.966745906	PNLIP	2.924034656	AMER2	2.904669856
TLX3	2.903886912	PDX1	2.882186281	DPYSL5	2.869458768
LCN15	2.843884331	VTN	2.819241353	ZPLD1	2.795929776
ISX	2.795438433	EPPIN	2.734479911	ALPP	2.699771711
PTPRZ1	2.695461275	INSL4	2.691308392	CHAT	2.659157612
MAGEC3	2.626652586	DAB1	2.581804555	RDH8	2.559245587
XKR7	2.556307418	CIDEC	2.535297601	ROS1	2.520534946
CSN3	2.519649538	VSTM2L	2.490355446	HTR1D	2.489417462
FAM83A	2.455106896	S100A7	2.43745305	HMGA2	2.423695315
ANKFN1	2.408489181	UBE2U	2.401187787	TRPV5	2.378341308
LCE1C	2.377491995	DRGX	2.375422577	SLC18A3	2.366620248
KLF17	2.362440353	ZIC2	2.35428125	SPACA3	2.348805744
FCRL4	2.346660183	CRP	2.332869284	SPANXB1	2.326683931
UTS2R	2.314650465	MATN4	2.311903817	ZNF114	2.30971043
ADIPOQ	2.296860368	KISS1	2.295428739	LIN28B	2.291059085
ANXA8L1	2.248521884	MAGEB1	2.242953797	SPANXN3	2.242130571
IL22RA2	2.240150546	C1QL2	2.209979502	AGBL1	2.206686442
TLX2	2.202836841	RLBP1	2.159036842	NPPB	2.154907807
HTR5A	2.149124359	SERPINB3	2.14782693	SBSN	2.1417701
SPINK6	2.114686901	FOXE1	2.096651213	GNG13	2.082021332
ALOXE3	2.054881574	RTP3	2.051444937	OTX1	2.040341385
HMX2	2.030173909	KIRREL3	2.025763852	DMRTA2	2.018437908
KRT6A	2.006147507	IRS4	−7.069279584	AQP6	−6.952633679
LY6L	−6.466292518	HHATL	−6.178662879	CRISP3	−5.942489086
PAGE5	−5.566923649	HBG1	−5.565412617	SFTPB	−5.46165805
MDFIC2	−4.7346839	MAGEA11	−4.702227866	CCKAR	−4.620218512
NTSR2	−4.412067953	LRRTM1	−4.295989741	CLDN8	−4.291159779
PAGE2B	−4.290100156	DCAF4L2	−4.285297367	CHRM1	−4.203135741
FEZF2	−4.181641013	SERTM2	−4.084855062	PSG4	−4.069117346
DEFB125	−4.034642804	ATP6V0A4	−4.03380667	ATP6V1G3	−3.918376267
FXYD4	−3.882031698	C10orf71	−3.845620551	ST8SIA3	−3.817050292
TTR	−3.8141048	PAGE4	−3.813169574	FGF9	−3.781764959
POU3F4	−3.771004791	ATP6V0D2	−3.753224136	PSG9	−3.751868431
SPOCK3	−3.749385525	TMEM213	−3.705206888	KBTBD12	−3.684155012
KRTAP5–8	−3.632121999	PIP	−3.541015006	TMEM215	−3.537175656
RHBG	−3.513276723	CTNNA2	−3.497574449	GJD2	−3.465274322
GLB1L3	−3.462356811	SLC4A1	−3.459997603	NUPR2	−3.451627461
HBG2	−3.360260797	NR5A1	−3.354792948	VWA5B1	−3.340662569
MLANA	−3.311141752	OMG	−3.302149224	BSND	−3.275017729
AQP10	−3.234439151	FER1L6	−3.223091448	SLC26A7	−3.196657291
KLK1	−3.168181356	ATP6V1B1	−3.166112958	RHCG	−3.157008772
FGL1	−3.146889407	TNNT3	−3.130099704	SLC24A2	−3.090435759
PLK5	−3.073715835	PSG5	−3.063389834	TYR	−3.036736515
CD177	−2.967875945	CDH7	−2.947214145	XAGE5	−2.941242246
AQP5	−2.928574991	LGI1	−2.920563422	SCRT1	−2.915273241
LCN1	−2.897125323	CRISP2	−2.891236689	CGA	−2.880719932
FOXI1	−2.856870004	SLC4A9	−2.85058536	GREM2	−2.846325204
ADAM7	−2.823853478	MYMX	−2.780243665	FOXI2	−2.747040565
BPIFA2	−2.744920257	NXPH2	−2.73264296	FAM24B	−2.005641145
CLCNKB	−2.711841094	DNTT	−2.703518233	FRG2C	−2.696015544
TMEM61	−2.688842068	CASP14	−2.687885646	GIMD1	−2.686569536
LHFPL4	−2.682599598	ADCYAP1	−2.68255206	TBATA	−2.65671051
DMRT2	−2.645831657	MCCD1	−2.625093054	PAGE2	−2.615268476
GPRC6A	−2.613101443	WFIKKN2	−2.598374715	UGT2B4	−2.586510771
IGF2	−2.56153826	KERA	−2.560942199	FRG2B	−2.549870167
SLC7A13	−2.544471449	MOG	−2.537312543	ASCL4	−2.534282307
C11orf53	−2.519948822	PSCA	−2.507368106	GCGR	−2.506059534
PLA2G4F	−2.494234559	DAZ1	−2.461947613	NKX6–1	−2.457759032
RHAG	−2.444447278	LUZP2	−2.426420149	HBM	−2.424034763
NMRK2	−2.412559163	TRIM50	−2.4050669	LRRC52	−2.396507205
GRIK1	−2.380726671	CRYAA	−2.361368316	ADRB1	−2.352091261
AHSP	−2.350914787	ASB5	−2.345814708	CNMD	−2.339953179
GGTLC3	−2.332560999	GCG	−2.325940672	PSG8	−2.303814006
STAP1	−2.295027287	RGS8	−2.290434876	STAC2	−2.269340054
CYP1A1	−2.246907308	KRTAP5–3	−2.240169508	HBD	−2.234219697
RBBP8NL	−2.232288152	UGT2B28	−2.229968426	ATP13A5	−2.22816884
SMOC1	−2.226575753	DEFA4	−2.194637278	FRMD7	−2.190289838
CA1	−2.182904697	CLNK	−2.179307919	SRARP	−2.162262658
ERP27	−2.157025947	KLK4	−2.152704502	FAM133A	−2.145658322
PNMT	−2.136928193	CEACAM7	−2.131707182	NRK	−2.11265576
SMIM5	−2.105569769	DEFA3	−2.104237638	TDGF1	−2.101766107
ADGRF1	−2.098885814	GRM1	−2.096205239	HEMGN	−2.091490619
UGT1A4	−2.087390147	AL445989.1	−2.918112259	PRG4	−2.083544157
ABCB5	−2.082109144	PGPEP1L	−2.077264255	PCP4	−2.063618468
HAO1	−2.062354203	HSPB3	−2.051568162	MYH8	−2.04723169
THBS4	−2.085595685	AL035425.2	−4.867341747	C20orf141	2.010402307
TMPRSS11E	−4.867341747	HEPACAM2	−2.731391743