Screening of primary open-angle glaucoma diagnostic markers based on immune-related genes and immune infiltration

(2022) 23:67
Suo et al. BMC Genomic Data
/>
BMC Genomic Data

Open Access

RESEARCH

Screening of primary open‑angle glaucoma
diagnostic markers based on immune‑related
genes and immune infiltration
Lingge Suo1,2†, Wanwei Dai1,2†, Xuejiao Qin3, Guanlin Li4, Di Zhang1,2, Tian Cheng5, Taikang Yao5 and
Chun Zhang1,2* 

Abstract 
Purpose:  Primary open-angle glaucoma (POAG) continues to be a poorly understood disease. Although there were
multiple researches on the identification of POAG biomarkers, few studies systematically revealed the immune-related
cells and immune infiltration of POAG. Bioinformatics analyses of optic nerve (ON) and trabecular meshwork (TM)
gene expression data were performed to further elucidate the immune-related genes of POAG and identify candidate
target genes for treatment.
Methods:  We performed a gene analysis of publicly available microarray data, namely, the GSE27276-GPL2507,
GSE2378-GPL8300, GSE9944-GPL8300, and GSE9944-GPL571 datasets from the Gene Expression Omnibus database.
The obtained datasets were used as input for parallel pathway analyses. Based on random forest and support vector machine (SVM) analysis to screen the key genes, significantly changed pathways were clustered into functional
categories, and the results were further investigated. CIBERSORT was used to evaluate the infiltration of immune cells
in POAG tissues. A network visualizing the differences between the data in the POAG and normal groups was created. GO and KEGG enrichment analyses were performed using the Metascape database. We divided the differentially
expressed mRNAs into upregulated and downregulated groups and predicted the drug targets of the differentially
expressed genes through the Connectivity Map (CMap) database.
Results:  A total of 49 differentially expressed genes, including 19 downregulated genes and 30 upregulated
genes, were detected. Five genes ((Keratin 14) KRT14, (Hemoglobin subunit beta) HBB, (Acyl-CoA Oxidase 2) ACOX2,
(Hephaestin) HEPH and Keratin 13 (KRT13)) were significantly changed. The results showed that the expression profiles

of drug disturbances, including those for avrainvillamide-analysis-3, cytochalasin-D, NPI-2358, oxymethylone and
vinorelbine, were negatively correlated with the expression profiles of disease disturbances. This finding indicated that
these drugs may reduce or even reverse the POAG disease state.
Conclusion:  This study provides an overview of the processes involved in the molecular pathogenesis of POAG in
the ON and TM. The findings provide a new understanding of the molecular mechanism of POAG from the perspective of immunology.

†

Lingge Suo and Wanwei Dai contributed equally as first authors.

*Correspondence:
1
Department of Ophthalmology, Peking University Third Hospital, 49 North
Garden Road, Haidian District, Beijing 100191, PR China
Full list of author information is available at the end of the article

© The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​
mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.


Suo et al. BMC Genomic Data

(2022) 23:67

Page 2 of 16

Keywords:  Bioinformatics analysis, Primary open-angle glaucoma, Optic nerve, Trabecular meshwork, Immune
infiltration

Background
Glaucoma is the leading cause of irreversible blindness
worldwide. With the growing number and proportion
of older persons in the population, it is projected that
111.8 million people will have glaucoma in 2040 [1]. Primary open-angle glaucoma (POAG) is the most common
type of glaucoma, accounting for 60–70% of all glaucoma
patients [2]. In POAG, the anterior and posterior segments of the eye are affected, and serious damage may
be inflicted upon the trabecular meshwork (TM) and
optic nerve (ON) [2–4]. The TM is a specialized eye tissue essential for the regulation of aqueous humor outflow
and control of intraocular pressure (IOP), disturbances
of which may lead to elevated IOP and glaucoma [5]. In
general, POAG has an insidious onset and develops painlessly and quietly, visual problems often late in the course
of the disease, when significant and irreversible ON damage occurs [1]. Neuroprotective therapies are not available, and current treatments are limited to lowering
IOP, which can slow disease progression at early disease
stages. However, over 50% of glaucoma cases are not
diagnosed until irreversible ON damage has occurred [6].
Numerous POAG patient data have been collected
in research [5, 7, 8], but the molecular pathogenesis of
POAG remains largely obscure. Therefore, an effective
treatment option that addresses these molecular changes
is still missing. In recent years, accumulating evidence has
shown that immune cell infiltration plays an important
role in POAG development [9]. Zhang et  al. generalized
that POAG may be associated with systemic disorders,
mainly those related to the nervous system, endocrine system and immune systems. It has been firmly established

that the neuroendocrine system and immune system
closely interact through mediators, such as hormones,
neuropeptides, neurotransmitters and cytokines [10].
Cytokines mediate the biological effects of the immune
system, and our previous study revealed an imbalance of
T-helper (Th) 1-derived and Th2-derived cytokines in the
serum of patients with glaucoma [11]. We also collected
data from irises of normal individuals and those with
POAG or chronic angle-closure glaucoma (CACG) [12].
Bioinformatics is an interdisciplinary subject that
combines a broad spectrum of domains, including the
fields of molecular biology, information science, statistics and computer science [13]. Machine learning,
a trendy subfield of artificial intelligence (AI), focuses
on extracting and identifying insightful and actionable
information from big and complex data using different

types of neural networks [14]. It is of great significance
to reveal the molecular mechanism of disease by using
these emerging technologies. Using omics technologies, we are able to measure the expression of several
thousand molecules from one sample of affected tissue,
leading to an exponential increase in data [15]. The data
were used in bioinformatics analyses to identify key transcription factors (TFs) associated with POAG to examine
the pathogenesis of glaucoma and may provide a basis
for the diagnosis of glaucoma and drug development.
CIBERSORT is a method to describe the composition
of immune cells in complex tissues based on their gene
expression profiles [16]. Few studies have used CIBERSPORT to analyze immune cell infiltration in POAG.
In this study, we identified the key genes from TM tissue and ON tissue in patients with POAG compared
with normal controls. The aim of this study was to gain
a deeper understanding of the molecular pathogenesis

of POAG by applying integrative bioinformatics analysis
to the available human gene expression data of the TM
and ON tissues in patients with POAG and controls. The
obtained results enable us to identify possible drug targets to modulate the disease outcome.

Results
Systematic search

After the systematic search, the datasets of the four different human microarray studies were selected for further
analyses. After correcting the batch effect, we combined
the four GEO datasets GSE27276, GSE2378, GSE9944
(GPL8300) and GSE9944 (GPL571) into the expression profiles of 110 samples (control group: 67 cases; POAG group:
43 cases) (Fig.  1A-B). Difference analysis was performed
by the limma package. The screening conditions of different genes were P value < 0.05 and |logFC|> 0.585. Finally, 49
differentially expressed genes were screened, including 30
upregulated genes and 19 downregulated genes (Fig. 1C).
Pathway analysis

We further analyzed the pathways of these 49 candidate
genes in the Metascape database. The results showed that
these candidate genes were mainly enriched in structural
molecular activity, epidermis development, extractive
matrix, oxidoreductase activity, aminoglycan metabolic
process, aging and other pathways (Fig.  2A). Moreover,
we analyzed the protein–protein interaction (PPI) network of genes in different gene sets by Cytoscape software (Fig. 2B).


Suo et al. BMC Genomic Data

(2022) 23:67

Page 3 of 16

Fig. 1  Two-dimensional PCA cluster plot before and after PCA for the combined expression profile. A, B shows two-dimensional PCA cluster
plots before and after PCA for the combined expression profile. After correcting the batch effect, we combined the four GEO datasets GSE27276,
GSE2378, GSE9944 (GPL8300) and GSE9944 (GPL571) into the expression profiles of 110 samples (control group: 67 cases; POAG group: 43 cases). C
DEG volcano plot; red represents upregulated differentially expressed genes, and green represents downregulated differentially expressed genes

Key genes

We analyzed the above 49 differentially expressed genes
by random forest and SVM to screen the key genes
(Fig.  3A-B). According to the comprehensive scores of
the two machine learning methods, we obtained the top

5 genes as the key gene sets, which were (Keratin 14)
KRT14, (Hemoglobin subunit beta) HBB, (Acyl-CoA
Oxidase 2) ACOX2, (Hephaestin) HEPH and Keratin 13
(KRT13) (Fig. 3C). The expression of five key genes in the
POAG group and normal group is shown in Fig. 4.


Suo et al. BMC Genomic Data

(2022) 23:67

Page 4 of 16

A

B

Fig. 2  GO and PPI network analyses of DEGs. A GO biological function enrichment analysis. B PPI network analysis graph. GO, Gene Ontology; PPI,
protein–protein interaction; DEGs, differentially expressed genes


Suo et al. BMC Genomic Data

(2022) 23:67

Page 5 of 16

A

B

C

Fig. 3  Selection of diagnostic biomarkers and identification of key genes. A Select POAG biomarkers by random forest. B Select POAG biomarkers
by SVM. C Key genes extracted from the random forest and SVM methods. SVM, support vector machine; Keratin 14, KRT14; Hemoglobin subunit
beta, HBB; Acyl-CoA Oxidase 2, ACOX2; Hephaestin, HEPH; and Keratin 13, KRT13


Suo et al. BMC Genomic Data

A

(2022) 23:67

Page 6 of 16

B

Kruskal−Wallis, p = 9.3e−09

12.5

C

Kruskal−Wallis, p = 2.2e−05

10

Kruskal−Wallis, p = 9.1e−07

1.3e−05

2.7e−10

10.0

a

Control

a

POAG

7.5

10.0

a

Control

a

POAG

7.5

5.0

5.0
Control

D

9

8

a

Control

a

POAG

7

6
Control

POAG

ACOX2 Expression

9.5e−08

HBB Expression

KRT14 Expression

12.5

POAG

E

Kruskal−Wallis, p = 1.1e−07

11

Control

POAG

Kruskal−Wallis, p = 2.4e−08

8

7

a

Control

a

POAG

6

KRT13 Expression

HEPH Expression

7e−09

10

2.5e−09

9

8

a

Control

a

POAG

7

6
5
5
Control

POAG

Control

POAG

Fig. 4  The expression of five key genes in patients with the POAG group and participants in the normal group. A KRT14 is downregulated in
patients with POAG. B HBB is upregulated in patients with POAG. C ACOX2 is upregulated in POAG. D HEPH is upregulated in POAG. E KRT13
is downregulated in patients with POAG. P value < 0.05. Red represents normal groups, and green represents POAG groups. Keratin 14, KRT14;
Hemoglobin subunit beta, HBB; Acyl-CoA Oxidase 2, ACOX2; Hephaestin, HEPH; and Keratin 13, KRT13

Immune cell infiltration

The microenvironment is mainly composed of immune

cells, extracellular matrix, a variety of growth factors,
inflammatory factors and special physical and chemical
characteristics, which significantly affect the sensitivity
of disease diagnosis and clinical treatment. By analyzing
the relationship between key genes and immune infiltration in the POAG dataset, we further explored the
potential molecular mechanism of key genes affecting the
progression of POAG. The results show that the proportion of immune cells in each patient and the correlation
between immune cells are shown in Fig. 5A-B. Compared
with the normal group, the T-cell regulatory (Treg) level
of samples in the POAG group was significantly higher
(Fig. 5C).
We further explored the relationship between key
genes and immune cells. The five key genes were highly
correlated with immune cells. KRT14 was positively correlated with plasma cells and neutrophils and negatively
correlated with regulatory T cells (Tregs) and mast cell
resetting. HBB was positively correlated with activated
NK cells and monocytes and negatively correlated with
resting mast cells and resting dendritic cells. ACOX2
was positively correlated with CD4 memory resting T
cells and monocytes and negatively correlated with cellular helper T cells and naïve CD4 T cells. HEPH was
positively correlated with memory CD4 + T-cell resetting
and regulatory T cells (Tregs) and negatively correlated
with naive CD4 + T cells and follicular helper T cells.

KRT13 was positively correlated with follicular helper
and plasma cells and negatively correlated with regulatory T cells (Tregs) and resting mast cells (Fig.  6A). We
further obtained the correlation between these key genes
and different immune factors from the TISIDB database, including immune modulators, chemokines and
cell receptors (Fig.  6B-E). These analyses confirmed that
these key genes are closely related to the level of immune

cell infiltration and play an important role in the immune
microenvironment.
Key gene‑related pathways

We used these five key genes in the gene set of this analysis to further explore the transcriptional regulatory network involved in key genes. Relevant transcription factors
were predicted through the Cistrome DB online database,
including 55 transcription factors predicted by KRT14, 92
transcription factors predicted by HBB, 71 transcription
factors predicted by ACOX2, 106 transcription factors
predicted by HEPH and 57 transcription factors predicted
by KRT13. Finally, a comprehensive transcriptional regulatory network of key POAG genes was constructed by
visualization through Cytoscape (Fig. 7).
We studied the specific signaling pathways enriched by
five key genes to explore the potential molecular mechanism of key genes affecting the progression of POAG.
We selected the significantly enriched pathways shown
in Figs.  8 and 9. The pathways enriched with KRT14 by
GO analysis included cell substrate junction assembly,


Suo et al. BMC Genomic Data

A

(2022) 23:67

Page 7 of 16

1.00

Relative Percent

0.75

NK cells activated
Monocytes
Macrophages M0
Macrophages M1
Macrophages M2
Dendritic cells resting
Dendritic cells activated
Mast cells resting
Mast cells activated
Eosinophils
Neutrophils

B cells naive
B cells memory
Plasma cells
T cells CD8
T cells CD4 naive
T cells CD4 memory resting
T cells CD4 memory activated
T cells follicular helper
T cells regulatory (Tregs)
T cells gamma delta
NK cells resting

0.50

0.25

G
GSM
GSM674
GSM674412
GSM674413
GSM674414
GSM674415
GSM674416
GSM674417
GSM674418
GSM674419
GSM674420
GSM674421
GSM674422
GSM674423
GSM674424
GSM674425
GSM674426
GSM674427
S 6 4
G M 674428
GSM74429
GSM44930
GSM44911
GSM44912
GSM44 13
G SM44942
GSM 44943
GSM251944

GSM251607
GSM251608
GSM251609
GSM251610
GSM251611
GSM251612
GSM251495
GSM251496
GSM251497
GSM251498
GSM251499
GSM251500
GSM251501
GSM251502
GSM251503
GSM251504
GSM251505
GSM251506
GSM251507
GSM251508
GSM251509
GSM251510
GSM251511
GSM251512
GSM251513
GSM251514
GSM251515
GSM251516
GSM251517
GSM251518

GSM251519
GSM251520
GSM251521
GSM251522
GSM251523
GSM251524
GSM251525
GSM251526
GSM251527
GSM251528
GSM251529
GSM674530
GSM674408
GSM674409
GSM674410
GSM674411
GSM674431
GSM674432
GSM674433
GSM674434
GSM674435
GSM674436
GSM674437
GSM674438
GSM674439
GSM674440
S 6 4
G M674441
GSM74442
GSM44943

GSM44908
GSM44909
GSM44910
GSM44 38
G SM44939
GSM 44940
GSM251941
GSM251599
GSM251600
GSM251601
GSM251602
GSM251603
GSM251604
GSM251605
GSM251613
GSM251614
GSM251615
GSM251616
GSM251617
GSM251618
GSM251922
GSM251923
GSM251924
SM25 9
2
2519 5
1926
27

0.00

B

**

*

***

C

**

*

***
*

*

**

*

***

**

***

*

**
**

***

**

*

Tissue

***

***

***

***
***

*

a

a

Control

POAG

*

*

**

*

*
*

***
**

*

***

*

*

***

**

***

**

***

*

ils

ils

ph

ph
no

tro

si

eu
N

g
in

ed
Eo

st

tiv
at

re
M

as

tc

el

tc

ls

el

ls

ac

g

ed

in
st

tiv

at

re

ac
as

ce
ic

dr
it
en

en
D

lls

lls

ag
ph

ce
c
iti

ac
M

M

1

2
es

es
ag
ph

ro

ro
ac
M

M

M

0

es

M

yt

es

oc

ag

on
M

ph
ro
ac

lls
ce
K
N

M

g

ed

in
st

tiv
at

re
lls

ac

)

de
a
m

m

ce
K
N

ga
lls
ce

dr
D

lls

lta

er

gs

lp
he

(T
re

ar

ry

ul

to
la

re

gu

fo
l
ce

T

g

ed

in
4

D

T

C

ce

m

lls

em

or

lic

y

or

y

ac

re

st

tiv
at

na
4
D
C

em
m
4
D

C
lls

lls
ce

T

ive

8

lls

D

T

m
lls

ce
B

ce
B

B
B ce
ce lls
lls n
T
Pl m aiv
T ce
as em e
ce lls T
lls C c T ma ory
C D4 ells ce ce
D m C lls lls
4
T m em D4 CD
T ce em or n 8
ce lls o y ai

lls fo ry re ve
r ll a st
T egu icul ctiv ing
ce la ar at
lls tor he ed
g y lp
N am (T er
N K c m reg
K e a s)
ce lls de
lls re lta
a st
M
a M ctiv ing
M crop ono ate
ac h c d
D Ma rop age yte
D en cr ha s s
en dr op g M
dr itic ha es 0
iti c g M
c e e 1
M ce lls s M
M as lls a res 2
as t c c tin
t c el tiv g
el ls r ate
ls e d
a s
Eo ctiv ting

s a
N ino ted
eu p
tro hils
ph
ils

lls

*

ce

*

ce

**
**

ce

*
*

C

***

lls

*

ce

**

T

*
***

**

***

T

***

**

T

*

em
or
y

*
**

m
a

**

as

***

*

Pl

*

**

**

*

***

***

***

−0.5

***

0.0

*

0.0

*

ive

***

*

0.5

**

na

**

1.0

**
*

*

Pearson
Correlation

*** *** ***

**

*
***

**
***

*
**

**

0.1

*

*

0.2

*

0.3

***
*

Fraction

*
**

***

*

0.4

**

**

lls

Neutrophils
Eosinophils
Mast cells activated
Mast cells resting
Dendritic cells activated
Dendritic cells resting
Macrophages M2

Macrophages M1
Macrophages M0
Monocytes
NK cells activated
NK cells resting
T cells gamma delta
T cells regulatory (Tregs)
T cells follicular helper
T cells CD4 memory activated
T cells CD4 memory resting
T cells CD4 naive
T cells CD8
Plasma cells
B cells memory
B cells naive

Fig. 5  Correlation plots of immune cell infiltration analysis. A The proportion of 22 immune cells. B Correlation heatmap of 22 immune cells. Red
represents a positive correlation, purple represents a negative correlation, and the darker the color is, the stronger the correlation. C Plot of the
proportion of infiltration by 22 types of immune cells in normal control samples versus in POAG samples. Blue represents control samples; yellow
represents POAG samples

cell junction assembly and other pathways. The pathways enriched by KEGG included ladder, cancel and so
on Butanoate metabolism and other channels [17]. The
pathways enriched with HBB by GO analysis included
brown fat cell differentiation and corporate cytoskeleton organization. The pathways enriched by KEGG
include Angel processing and presentation and focal
adhesion. The pathways enriched with ACOX2 by GO
analysis included spindle localization and transitional
initiation. The pathways enriched by KEGG include
promote metabolism and pyruvate metabolism. The

pathways enriched with HEPH by GO analysis included
numeric expression repair DNA recognition and lamellipodium organization. The pathways enriched by KEGG
included glycerophospholipid metabolism and beta alanine metabolism. The pathways enriched with KRT13 by
GO analysis included autophagosome organization and
column cuboidal epithelial cell differentiation. The pathways enriched by KEGG included the circuit cycle, TCA
cycle, and cytokine receptor interaction (Fig. 9).

Gene regulatory network analysis of key genes in POAG

We predicted and analyzed the five key genes through
the miRWalk database and ENCORI database to
obtain their possible miRNAs and lncRNAs. First, the
mRNA–miRNA relationship pairs related to these five
key mRNAs were extracted from the miRWalk database. We retained only 35 mRNA–miRNA relationship
pairs with TargetScan of 1 or miRDB of 1 (including 4
mRNAs and 13 miRNAs). Then, we predicted the interacting lncRNAs according to these miRNAs, in which
1112 pairs of interactions (including 2 miRNAs and 823
lncRNAs) were predicted. Finally, we constructed the
ceRNA network through Cytoscape (V3.7) (Fig. 10).
POAG biomarkers

We discussed the prediction efficiency of key genes
through the ROC curve verified by diagnostic efficiency. The results showed that the area under the
AUC of KRT14 was 0.825; the area under the AUC of

(See figure on next page.)
Fig. 6  The relationship between key genes and immune cells. A The five key genes (KRT14, HBB, ACOX2, HEPH and KRT13) were highly correlated
with immune cells. B The relationship between key genes and chemokines. C The relationship between key genes and immunoinhibitors. D The
relationship between key genes and MHC E The relationship between key genes and immunostimulators. MHC, major histocompatibility complex.
Keratin 14, KRT14; Hemoglobin subunit beta, HBB; Acyl-CoA Oxidase 2, ACOX2; Hephaestin, HEPH; and Keratin 13, KRT13

Suo et al. BMC Genomic Data

(2022) 23:67

Page 8 of 16

A

B
KRT13

**

***

***

KRT14

KeyGene

***

***

*

**

***

***

**

**

*

**

*

**

*

**

*

Pearson
Correlation

**

***

0.50

HBB

**

**

*

*

**

**

***

*

**

0.25

***

**

***

−0.25

L2
XC
L9
C
XC
L1
3
C
C
L2
C
XC
L3
C
XC
L6
C
XC
L1
2
C
X3
C
L1

**

C

L4
C

XC

C

C

C

*

C

0

L1
C

2

L1

XC

5

L2

L2

C

C
C

C

8

L8

L1
C

C

C

C

C

**

***

XC

L8

3

*

C

0

L2
C

L1

L2

7

L5
C

C

C

C

3

1

L1

C

C

L2

1

C

C

C

L1

XC

L1

L1
1

XC

L1
9

C

C
C

C

6

L7

L1

C

C

C

C

***

***

**

L1

*

C

7
C
C

C

***

**

XC

**

**

C

***

*

***

C

ACOX2

L2

*

XC

*

L5

0.00

HEPH

chemokine

D

***

**

**

2

H

Immunoinhibitor

B
H
LA P
−D
O
H
LA A
−D
O
H
LA B
−D
R
H
A
LA
−D
M
H
A
LA
−D
M
B

1

TA
P

TA
P

H
LA
−

H
LA

IL
10
C
D
9
TG 6
FB
R
1
C
D
16
0
IL
10
R

B

KD
R
LA
G
3

PD
C
D
LG 1
AL
S9

ID

TG

***

***

TA
P

***

B1

*

***

**

ACOX2

D
Q

FB
1

***

***

***

**

**

*

***

Pearson
Correlation

0.2
0.0

*

−A

***

−0.4

*

***

HEPH

E

***

O
1

ACOX2

*

**

***

**

LA
−

**

*

−F

HEPH

***

−0.2
−0.4

P
H
LA A1
−D
PB
1

−0.2

***

0.4

HBB

B2
M

0.0

**

**

−D

0.2

*

***
***

LA

HBB

*

*

KRT14

Pearson
Correlation

H
LA

***

KRT13

**

LA

*

H

KRT14

KeyGene

***

**

−G

***

KeyGene

KRT13

H

C

MHC

E
**

**
**

*

1
D

EN

TP

LG

80
D

40

C

D
C

4

14
SF

FR

R

27
C

XC

TN

D

C

LT
A

28
D

C

86
D

N

FR

TN

PV

IL
2R

B

0.0
−0.4

*

C

***

E

***

0.2
−0.2

*

T5

***

S

***

SF
TN 8
FS
F
TN 4
FS
F9

**

R

***

A

***

Immunostimulator

Fig. 6  (See legend on previous page.)

Pearson
Correlation

*

*

**

IC

70

C

D

TN

D

48

*

C

G
SL

O

13
B

SF

IC

25
SF

FR

FR

*

***

TN

4

C

D

*

40

SF

***

TN

FR

IL
6

*

SF

*

TN

TN

FR

SF

9

*

FR

ACOX2

17

***

4

HEPH

*

*

*

M

***

F1

**

FS

HBB

TN

KeyGene

KRT14

***

**

*

O

**

IC

KRT13


Suo et al. BMC Genomic Data

(2022) 23:67

HCFC1
ELL2
IRAK4 LDB1
PHIP
RBPJ
E2F6
CEBPD
UBE2I
EZH1
RARA
MYB
NFE2
FANCD2 NFYB
KMT2A
NEUROG2
GFI1B
LARP7
TBP
ZNF24
ASCL1STAT2
BHLHE40
USF1

RBBP5
SIRT1
NOTCH1
RING1

NFIA

HBB

WDR5
POLR2M

T
ZNF423

Page 9 of 16

EGR3

STAT5B

ZEB2 JUNB
MAZ
CDK9
ZNF589

ZNF143
PROX1
TFAP2C
HOXC9 TRIM24

KRT13

SMARCB1
GRHL2

SMAD1
EP300
GATA3
TEAD4
IKZF1CEBPBGATA1
NR2F6
GATA4NR2F1
RCOR1
NR2F2
NR3C1
GATA2
CTCFTAL1 BCL11A
POLR2A
SMC3
JUND
ESR1 SOX2
FOXA2
IRF5
TERC
SPI1
MYC
REST
MAX
SMC1A

PPARG
CREB1
SUMO2
GATA6 STAT3
BRD4 RAD21
NKX2-1
STAT1 TET2 ATF3CEBPA
YY1
PR
PKNOX1
AR TP63 ATF1
FOXA1
RUNX1 MYNN
TFAP2A
CDK8
KLF1 EZH2

OSR2
PRDM10
IKZF2

REPIN1

RNF2
ZNF766
SMARCE1STAT5ASMARCA4
MTA2
DPF2
RFX1
TRIM28

TCF7 ARID3A
ZBTB7A JUN
HMBOX1

EGR2

KRT14

CREBBP
CDK2
STAT4
TP53 POLR3D
POLR2AphosphoS5
ZNF41
SPIB
MBD3
NBN
DACOR1
NRF1
ZNF793
NR1H3
HIRA
KLF11
INTS11

H2AZ
FLI1
HDAC2

HEPH

RELA
SNAI2
EOMES

FAIRE

ACOX2

BCL3 CUX1HDAC1

TCF12

IRF1

HEY1

FOSL1

PML

ELF1KDM1A

E2F5
SMAD3

NCOA1
UBTF
NFRKB
PYGO2

KLF16

FOS

SUPT5HBCLAF1 ZNF83
POU5F1
HNF4A
HOXB13
CDX2 CBX3 ZKSCAN8
MBD2
KRAB
PRAME ZNF197
ZFX

EBF1
ZMIZ1

PBX2

MED1

MAFG LHX2
KDM5B
TBL1XR1
MAFF
SIRT6
TCF7L2
ZNF384
TTF1

RXRA

GABPA

MEF2A
ZKSCAN1

ZBTB17

FOXO3

ZBTB33

SOX13
HNF4GHES2

CEBPG
ATF2
DAXX
JMJD1C

SP1
ZNF382
ERG
SRF
FOXO1
SUMO1
RFX5 OTX2
ARRB1
LYL1

Fig. 7  A comprehensive transcriptional regulatory network of key POAG genes (KRT14, HBB, ACOX2, HEPH and KRT13) was constructed by
visualization through Cytoscape. Keratin 14, KRT14; Hemoglobin subunit beta, HBB; Acyl-CoA Oxidase 2, ACOX2; Hephaestin, HEPH; and Keratin 13,
KRT13

HBB was 0.740; the area under the AUC of ACOX2
was 0.778; the area under the AUC of HEPH was
0.801; and the area under the AUC of KRT13 was
0.816. These results show that these five key genes
have good prediction efficiency for POAG and may
better predict the occurrence and development of diseases (Fig. 11A-E).
Drug targeting prediction in POAG

We divided the differentially expressed mRNAs into
upregulated and downregulated groups and predicted
the drug targets of the differentially expressed genes
through the Connectivity Map database. The results
showed that the expression profiles of drug disturbances,
such as avrainvillamide-analysis-3, cytochalasin-D, NPI2358, oxymethylone and vinorelbine, were negatively
correlated with the expression profiles of disease disturbances (Fig.  12). This finding indicated that these drugs
may reduce or even reverse the POAG disease state.

Discussion
POAG is a chronic retinal neurodegeneration disease
characterized by changes in the anterior and posterior
segments of the eye; in addition, serious damage may
be detected in the TM and ONH [1, 5, 8]. Lowering IOP
using drugs or surgery is the only intervention currently
available [6]. However, clinical evidence indicates that
lowering IOP does not prevent progression in all POAG

patients. Consequently, non-IOP factors are involved
in the disease [2]. With the rapid development of science and technology, bioinformatics provides a powerful strategy for the screening of molecular markers [18].
Biomarkers reflect changes at the molecular level and can
accurately monitor pathological changes in the TM and
ONH and provide important information for the diagnosis of POAG [5, 15, 19–23]. Nevertheless, the detection of
POAG lesions using molecular biology methods is suboptimal, and treatments are currently in a limited pharmacotherapy phase [6].


Suo et al. BMC Genomic Data

A

(2022) 23:67

Page 10 of 16

B

0.6

0.3

0.0
GO_ESTABLISHMENT_OF_MITOTIC_SPINDLE_LOCALIZATION

Enrichment Score

0.3

Enrichment Score

0.6

0.0
KEGG_CITRATE_CYCLE_TCA_CYCLE
KEGG_GLYCOSPHINGOLIPID_BIOSYNTHESIS_LACTO_AND_NEOLACTO_SERIES

GO_NUCLEOTIDE_EXCISION_REPAIR_DNA_DAMAGE_RECOGNITION
−0.3

−0.3

GO_PROTEIN_TARGETING

KEGG_MATURITY_ONSET_DIABETES_OF_THE_YOUNG

GO_REGULATION_OF_PROTEIN_TARGETING_TO_MITOCHONDRION

KEGG_PROPANOATE_METABOLISM

GO_SPINDLE_LOCALIZATION

KEGG_PYRUVATE_METABOLISM

GO_TRANSLATIONAL_INITIATION

KEGG_VALINE_LEUCINE_AND_ISOLEUCINE_DEGRADATION

−0.6

high expression<−−−−−−−−−−−>low expression

high expression<−−−−−−−−−−−>low expression

C

D
0.50

0.0

−0.4

GO_ANTIGEN_PROCESSING_AND_PRESENTATION_OF_PEPTIDE_
OR_POLYSACCHARIDE_ANTIGEN_VIA_MHC_CLASS_II
GO_BROWN_FAT_CELL_DIFFERENTIATION

Enrichment Score

Enrichment Score

0.4
0.25

0.00
KEGG_ANTIGEN_PROCESSING_AND_PRESENTATION
KEGG_FOCAL_ADHESION
−0.25

GO_CORTICAL_CYTOSKELETON_ORGANIZATION

GO_INACTIVATION_OF_MAPK_ACTIVITY

KEGG_SYSTEMIC_LUPUS_ERYTHEMATOSUS

GO_POSITIVE_REGULATION_OF_OSTEOCLAST_DIFFERENTIATION
−0.8

−0.50

GO_PROTEIN_LOCALIZATION_TO_LYSOSOME

KEGG_TYPE_I_DIABETES_MELLITUS

high expression<−−−−−−−−−−−>low expression

high expression<−−−−−−−−−−−>low expression

E

KEGG_GLYCINE_SERINE_AND_THREONINE_METABOLISM
KEGG_PROTEASOME

F

0.6

0.50

0.0

GO_LAMELLIPODIUM_ORGANIZATION
−0.3

0.25

0.00
KEGG_ALANINE_ASPARTATE_AND_GLUTAMATE_METABOLISM
KEGG_ARRHYTHMOGENIC_RIGHT_VENTRICULAR_CARDIOMYOPATHY_ARVC

GO_NUCLEOTIDE_EXCISION_REPAIR_DNA_DAMAGE_RECOGNITION
GO_PYRUVATE_METABOLIC_PROCESS
GO_REGULATION_OF_DIGESTIVE_SYSTEM_PROCESS

−0.6

Enrichment Score

Enrichment Score

0.3

−0.25

KEGG_BETA_ALANINE_METABOLISM
KEGG_GLYCEROPHOSPHOLIPID_METABOLISM

GO_REGULATION_OF_PROTEIN_TARGETING_TO_MITOCHONDRION

KEGG_PROPANOATE_METABOLISM

GO_RESPONSE_TO_MUSCLE_STRETCH

KEGG_TRYPTOPHAN_METABOLISM

high expression<−−−−−−−−−−−>low expression

high expression<−−−−−−−−−−−>low expression

Fig. 8  GSEA of GO and KEGG enrichment analysis for the key genes. A-B GSEA of GO and KEGG enrichment analysis for ACOX2. C-D GSEA of GO
and KEGG enrichment analysis for HBB. E–F GSEA of GO and KEGG enrichment analyses for HEPH. GSEA, gene set enrichment analysis. Hemoglobin
subunit beta, HBB; Acyl-CoA Oxidase 2, ACOX2; and Hephaestin, HEPH


Suo et al. BMC Genomic Data

(2022) 23:67

A

Page 11 of 16

B

0.6

0.3

0.0
GO_AUTOPHAGOSOME_ORGANIZATION

Enrichment Score

Enrichment Score

0.4

0.0
KEGG_CITRATE_CYCLE_TCA_CYCLE

GO_COLUMNAR_CUBOIDAL_EPITHELIAL_CELL_DIFFERENTIATION

KEGG_CYTOKINE_CYTOKINE_RECEPTOR_INTERACTION
−0.3

GO_EPIBOLY
−0.4

KEGG_JAK_STAT_SIGNALING_PATHWAY

GO_NEGATIVE_REGULATION_OF_NEUROLOGICAL_SYSTEM_PROCESS

KEGG_MATURITY_ONSET_DIABETES_OF_THE_YOUNG

GO_NUCLEOTIDE_EXCISION_REPAIR_DNA_DAMAGE_RECOGNITION

KEGG_NEUROACTIVE_LIGAND_RECEPTOR_INTERACTION

GO_REGULATION_OF_SENSORY_PERCEPTION

KEGG_VALINE_LEUCINE_AND_ISOLEUCINE_DEGRADATION

high expression<−−−−−−−−−−−>low expression

C

high expression<−−−−−−−−−−−>low expression

D

0.50

0.6

0.25

0.0
GO_CELL_JUNCTION_ASSEMBLY
−0.3

GO_CELL_SUBSTRATE_JUNCTION_ASSEMBLY

Enrichment Score

Enrichment Score

0.3

0.00

KEGG_BLADDER_CANCER

−0.25

KEGG_NITROGEN_METABOLISM

GO_NEGATIVE_REGULATION_OF_SYNAPSE_ORGANIZATION
GO_PROTEIN_REFOLDING
−0.6

GO_TRANSFERRIN_TRANSPORT

high expression<−−−−−−−−−−−>low expression

KEGG_BUTANOATE_METABOLISM
KEGG_DORSO_VENTRAL_AXIS_FORMATION

GO_NEGATIVE_REGULATION_OF_ERBB_SIGNALING_PATHWAY

−0.50

KEGG_P53_SIGNALING_PATHWAY
KEGG_VIBRIO_CHOLERAE_INFECTION

high expression<−−−−−−−−−−−>low expression

Fig. 9  GSEA of GO and KEGG enrichment analysis for the key genes. A-B GSEA of GO and KEGG enrichment analysis for KRT13. C-D GSEA of GO and
KEGG enrichment analysis for KRT14. GSEA, gene set enrichment analysis. Keratin 14, KRT14; and Keratin 13, KRT13

In the present study, to screen the pathogenic genes
involved in POAG, an integrated analysis was performed
by using microarray datasets in glaucoma derived from

the GEO database. The functional annotation and potential pathways of DEGs were additionally examined by GO
and KEGG enrichment analyses. A POAG-specific transcriptional regulatory network was constructed to identify
crucial transcription factors that target the key genes in
patients with POAG. We used random forest and the SVM
method to screen five potential key genes in both human
TM and ONH tissue. A growing number of researchers
realize that immune infiltration is related to the diagnosis of
POAG [10–12]. Therefore, analyzing the pattern of POAG
immune cell infiltration and finding specific diagnostic
markers have profound significance for POAG patients.
Subsequently, the CIBERSORT algorithm performed
deconvolution analysis on the immune microenvironment
to assess the proportion of immune cells in POAG.

We identified key differentially expressed genes
between POAG and normal tissues by performing
a combined analysis of TM and ONH. A total of 49
DEGs were identified, including 19 downregulated
genes and 30 upregulated genes. These 49 differentially expressed genes were further analyzed by random forest and SVM to screen the key genes. Five
genes (KRT14, HBB, ACOX2, HEPH and KRT13)
were significantly changed. We found that these key
genes were highly correlated with immune cells. The
immune microenvironment is composed of a variety of
lymphocytes, such as T cells, B cells and macrophages,
etc. [24]. From the immune infiltration analysis, we
found that there was a significant difference in the relative cell content of 22 types of immune cells (e.g., B
cells naïve, B cell memory, plasma cells, T cells CD8, T
cells CD4 naïve, T cells CD4 memory resting) in normal control samples versus in POAG samples.

Suo et al. BMC Genomic Data

(2022) 23:67

Page 12 of 16

hsa-miR-1227-3p
hsa-miR-544b
hsa-miR-6508-5p
AC025048.4

TRDN-AS1
STEAP3-AS1
LIX1L-AS1
AC011447.3

BX649632.1

AC000068.1
AC023302.1

hsa-miR-6851-3p
hsa-miR-6511a-5p

AC012074.1

SNHG11

AC096531.2 AC245140.2

hsa-miR-1910-3p

hsa-miR-6876-3p

HEPH

ADAMTSL4-AS1
AC069278.4
AC004491.1
LINC00599
SOX2-OT
AC092718.4 AC009812.4
C17orf77
AC115837.2
AC007405.1
LINC01471
SPRY4-AS1
LINC01342 AC138430.1 AC019118.2
AC006213.2
LMCD1-AS1AC097634.3
AC026471.4
PINK1-AS
AC244197.2
CSNK1G2-AS1
AC008555.8
AC087741.1
AC124057.1
AC010834.1
AL445183.2
AL355388.2

AC226118.1
AC022400.3
AC023194.3
MHENCR
AC012100.3
SDCBP2-AS1
AC018659.2
LINC01639
AC002076.1
AC068473.3
SLC2A1-AS1
AC006538.1
AC093525.7
CTD-2132N18.4
AC037459.3AC010336.1
AL033543.1
AC110048.2
ZNF433-AS1
AL159152.1
AC127024.2
MAGI2-AS3
AC012181.2
Z95115.1
AL118522.1
BISPR
MIR503HG
AC025569.1
AFAP1-AS1
AL135999.1
AC090970.2

AC022558.1 AC005089.1
CARMN
AL121772.3
AC004466.1
C1orf220
AC106882.1
MIR34AHG
AL353795.3
AC007383.4
VASH1-AS1
AC106820.4
AC015799.1
AP003419.3
SPINT1-AS1
AC080080.1AL583810.1 AC144831.1
PWAR1 THAP7-AS1
AC023509.1
AC009061.2
HEIH AC145423.3
AC009145.4
MIR193BHG
H19
EBLN3P AC012513.3
LINC00472
U73166.1
AC004803.1
LINC01239
AL591848.4
EP300-AS1
AL583810.2

AC008109.1
TTTY14
LINC00304AC080023.1
FENDRR
AC002116.1
AL158835.4 AL353898.3
AL133243.4
ZNF460-AS1
AC139795.2
AC008686.1 AC237221.2 LINC00637
PRICKLE2-AS1
AC012313.3
TTC28-AS1
AC012676.5
AC018629.1
LINC00689
AC010327.3
AL049569.1
LINC01114
OVCH1-AS1
LINC02466 HCG11
AC092117.1
AP006623.1
AC092279.2
AC026461.1
AC138951.2
AC011472.1
AC018647.2
AP003108.4
AC016722.3

AC005076.1
AL121832.3
SNHG5
AP000347.2
HNF1A-AS1
AC040169.1
RP4-565E6.1
AC020909.2
AC105206.1
AC092171.2
AC012467.2
AC012313.9
LINC01165
GASAL1
AL136040.1
AC022150.4
AC027031.2
AL024507.2SNHG20
AC013448.1
AC083973.1
AL121983.2
AC239809.3
AL160006.1
AF127577.4
AC073487.1
AC015883.1
AC130462.1
AC110285.2
AC233702.9
AC022498.2 AC004656.1

SEPSECS-AS1
AC006333.2
NUTM2B-AS1
ST20-AS1 SLC25A5-AS1
AC073857.1
AATBC
AC091053.1
AC079922.2
AL021707.7
TBX2-AS1 AC131212.2
AC011498.7
AKT3-IT1
AC005394.2
HOXC-AS1
AC005899.5 AC093292.1
AC010913.1
AC079316.2
LINC01252
AP000577.1
AC025171.1
LINC01748
AC087289.5
PAXIP1-AS1
AC007383.2
AC003102.1
GATA2-AS1
AC099811.1
AC093827.4
SLC16A1-AS1
RAP2C-AS1 AC126696.3

Z69733.1
LINC00894
NBR2
AC008105.1
AC006141.1
KCNQ1OT1
SOS1-IT1
NPSR1-AS1 AC120114.4
ASB16-AS1
TMEM161B-AS1
AC015813.1
AC010761.1 AC092756.1
AC114980.1
AC008764.8
AC011978.2
AP002754.1
AC012358.3
FAM95B1
AP004147.1 AC122688.3
AP001157.1
AL354707.1
LINC00958
AC116049.2
AL845472.1
LINC01772
MMP25-AS1
GAS6-AS1
AC015712.6
CCDC18-AS1 AC139887.2
AC026271.3

AC147651.1
AL162231.4
AC015909.5OLMALINC AL356481.3
AC099494.3
JPX
AL355102.3
AC023794.3 AC013403.2
AC137834.2
AC015726.1
AC087203.3
AC092068.3
SLC12A5-AS1
AC005831.1
AL133243.3
AL117339.4
PSMA3-AS1 RAD51-AS1
MALAT1
AL008721.2
AC022075.1
AC099778.1
AL022069.1
AC026362.1FTX
LINC02001
SLC5A4-AS1
AC090673.1
AC007610.4
AC007036.3AC009078.3
AL031595.3
SBF2-AS1 LINC02575
STX18-AS1

AC114488.3
AL450998.2
AC005480.1
MIR4697HG
C22orf34
AC017116.1
AC009948.4
A1BG-AS1
AL049649.1
AL162586.1
OIP5-AS1
LINC00704
SNHG7 AC027682.7
AC004918.3
AC110285.1
AC073548.1
AC020658.6
AL355987.4
AC021491.1
AC106739.1
LINC01010PWAR5
AC008467.1
AC068152.1
LINC01235 GMDS-AS1
AL365205.3
AC020922.3
AP001469.3
AL513190.1
AC009948.5
SLC25A25-AS1

AC016831.5 AC007191.1 AC108704.1
AC011840.2
AC099343.3
AC010999.3
AC004490.1
NRSN2-AS1 LINC01446
AC013652.1
RNF139-AS1
AL035425.3
STX17-AS1
MIAT
AC087393.2CXXC4-AS1AC140479.5
AC106820.3
LINC00982
AC009088.1
AL158835.1
AC027307.3
AC104581.4
ATP2C2-AS1
AC002454.1
AL139287.1
CASC15
AL355385.1
LINC01444
AC174065.1
IGF2-AS
GUSBP11
PTPRD-AS1
AC078880.4
AC009041.2

RAB30-AS1
AL136084.3
NR2F2-AS1
AL137003.1
AC092139.3
AL161669.1
AC068888.1
AC022098.1
AL928654.1
TEN1-CDK3
AC093278.2
AL662844.4
LINC01220
AC027601.4
AC021078.1
TPT1-AS1
LINC02387
SH3BP5-AS1
AC092135.3
AC132872.3
MIR4435-2HG
LINC00330 AC105105.2
AC068338.2 AC090181.2
LRRC75A-AS1 MATN1-AS1 LINC01806
FBXL19-AS1
AL358472.4
AC008957.3
AC110058.1
AL138889.1
MIRLET7BHG

STPG3-AS1
AL512353.1
LINC01694
AL596244.1
GATA3-AS1
AC099850.1
AC131212.3
AC021945.1
AC010327.5
AC008915.2
Z94721.2
LINC01719
AC118344.2
AC104581.3
AC010336.4
AC012313.8
AP000251.1
AL162231.2
AC010680.3
AL132780.4
AC011773.4
AC096992.2
AC005154.1
AL590705.1 WWTR1-IT1
AC023886.1
AC012510.1
AC005229.4
LINC00824
PEG13
RUSC1-AS1

NNT-AS1 AL928654.4
LINC02389
AC009084.3
UCKL1-AS1
AL358473.1
SNHG14
SNAI3-AS1
AC004687.1
AGAP2-AS1
AL035681.1
AP003392.4
AC068888.2
AC245452.1
AP003071.3
AL691432.2
AC116407.4
LINC00926 AL645728.1
AC004925.1
SNHG12
AC024940.6
AL117335.1AL662907.2
AC078777.1 ALOX12-AS1
U62317.4
AL390957.1
AC109460.3
VCAN-AS1
AC009716.2
GABPB1-AS1
AC115618.3
AC103702.1 MINCR

AL359878.1
AC105460.1
AC024896.1
SNHG22
AC079414.3
CACNA1C-AS1
AC016738.1
RPARP-AS1
AC025423.4 AL356488.3
LINC00963
AC010969.2
AL136309.2
APCDD1L-AS1
MAFG-AS1
AC092384.3
AC097376.2
AL035413.1
C9orf106 AC139530.1
AP002748.3
AL356512.1
AC005005.4
HELLPAR
AL157392.3
AC007066.2AC005520.2
AL390719.2MEG3
RAMP2-AS1
AF117829.1
AC026471.1
AL133367.1
AC125807.2

AC145098.2
AL662795.1
AC091152.2
AC090515.2LINC01515
AL021392.1
AC015922.4
SPAG5-AS1
ZFHX2-AS1
NOP14-AS1
AL359258.1
AC009118.2
LINC02019AC243964.3
HOTAIR
AC003965.2
AP001267.2
XIST
AC009948.1 AC011498.1
AC092301.1
NAPA-AS1
AC012501.1
AP006621.1
AP004609.3
AC069281.2
AC011462.2
AC008969.1 AC027290.2
AC002070.1
LINC00886
AP001020.2
PPP1R12A-AS1
LINC00863

AC007743.1
AC104667.2
AL596202.1
AC034111.1
AP000692.1
LINC01014LINC00630
BAIAP2-AS1
AC067852.1
AL137782.1
BAALC-AS1 LIPE-AS1
AC005329.1
AP000654.1
AC090114.2
AC060780.1
PTOV1-AS2 MIR31HG
AL354892.3
PART1
SNHG1SLC9A3-AS1
AC016876.2
AL390066.1
AC122710.3
ATG10-IT1
LINC01547
AC138150.2 AL645608.1
AC020915.3
AC012073.1
AC012464.3
PVT1
AL139220.2
LINC00921

AL022311.1
AC010834.2AL133415.1
DLX6-AS1 ARRDC1-AS1TRIM52-AS1
AL133551.1
LINC00896
AC111170.3 NFYC-AS1
PTPRG-AS1
LINC01090
AC084125.2
AC138028.2
AF127577.3
AL356488.2
AP003469.4
AC013564.1
BACH1-IT2
LINC00294
CYP1B1-AS1
COLCA1 AC004943.2
LINC01184
AC016876.3
FALEC
AL138781.2
AC009779.2
AC000123.3
AP003108.1
TTTY15
RASSF8-AS1 LINC00839
LINC00957
AC012615.2
AL121672.1

AC138655.1
NUP50-AS1
SNHG17
AP000894.2
AC004233.3
LINC00667 AC005224.3
AC124319.1
AC067852.3
AC079336.7
Z95331.1
AC012615.6
AC134407.3 LINC01105
AC010997.5
AC138028.6
AC015689.1
LINC01521
AC114316.2
AP000350.6
AC026740.1
AC104982.1
WASIR1
MIR29B2CHG
AC011815.1
ZNF561-AS1
EML2-AS1
AL049836.1
AL137145.1
AC092127.1
AL160314.2
NARF-AS1

AC006305.1
LINC01560
AL031009.1
AL158166.1
AC021016.3
SUCLG2-AS1
UMODL1-AS1
AC025048.2
AC130324.1
FIRRE
AC020763.4
XXyac-YX155B6.6
AC011933.2
NEAT1
AC106782.2
MIR9-3HG
AL513534.1
CIRBP-AS1
AC099487.1
AC007878.1
CSAG2
AC008875.1
AC104109.4
AL592078.1
AC005261.1
AC083843.3
LINC01703
FOXN3-AS1 Z97192.1
CRNDE
AC099508.2 LINC01126 AC026979.2

AC073335.2
AL109613.1
AC022400.7
AL158212.3
STAG3L5P-PVRIG2P-PILRB NORAD
AL845472.2
AC009133.1
AC239804.1 CSAG3
AC102945.2
AC126474.2
AP002364.1
SOCS2-AS1
TTN-AS1
AL139099.2
AC009090.5
AC103691.1
AC023794.7
AL035252.3
AC073957.2
AC027020.2
AC007388.1
AC010530.1
MCCC1-AS1
AL096870.2
AL031587.4
AC005562.1
AC016705.2
SNHG15
AC078846.1
OSER1-AS1

AL590705.5
AC087388.1
AC027682.2
AP001273.1
AL645608.8LINC01836
LINC01743
AC016876.1
CCDC183-AS1
AC022819.1 AC006946.2 AC087885.1
AC011491.2 AC027796.4
AL683813.1
AC092117.2
LINC01006
PPP3CB-AS1
AC006206.1
MALINC1
AC079466.1
AP003108.5
AC020917.4 FP325335.1
AC026356.1
LINC01138
AC005332.7
AC103746.1
AP001005.3
AC073842.2
BBOX1-AS1
LINC01089 WDFY3-AS2
AC148477.2
AL450326.1
AL035587.1

AL136988.1
AL022323.1AC090337.1
AC092118.2
EPB41L4A-AS1
AL157700.1
AC007216.3
LINC01885 AC105105.4
AC055713.1
ST8SIA6-AS1
BMS1P4
AL391244.1
RN7SL832PMIR200CHG
PAXIP1-AS2
AC007848.1
AC092111.1
AC069277.1
AC008124.1 AC110285.6AC148477.3
AL390038.1
AC003681.1AC025265.1
SLFNL1-AS1
LINC00482
AC093227.1
WASIR2
LINC00511
AL445645.1
AC012181.1
AP000766.1
AL137077.2AL137786.1
THAP9-AS1
AC008395.1

AC009549.1
AC027307.1
AP000662.1
TMEM9B-AS1
AC010536.1
AC026362.2 AL590627.2
AC008738.7 AC092069.1
AC005785.2
AC011337.1 AL355472.4
AC008771.1SNHG19
AC084125.4
AL645568.1
AF111169.1
AC010478.1
AC092142.1
ZEB1-AS1
AC027237.3 AC004951.1 AC008764.10
AC079228.1
AC007249.2
AC018521.1
AP000873.2
AC104447.1
AL592158.1
LINC02478
VPS9D1-AS1
LINC02381
AC064836.3
AC016394.2
CCDC144NL-AS1
FP325330.3

AC091181.1
AC013643.3
AC242426.2
AC026124.1
LINC00649
AC011471.3
LINC00641
SRP14-AS1 AL157931.2
CTBP1-AS2
AC002310.2
AC018628.1
AL365181.2
AL157702.2
AC078883.1
AJ239328.1
AP002761.4
AC006504.1
AL596451.1
Z83843.1
AC091230.1
AC022107.1
AC012065.4
AL160394.1
AC022211.4

HBB

AC040169.3

hsa-miR-520g-3p

AC008014.1

hsa-miR-361-3p

hsa-miR-6881-5p
hsa-miR-6774-5p
hsa-miR-5572
AC104964.4
hsa-miR-6751-3p

AL691482.3
AC011558.1

BASP1-AS1

KRT13

KRT14

Fig. 10  The five key genes through the miRWalk database and ENCORI database to obtain their possible miRNAs and lncRNAs. Keratin 14, KRT14;
Hemoglobin subunit beta, HBB; Acyl-CoA Oxidase 2, ACOX2; Hephaestin, HEPH; and Keratin 13, KRT13

KRT14 is a member of the type I keratin family of
intermediate filament proteins [25]. KRT 14 is expressed
in a variety of cells in humans. It is recovered as a heterodimer with KRT5 and forms the cytoskeleton of epithelial cells [26, 27]. Gautam et  al., using multispecies
single-cell transcriptomic analysis of the human eye,
found that KRT 14 was expressed in corneal epithelial
cells [28]. However, previous studies did not perform a
deeper investigation of POAG. In our study, we found

that KRT14 was expressed in both TM and ON tissues
in humans. KRT14 was highly correlated with immune
cells (plasma cells, neutrophils, etc., and negatively correlated with regulatory T cells (Tregs) and mast cell
resetting. HBB (hemoglobin subunit beta) is encoded by
the HBB gene on human chromosome 11 [20]. HBB was
positively correlated with activated NK cells and monocytes and negatively correlated with resting mast cells
and resting dendritic cells. The ACOX2 gene encodes
the enzyme Acyl-CoA oxidase 2 in human autosome
3, oxidizing the Coenzyme A esters of bile acid and ditrihydroxycholestanoic acid intermediates. It is located
in the peroxisomes of cells and has a tripeptide at the
C-terminal end of the protein formed by serine-lysineleucine that serves as a peroxisome localization signal.
Deficiency of this enzyme causes an accumulation of

fatty acids and bile intermediates, generating Zellweger
syndrome. ACOX2 was positively correlated with CD4
memory resting T cells and monocytes and negatively
correlated with cellular helper T cells and naïve CD4 T
cells. HEPH (Hephaestin) is involved in the metabolism
and homeostasis of iron and possibly copper [29]. It is a
transmembrane copper-dependent ferroxidase responsible for transporting dietary iron from intestinal enterocytes into the circulatory system. HEPH was positively
correlated with memory CD4 + T-cell resetting and regulatory T cells (Tregs) and negatively correlated with naive
CD4 + T cells and follicular helper T cells. The KRT13
(keratin 13) gene encodes a type I cytokeratin that is
expressed in the differentiated cells of noncornified stratified squamous epithelia [30, 31]. KRT13 is positively correlated with T-cell follicular helper and plasma cells and
negatively correlated with T-cell regulatory (Tregs) and
resting mast cells.
The ocular innate immune response relevant to glaucoma involves the complement cascade, microglia, astrocytes, and Müller cells. Autoantibodies, tumor necrosis
factor (TNF)-alpha, and other products of adaptive B
and T cells are also featured prominently in the glaucomatous adaptive immune response. However, research
has not provided clear immune infiltration among these

Suo et al. BMC Genomic Data

(2022) 23:67

Page 13 of 16

Fig. 11  Verification of biomarkers. A Prediction ROC for ACOX2. B Prediction ROC for HBB. C Prediction ROC for HEPH. D Prediction ROC for KRT13. E
Prediction ROC for KRT14. Keratin 14, KRT14; Hemoglobin subunit beta, HBB; Acyl-CoA Oxidase 2, ACOX2; Hephaestin, HEPH; and Keratin 13, KRT13

cells [9]. Thus, in our study, based on the immune infiltration analysis between key genes and immune cells, we
divided the differentially expressed mRNAs into upregulated and downregulated groups and predicted the drug
targets of the differentially expressed genes through the
Connectivity Map database. The results showed that the
expression profiles of drug disturbances such as avrainvillamide-analysis-3, cytochalasin-D, NPI-2358, oxymethylone and vinorelbine were negatively correlated with
the expression profiles of disease disturbances. NPI-2358
and vinorelbine are tubulin inhibitors. Cytochalasin-d is
an actin polymerization inhibitor. Oxymetholone is an
androgen receptor agonist. avrainvillamide-analog-3is
a nucleophosmin inhibitor. This finding indicated that
these drugs may reduce or even reverse the POAG disease state.
There were several limitations to this study. First, due
to the limited number of samples, there is still a need to
confirm these preclinical observations in future clinical
studies for novel biomarkers. Second, CIBERSORT is
based on the principle of linear support vector regression and uses gene expression data in reverse to deduce

the result of immune cell infiltration. Indeed, it is not
based on experimental data, and further verification

of immune cell infiltration by a large number of experiments is needed. Third, we performed mining and analysis of previously published data; although some previous
studies showed similar results, the related molecules and
their mechanisms at the molecular, cell, and tissue levels
require validation.
In conclusion, the overlap of the random forest and
SVM algorithms was obtained, and five key genes were
eventually recognized in both human TM and ONH tissue. We found that it may be used as a diagnostic marker
for POAG. To understand POAG development, GO and
GSEA of the selected genes supplied a more specific
molecular mechanism. To date, the relationship between
key genes and immune infiltration in both TM and ONH
tissues has been rarely reported. The mechanism of key
genes and immune infiltration-related factors in the diagnosis of POAG remains to be explored. Further investigation of these immune cells may identify targets of
immunotherapy for POAG and help POAG patients benefit from immunomodulatory therapy.


Suo et al. BMC Genomic Data

(2022) 23:67

Page 14 of 16

Fig. 12  Drug targeting prediction in POAG. A Avrainvillamide-analysis-3. B NPI-2358. C vinorelbine. D cytochalasin-D. E oxymethylone

Materials and methods
Material and data

The datasets used in the present study were downloaded
from the National Center of Biotechnology Information
(NCBI) Gene Expression Omnibus (GEO). Four sets of

POAG mRNA (GSE27276-GPL2507, GSE2378-GPL8300,
GSE9944-GPL8300, GSE9944-GPL571) data were
obtained. The detailed information is shown in Table  1.
In total, 43 POAG patients and 67 normal individuals
were enrolled in our study from these four datasets.
Screening of key genes

The differentially expressed genes (DEGs) in the GEO
dataset (P < 0.05 and |logFC|> 0.585) were screened by
difference analysis, and then the candidate gene sets were
further screened by random forest and SVM algorithms.
Among them, random forest is an integrated learning
algorithm based on a decision tree. Multiple samples are
selected from the sample sets as a training set by sampling with replacement. The decision tree is generated

from the obtained sample sets by sampling. At each generated node, features without repetition are randomly
selected. Based on optimal features to divide the sample
sets, the prediction results are determined. In this study,
the features were evaluated by the random forest algorithm. Using %IncMSE to evaluate the significance, 1000
trees were built and repeated 50 times. SVM is a machine
learning method based on SVM. Optimal variables were
identified by deleting SVM-generated feature vectors.
For the sake of further recognizing the diagnostic value
of disease biomarkers, a support vector machine model
was established according to the “e1071” software package. Finally, the top features were retained for follow-up
analysis.
GO and KEGG functional annotation

The differentially expressed genes were functionally
annotated using the Metascape database (www.​metas​

cape.​org) to comprehensively explore the functional
correlation of these genes. For the specific genes, we


Suo et al. BMC Genomic Data

(2022) 23:67

Page 15 of 16

Table 1  Characteristics of four datasets of POAG mRNA data
Series matrix file data Annotation file

Number of participants in the
POAG group

Number of participants in the
normal group

Sample type

GSE27276

GPL2507

19

17

Trabecular meshwork

GSE2378

GPL8300

7

6

Optic nerve head

GSE9944

GPL8300

13

6

Optic nerve head

GSE9944

GPL571

6

36

Optic nerve head

performed Gene Ontology (GO) analysis and Kyoto
Encyclopedia of Gene Genome (KEGG) pathway analysis. Minimum overlap ≥ 3 and P ≤ 0.01 were considered
statistically significant.
Analysis of immune cell infiltration

CIBERSORT is widely used to evaluate the type of immune
cell in the microenvironment. The tool is based on the
principle of a linear support vector to perform deconvolution analysis on the expression matrix of immune cell
subtypes. It contains 547 biomarkers and 22 phenotypes
of human immune cells, covering plasma cells, B cells, T
cells, and myeloid cell subsets. Using the CIBERSORT
algorithm, this study analyzed data from POAG patients
and quantified the relative proportions of 22 infiltrating
immune cells. Furthermore, this study performed Pearson
correlation analysis on immune cells and gene expression.
Gene regulatory network analysis of key genes

The Cistrome DB is an up-to-date, scalable, and powerful tool to process large batches of ChIP-seq and DNaseseq  datasets, which map the genome-wide locations of
transcription factor-binding sites, histone posttranslational modifications and regions of chromatin accessible
to endonuclease activity [32]. Currently, the Cistrome DB
contains approximately 30,451 human and 26,013 mouse
samples. This study explored the regulatory relationships
between transcription factors and key genes based on the
Cistrome DB database with genome file setting as hg38
and the transcription initiation site 10 KB, which was visualized through Cytoscape.
Gene set enrichment analysis (GSEA)

GSEA uses a predefined gene set to rank the genes
according to the degree of differential expression in the

two types of samples and then tests whether the preset
gene set is enriched at the top ranking or bottom ranking. This study compared the differences in KEGG and
GO signaling pathways between participants in the high
expression group and the low expression group by GSEA
to explore the molecular mechanism of key genes in the
two groups, in which the number of replacements was set
to 1000 and the type of replacement was set to phenotype.

Drug targeting prediction

The connectivity map (CMAP), which was developed by the Broad Institute, is a resource created to
enable data-driven studies on drug mode-of-action
and drug repositioning; it is mainly used to reveal
the functional relationship among small molecular
compounds, genes and disease status. It contains the
microarray data of 1309 small molecule drugs before
and after treatment of 5 human disease cell lines. It
provides various treatment conditions, including different drugs, different concentrations, different treatment durations and so on.
Statistical analysis

All statistical analyses were performed in R language
(version 3.6). All statistical tests were bilateral (P < 0.05).
Acknowledgements
Not applicable.
Authors’ contributions
Study design: LGS, XJQ; Experimentation: LGS, GLL, Data analysis: LGS, WWD,
DZ, TC, TKY; figure and manuscript preparation: LGS, WWD; editing and review:
SLG, XJQ, CZ; supplemental files preparation: LGS. The author(s) read and
approved the final manuscript.
Funding

This study was funded by Chinese State Natural Science Foundation (Grant
No. 81670851), National Science and Technology Key R&D Program (Grant No.
2019YFC0118402), Capital’s Funds for Health Improvement and Research of China
(Grant No. CFH-2020–2-40911), and Peking University Medicine Fund of Fostering
Young Scholars’ Scientific & Technological Innovation (BMU2022PYB018).
Availability of data and materials
The selected datasets generated and/or analyzed during the current study are
available in the Gene Expression Omnibus (GEO) repository (https://​www.​ncbi.​
nlm.​nih.​gov/​geo/​query/​acc.​cgi?​acc=​GSE27​276; https://​www.​ncbi.​nlm.​nih.​
gov/​geo/​query/​acc.​cgi?​acc=​GSE23​78; https://​www.​ncbi.​nlm.​nih.​gov/​geo/​
query/​acc.​cgi?​acc=​GSE99​44). Based on these datasets, our data generated
or analyzed during this study are available from the corresponding author on
reasonable request.

Declarations
Ethics approval and consent to participate
Our data were downloaded from NCBI databases. No specific permits are
required for sample collection in this study.
Consent for publication
Not applicable.


Suo et al. BMC Genomic Data

(2022) 23:67

Competing interests
The authors have declared that no conflict of interest exists.
Author details
1

 Department of Ophthalmology, Peking University Third Hospital, 49
North Garden Road, Haidian District, Beijing 100191, PR China. 2 Beijing Key
Laboratory of Restoration of Damaged Ocular Nerve, Peking University Third
Hospital, 49 North Garden Road, Haidian District, Beijing 100191, PR China.
3
 Department of Ophthalmology, The Second Hospital of Shandong University,
Shandong, China. 4 Clinical Stem Cell Research Center, Peking University Third
Hospital, Beijing, China. 5 School of Basic Medicine Sciences, Peking University,
Beijing, China.
Received: 1 April 2022 Accepted: 29 June 2022

References
1. Kang JM, Tanna AP. Glaucoma. Med Clin North Am. 2021;105(3):493–510.
2. Kwon YH, et al. Primary open-angle glaucoma. N Engl J Med.
2009;360(11):1113–24.
3. Gharahkhani P, et al. Genome-wide meta-analysis identifies 127
open-angle glaucoma loci with consistent effect across ancestries. Nat
Commun. 2021;12(1):1258.
4. Sacca SC, et al. From DNA damage to functional changes of the trabecular meshwork in aging and glaucoma. Ageing Res Rev. 2016;29:26–41.
5. Liu Y, et al. Gene expression profile in human trabecular meshwork
from patients with primary open-angle glaucoma. Invest Ophthalmol
Vis Sci. 2013;54(9):6382–9.
6. Nucci C, et al. Neuroprotective agents in the management of glaucoma. Eye. 2018;32(5):938–45.
7. Lukas TJ, et al. Susceptibility to glaucoma: differential comparison of
the astrocyte transcriptome from glaucomatous African American and
Caucasian American donors. Genome Biol. 2008;9(7):R111.
8. Hernandez MR, et al. Differential gene expression in astrocytes from
human normal and glaucomatous optic nerve head analyzed by cDNA
microarray. Glia. 2002;38(1):45–64.
9. Kamat SS, Gregory MS, Pasquale LR. The Role of the Immune System

in Glaucoma: Bridging the Divide Between Immune Mechanisms in
Experimental Glaucoma and the Human Disease. Semin Ophthalmol.
2016;31(1–2):147–54.
10. Zhang T, Xie X, Lu F. Primary open-angle glaucoma: Neuroendocrine–
immune disorder? Med Hypotheses. 2014;83(4):514–5.
11. Huang P, et al. Serum Cytokine Alteration is Associated With Optic
Neuropathy in Human Primary Open Angle Glaucoma. J Glaucoma.
2010;19(5):324–30.
12. Wong M, et al. T-Helper1/T-Helper2 Cytokine Imbalance in the Iris of
Patients with Glaucoma. PLoS ONE. 2015;10(3):e0122184.
13. Raut SA, Sathe SR, Raut A. Bioinformatics: Trends in gene expression
analysis. In: 2010 International Conference on Bioinformatics and
Biomedical Technology.  IEEE. 2010. pp. 97–100.
14. Zheng P, et al. Editorial: Artificial Intelligence in Bioinformatics and Drug
Repurposing: Methods and Applications. Front Genet. 2022;13:870795.
15. Liesenborghs I, et al. Comprehensive bioinformatics analysis of trabecular
meshwork gene expression data to unravel the molecular pathogenesis
of primary open-angle glaucoma. Acta Ophthalmol. 2020;98(1):48–57.
16. Newman AM, et al. Robust enumeration of cell subsets from tissue
expression profiles. Nat Methods. 2015;12(5):453–7.
17. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes.
Nucleic Acids Res. 2000;28(1):27–30.
18. Zhou Z, et al. Adenosine A3 receptor activated in H2O2 oxidative stress
of primary open-angle glaucoma. Ann Transl Med. 2021;9(7):526.
19. Qiu H, Zhu B, Ni S. Identification of genes associated with primary
open-angle glaucoma by bioinformatics approach. Int Ophthalmol.
2018;38(1):19–28.
20. Feng J, Xu J. Identification of pathogenic genes and transcription factors in glaucoma. Mol Med Rep. 2019;20(1):216–24.
21. Wang J, et al. Integrated microarray analysis provided novel insights to
the pathogenesis of glaucoma. Mol Med Rep. 2017;16(6):8735–46.

Page 16 of 16

22. Zhu J, et al. Rottlerin acts as a therapeutic in primary open-angle
glaucoma by targeting the trabecular meshwork via activation of Rap1
signaling. Pharmacol Res. 2020;159:104780.
23. Liesenborghs I, et al. The Molecular Processes in the Trabecular Meshwork After Exposure to Corticosteroids and in Corticosteroid-Induced
Ocular Hypertension. Invest Ophthalmol Vis Sci. 2020;61(4):24–24.
24. Wang L, et al. Network Integration Analysis and Immune Infiltration
Analysis Reveal Potential Biomarkers for Primary Open-Angle Glaucoma. Front Cell Dev Biol. 2021;9:793638.
25. Hanukoglu I, Fuchs E. The cDNA sequence of a human epidermal
keratin: Divergence of sequence but conservation of structure among
intermediate filament proteins. Cell. 1982;31(1):243–52.
26. Coulombe PA, et al. Point mutations in human keratin 14 genes of epidermolysis bullosa simplex patients: genetic and functional analyses.
Cell. 1991;66(6):1301–11.
27. Schweizer J, et al. New consensus nomenclature for mammalian keratins. J Cell Biol. 2006;174(2):169–74.
28 Gautam P, et al. Multi-species single-cell transcriptomic analysis of
ocular compartment regulons. Nat Commun. 2021;12(1):5675.
29. Chen H, et al. Decreased hephaestin activity in the intestine of copperdeficient mice causes systemic iron deficiency. J Nutr. 2006;136(5):1236–41.
30. Moll R, et al. The catalog of human cytokeratins: patterns of expression
in normal epithelia, tumors and cultured cells. Cell. 1982;31(1):11–24.
31. Waseem A, et al. Isolation, sequence and expression of the gene encoding human keratin 13. Gene. 1998;215(2):269–79.
32. Zheng R, et al. Cistrome Data Browser: expanded datasets and new tools
for gene regulatory analysis. Nucleic Acids Res. 2019;47(D1):D729–35.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission
• thorough peer review by experienced researchers in your field
• rapid publication on acceptance
• support for research data, including large and complex data types
• gold Open Access which fosters wider collaboration and increased citations
• maximum visibility for your research: over 100M website views per year
At BMC, research is always in progress.
Learn more biomedcentral.com/submissions