Abstract
Myocardial infarction (MI) is an acute and persistent myocardial
ischemia caused by coronary artery disease. This study screened
potential genes related to MI. Three gene expression datasets related
to MI were downloaded from the Gene Expression Omnibus database.
Differentially expressed genes (DEGs) were screened using the MetaDE
package. Afterward, the modules and genes closely related to MI were
screened and a gene co-expression network was constructed. A support
vector machine (SVM) classification model was then constructed based on
the [42]GSE61145 dataset using the e1071 package in R. A total of 98
DEGs were identified in the MI samples. Next, three modules associated
with MI were screened and an SVM classification model involving seven
genes was constructed. Among them, BCL6, CEACAM8, and CUGBP2 showed
co-interactions in the gene co-expression network. Therefore, ACOX1,
BCL6, CEACAM8, and CUGBP2, in addition to GPX7, might be feature genes
related to MI.
Keywords: myocardial infarction, differentially expressed genes,
weighed gene co-expression network analysis, gene co-expression
network, support vector machine
Introduction
Myocardial infarction (MI), a major cause of death and disability
worldwide, is caused by myocardial cell death due to prolonged ischemia
([43]Thygesen et al., 2007). The most important risk factors of MI
include age, smoking, hypertension, diabetes, and total and
high-density lipoprotein cholesterol levels ([44]Bao et al., 2022;
[45]Bruyninckx et al., 2008;[46]O’Gara, 2013). Chest pain is the most
common clinical manifestation of acute MI, which is often described as
stress or compression ([47]Fauci, 2014). The pain often radiates to the
left arm as well as to the jaw, neck, right arm, back, and upper
abdomen ([48]Marcus et al., 2007). Approximately 15.9 million people
worldwide developed MI in 2015 ([49]Ji et al., 2015; [50]Cowan et al.,
2018). MI is an emerging public health concern globally. Previous
studies have suggested that ALOX5AP ([51]arachidonate 5-lipoxygenase
activating protein) confers a risk of MI; thus, ALOX5AP is the first
specific gene conferring a substantial population-attributable risk
(PAR) of MI ([52]Helgadottir et al., 2004). TGF-β1 (Transforming growth
factor-beta 1) is involved in the modulation of cell growth and
differentiation, and plays an important role in cardiovascular
physiopathology and the repair of vascular injury ([53]Nikol et al.,
1992; [54]Cambien et al., 1996). Meanwhile, the ALDH2 (aldehyde
dehydrogenase 2) Lys/Lys genotype is a risk factor for MI due to its
influence on high-density lipoprotein (HDL) cholesterol level
([55]Gardemann et al., 1998; [56]Takagi et al., 2002). PLA1
(Phospholipase A1 member A) hydrolyzes fatty acids at the sn-1 position
of phosphatidylserine and 1-acyl-2-lysophosphatidylserine and its
abnormal expression is associated with coronary artery disease (CAD)
and MI ([57]Ji et al., 2019). Furthermore, high-throughput screening
revealed that Nox2 as a potential miRNA target for function improvement
following MI ([58]Wang et al., 2012; [59]Smyth and Smyth, 2013;
[60]Yang et al., 2017; [61]Bao et al., 2022; [62]Kim et al., 2022).
However, the genes closely related to MI development have not been
fully identified.
The present study searched microarray datasets related to human MI.
Three gene expression datasets on MI were downloaded from the Gene
Expression Omnibus database and differentially expressed genes (DEGs)
were identified using MetaDE. The genes associated with MI were further
screened by identifying disease-associated modules. With this
information, we constructed a gene co-expression network. To classify
the MI samples, a support vector machine (SVM) classification model
trained on the [63]GSE61145 dataset was used. With this trained model,
we focused on mining related genes associated with MI.
Methods
Microarray data
The [64]GSE61145, [65]GSE60993, and [66]GSE34198 gene expression
datasets related to human MI, which were developed based on the
[67]GPL6106, [68]GPL6884, and [69]GPL6102 platforms, respectively, were
downloaded from the Gene Expression Omnibus (GEO,
[70]http://www.ncbi.nlm.nih.gov/geo/) database. The [71]GSE61145
dataset contained data on 14 blood samples from patients with MI and 10
samples from normal controls. The [72]GSE60993 dataset included data on
a total of 24 samples (7 and 17 blood samples from normal controls and
patients with MI, respectively). Finally, the [73]GSE34198 dataset
contained 97 samples (48 and 49 blood samples from normal controls and
patients with MI, respectively).
The raw data were downloaded and the probes were annotated into gene
symbols based on platform annotation information. Because a single gene
could correspond to several probes (multiple values), the average gene
expression values were calculated for each gene. Afterward, log2
conversion was performed to transform the gene expression data from a
skewed distribution to an approximately normal distribution. The data
were then normalized using the limma package (MetaDE)
([74]http://www.bioconductor.org/packages/2.9/bioc/html/limma.html) in
R language.
Differentially expressed gene and meta-analyses
DEGs were screened by using the MetaDE package ([75]Wang et al., 2012)
in R based on the [76]GSE61145 and [77]GSE60993 datasets. The raw data
were downloaded and the probes were annotated into gene symbols based
on platform annotation information. The average gene expression values
were calculated for each gene. Afterward, log2 conversion was performed
to transform the gene expression data from a skewed distribution to an
approximately normal distribution. The data were then normalized using
the limma package in R language. The heterogeneity of gene expression
data based on different platforms was analyzed using the MetaDE.ES
method ([78]Kim et al., 2022), with tau^2 = 0 and Qpval >0.05.
Differential expression analysis of genes with homogeneous expression
was then performed between the disease and control groups, with an FDR
(false discovery rate) of < 0.05 defined as the threshold value.
Screening modules and disease-related genes based on the meta-analysis
Weighted gene co-expression network analysis (WGCNA) (MetaDE) is a
typical system biology algorithm used to construct gene co-expression
networks based on high-throughput mRNA expression data. The genes and
modules related to MI in this study were analyzed for DEGs based on the
WGCNA algorithm ([79]Langfelder and Horvath, 2008). The correlation
coefficient between gene expression was calculated using the function
[MATH:
Smn=|cor(m,
n)|<
/mrow> :MATH]
. Then, the coefficient was then weighted by the exponential adjacency
function
[MATH:
amn=power(Smn,β) :MATH]
. According to the principle of scale-free networks, the weight
coefficient β was determined for the adjacency function. To measure the
dissimilarity between nodes, the correlation matrix (S[mn]) was
transformed into an adjacency matrix (a[mn]). A hierarchical clustering
tree was then constructed based on the dissimilarity coefficients
between genes, with different branches of the clustering tree
representing different gene modules. Finally, t-tests were used to
analyze the correlations between network modules and disease states.
Construction of the gene co-expression network and enrichment analysis
The co-expressed modules that were closely associated with disease
state were analyzed and the module genes were collected to construct
the co-expression network. Genes related to disease were subjected to
gene ontology (GO) enrichment analysis using the clusterProfiler
package ([80]Yu et al., 2012) in R based on the hypergeometric
distribution algorithm. P < 0.05 was defined as the threshold value.
The formula for the hypergeometric distribution algorithm is shown in
[81]Eq. 1.
[MATH: p=1−∑i=0H−1(Mi)(N−MK−H)(NK), :MATH]
(1)
where N represents the number of genes with GO functional annotations;
K represents the number of DEGs among N genes, and M represents the
number of genes that are annotated with a particular GO function.
Support vector machine (SVM) classification modeling
To classify the samples, an optimal classification hyperplane must be
selected from numerous options to maximize the distance δ between the
sample set and the classification hyperplane. When ε = |wx [i ]+ b| =
1, the distance between the two types of sample points is 2 ((|wx [i ]+
b|)/‖w‖) =(2/‖w‖). The goal is to construct an optimal classification
hyperplane under the constraint of [82]Eq. 2 to maximize (2/‖w‖) and
minimize (‖w‖2/2).
[MATH: {wxi
+b≥1,yi=1wxi+b≤
−1,yi=−1i=1,2⋯,
l. :MATH]
(2)
Most classification issues can be treated as nonlinear separable
problems, and quadratic programming problems can be modified as follows
by introducing the slack variable ξi in [83]Eq. 3.
[MATH: {min1<
/mrow>2‖w‖2+C∑ξi
,ξi≥0cons
traint condition:
yi((wxi+
b))≥1
−ξi
,i=1,2⋯,l
:MATH]
(3)
where ξ [i ]is the slack variable and C is the penalty coefficient.
With [84]GSE61145 as the training dataset and all genes of interest as
classification factors, the SVM model was established using the e1071
package (MetaDE) in Rto distinguish the disease and control samples.
The classification factors were added individually until all of them
had been added to the SVM classifier. The classification accuracy of
the SVM classifier was then calculated and the genes that affected
classification accuracy were removed. The SVM model was then validated
in the [85]GSE60993 and [86]GSE34198 gene expression datasets.
Results
DEG screening
A total of 1,231 DEGs were identified by the MetaDE package
([87]Langfelder and Horvath, 2008; [88]Wang et al., 2012; [89]Yu et
al., 2012; [90]Meyer, 2013). The top 10 DEGs are listed in [91]Table 1,
including GZMK (granzyme K), HLA-DQA (histocompatibility complex, class
II, DQ alpha), and EOMES (eomesodermin). First, the heterogeneity of
gene expression data based on different platforms was analyzed using
the MetaDE.ES method, with tau2 = 0 and Qpval >0.05. Then, the
differential expression analysis of genes with homogeneous expression
was conducted between the disease and control groups, with an FDR
(false discovery rate) of <0.05 defined as the threshold value. A total
of 1,231 DEGs were identified. The top 10 DEGs with the smallest
p-values in the gene difference analysis between the disease and
control groups were selected; that is, the genes with the largest
difference between disease and control groups. The present study
analyzed the co-expressed modules that were closely associated with the
disease state and identified the module genes to construct a
co-expression network. Genes related to disease were subjected to gene
ontology (GO) enrichment analysis. KEGG pathway enrichment analysis was
not performed.
TABLE 1.
List of top 10 significant differentially expressed genes from
[92]GSE61145 and [93]GSE60993.
Symbol p FDR Q Qp tau2 logFC
GZMK 1.00E-20 2.47E-17 0.661264 0.416114 0 −4.38757
HLA-DQA1 1.22E-06 0.00028 0.067206 0.795449 0 −3.34521
EOMES 1.22E-06 0.00028 0.02484 0.874767 0 −3.27466
GZMA 8.51E-06 0.000625 0.599134 0.438909 0 −3.22898
GZMH 4.62E-05 0.001564 0.05081 0.82166 0 −2.76322
GZMM 5.67E-06 0.000478 0.164991 0.684602 0 −2.74506
KLRB1 4.05E-07 0.000133 0.336467 0.561876 0 −2.74359
NKG7 1.09E-05 0.00069 0.001251 0.971784 0 −2.718
IL2RB 1.38E-05 0.000778 0.644833 0.421966 0 −2.62686
[94]Open in a new tab
^a
FDR, false discovery rate; FC, fold-change.
Modules and genes closely related to disease
To satisfy the precondition of scale-free network distribution, we
selected a power of 18 as the adjacency parameter. The results of the
consistency analysis showed a high correlation between the [95]GSE61145
and [96]GSE60993 datasets (correlation coefficient = 0.86, p-value <
1e-200). Additionally, [97]GSE61145 was used as a training set to
identify disease-associated modules ([98]Figure 1A). Module
partitioning for the [99]GSE60993 dataset ([100]Figure 1B) showed high
consistency with the [101]GSE61145 dataset. We then calculated the
correlation coefficient between module and disease state (normal and MI
samples) for the [102]GSE61145 ([103]Figure 2A) and [104]GSE60993
([105]Figure 2B) datasets, respectively ([106]Table 1). According to
the correlation coefficients, the top three modules (black, pink, and
red) were identified.
FIGURE 1.
[107]FIGURE 1
[108]Open in a new tab
Tree diagrams for identifying the disease-associated modules based on
the [109]GSE61145 (A) and [110]GSE60993 (B) datasets. The abscissa
represents modules in different colors. The ordinate represents the
height of the system clustering tree based on the expression value.
FIGURE 2.
[111]FIGURE 2
[112]Open in a new tab
The disease-associated modules identified from the [113]GSE61145 (A)
and [114]GSE60993 (B) datasets. The abscissa represents modules in
different colors. The ordinate represents the overall correlation
coefficient between the genes in each module and the disease state.
Gene co-expression network construction and enrichment analysis
The correlation coefficients between genes in the top three modules and
disease state were calculated, which revealed 98 genes with correlation
coefficients >0.5. These included 30 genes (11 up-regulated and 19
down-regulated) in the black module, 19 genes (9 up-regulated and 10
down-regulated) in the pink module, and 49 genes (22 up-regulated and
27 down-regulated) in the red module. The gene co-expressed networks of
the 98 genes were then constructed ([115]Figure 3). GO analysis showed
the enrichment of 10 GO terms among the genes in the black module
([116]Table 2) and 15 GO terms among the genes in the red module
([117]Table 2). The GO terms enriched in the black module included
negative regulation of cell proliferation (p-value = 0.009704),
regulation of cell proliferation (p-value = 0.014724), and positive
regulation of macromolecule metabolic process (p-value = 0.019608). The
GO terms closely related to the genes in the red module mainly included
positive regulation of I-kappaB kinase/NF-kappaB cascade (p-value =
0.024598), regulation of I-kappaB kinase/NF-kappaB cascade (p-value =
0.029497), and positive regulation of signal transduction (p-value =
0.037345). No GO terms were significantly enriched among the genes in
the pink module.
FIGURE 3.
[118]FIGURE 3
[119]Open in a new tab
Gene co-expression network of the black, pink, and red modules. The
inverted and positive triangles represent up- and down-regulated genes
in the disease group, respectively. The node colors reflects the colors
of the disease modules.
TABLE 2.
Gene ontology functions enriched in the black (A) and red (B) modules.
Term Count p-value Genes
(A)
GO:0008285∼negative regulation of cell proliferation 4 0.009704 BCL11B,
RXRA, PEMT, BCL6
GO:0042127∼regulation of cell proliferation 5 0.014724 BCL11B, RXRA,
PEMT, BCL6, PURA
GO:0010604∼positive regulation of macromolecule metabolic process 5
0.019608 SLC11A1, PSMA5, BCL11B, RXRA, PEMT
GO:0046649∼lymphocyte activation 3 0.025327 SLC11A1, BCL11B, BCL6
GO:0019637∼organophosphate metabolic process 3 0.025564 GPD1L, PEMT,
ALG9
GO:0015807∼L-amino acid transport 2 0.02853 SLC36A1, SLC11A1
GO:0045321∼leukocyte activation 3 0.036326 SLC11A1, BCL11B, BCL6
GO:0000060∼protein import into nucleus, translocation 2 0.041902
SLC11A1, BCL6
GO:0001818∼negative regulation of cytokine production 2 0.046722
SLC11A1, BCL6
GO:0001775∼cell activation 3 0.049469 SLC11A1, BCL11B, BCL6
(B)
GO:0043123∼positive regulation of I-kappaB kinase/NF-kappaB cascade 3
0.024598 CFLAR, TNFRSF10B, RHOC
GO:0043122∼regulation of I-kappaB kinase/NF-kappaB cascade 3 0.029497
CFLAR, TNFRSF10B, RHOC
GO:0009967∼positive regulation of signal transduction 4 0.037345 CFLAR,
TNFRSF10B, ZAP70, RHOC
GO:0010647∼positive regulation of cell communication 4 0.04897 CFLAR,
TNFRSF10B, ZAP70, RHOC
GO:0006915∼apoptosis 5 0.062711 CFLAR, TNFRSF10B, RAF1, MTP18, SOD1
GO:0006575∼cellular amino acid derivative metabolic process 3 0.065021
SLC22A4, ICMT, SOD1
GO:0012501∼programmed cell death 5 0.065518 CFLAR, TNFRSF10B, RAF1,
MTP18, SOD1
GO:0010740∼positive regulation of protein kinase cascade 3 0.065708
CFLAR, TNFRSF10B, RHOC
GO:0006879∼cellular iron ion homeostasis 2 0.075126 HP, SOD1
GO:0055072∼iron ion homeostasis 2 0.086718 HP, SOD1
GO:0007242∼intracellular signaling cascade 7 0.090132 PDZD8, TNFRSF10B,
ZAP70, RAF1, RHOC, SOD1, RAB27A
GO:0043065∼positive regulation of apoptosis 4 0.092511 CFLAR,
TNFRSF10B, SOD1, RAB27A
GO:0043068∼positive regulation of programmed cell death 4 0.093994
CFLAR, TNFRSF10B, SOD1, RAB27A
GO:0010942∼positive regulation of cell death 4 0.094988 CFLAR,
TNFRSF10B, SOD1, RAB27A
GO:0007010∼cytoskeleton organization 4 0.095487 SVIL, SSH2, RAF1, SOD1
[120]Open in a new tab
Construction and evaluation of the SVM classification model
Based on the SVM classification model, we removed genes that could not
distinguish between the disease and control samples. Finally, seven
genes were obtained: ACOX1 (Acyl CoA oxidase 1), ADCK2 (aarF domain
containing kinase 2), AFF3 (AF4/FMR2 family member 3), BCL6 (B-cell
lymphoma 6), CEACAM8 (Carcinoembryonic antigen-related cell adhesion
molecule 8), CUGBP2 (CUG triplet repeat-binding protein 2) and GPX7
(glutathione peroxidase 7). The SVM classification model of these seven
genes could distinguish all samples in the [121]GSE61145 dataset. The
scatterplot of the [122]GSE61145 dataset is shown in [123]Figure 4A.
The [124]GSE60993 and [125]GSE34198 datasets were then used as
validation datasets to confirm the SVM classification model. As shown
in [126]Figure 4B, the SVM classification model correctly distinguished
23 (17 disease and 6 normal samples) of 24 samples in the [127]GSE60993
dataset. Additionally, the scatterplot of the [128]GSE34198 dataset
indicated that the SVM classification model correctly distinguished 90
(48 disease and 42 normal samples) of the 97 samples ([129]Figure 4C).
The efficiency receiver operating characteristic (ROC) curves of the
SVM classification model are shown in [130]Figure 5 and the efficiency
parameters of each dataset are listed in [131]Table 3.
FIGURE 4.
[132]FIGURE 4
[133]Open in a new tab
Scatterplots of the [134]GSE61145 (A), [135]GSE60993 (B), and
[136]GSE34198 (C) datasets. The purple and red dots represent the
normal and disease samples, respectively. The X and Y axes represent
the position vector coordinates of the samples.
FIGURE 5.
FIGURE 5
[137]Open in a new tab
Receiver operating characteristic (ROC) curves showing classifier
efficiency. The black, red, and green curves show the ROC curves of the
[138]GSE61145, [139]GSE60993, and [140]GSE34198 datasets, respectively.
TABLE 3.
Parameters for classifier performance.
Datasets Num.Samples Correct sample Correct rate Sensitivity
Specificity PPV NPV Auroc
[141]GSE61145 24 24 1.000 1.000 1.000 1.000 1.000 1.000
[142]GSE60993 24 23 0.958 1.000 0.857 0.944 1.000 0.983
[143]GSE34198 97 90 0.928 0.979 0.875 0.889 0.977 0.956
[144]Open in a new tab
PPV, positive predictive value; NPV, net present value; AUROC, area
under the receiver operating characteristic.
Discussion
MI is a major cause of death and disability worldwide and has imposed
burdens and impacted the health of the population ([145]ErikssonP.,
2014) While studies have focused on the mechanism and management of MI
at the molecular level ([146]Hak et al., 2000; [147]Erikson et al.,
2017; [148]Wongsurawat, 2018; [149]Yang et al., 2022), effective
therapy is lacking. The present study screened 1,231 DEGs based on
three microarray datasets. Based on WGCNA, the top three modules
related to disease (black, pink, and red) were screened. Afterward, a
total of 98 DEGs were screened from the top three modules to construct
the gene co-expression network. The SVM classification model was also
constructed and identified seven genes (including ACOX1, BCL6, CEACAM8,
CUGBP2, and GPX7) that were closely associated with MI.
ACOX1 is the first enzyme in peroxisomal fatty acid β-oxidation. It is
rate-limiting and plays a key role in fatty acid metabolism and fat
deposition ([150]Foraker et al., 2013). Both lipid abnormalities and
chronic inflammation have crucial involvement in atherosclerosis
initiation and progression ([151]Bhagavan et al., 2003). Lutein plays a
regulator role in gene expression and is involved in oxidative stress
and the lipid metabolism of ACOX1, thereby mitigating atherosclerosis
progression ([152]Bruyninckx et al., 2008). In addition, BCL6 is a
transcriptional repressor required for mature B-cell germinal center
(GC) formation and is also implicated in lymphomagenesis ([153]Jiao et
al., 2011; [154]Vik et al., 2015). Increasing Bcl6 expression reduces
inflammatory responses and limits atherosclerosis ([155]Han et al.,
2015). Meanwhile, CEACAM8 is a glycosylphosphatidylinositol-anchored
membrane glycoprotein with a molecular weight of around 95 kDa
([156]Basso et al., 2010). CEACAM8 is also known as Cluster of
Differentiation 66b (CD66b) and is expressed by neutrophils ([157]Lasa
et al., 2008; [158]Kulbacki et al., 2010; [159]Singer, 2013; [160]Wei
et al., 2015). Leucocyte activation is a crucial step in atherogenesis
([161]Chudasama et al., 2011). The expression of leucocyte integrins,
such as neutrophil and neutrophil CD66b, has been linked to
atherosclerosis ([162]Alipour et al., 2013). Furthermore, coronary
artery disease (CAD) reflects generalized inflammation ([163]Oostrom et
al., 2004). Additionally, CUG triplet repeat-binding protein 2 (CUGBP2)
plays a critical role in the apoptosis of breast cancer cells in
response to genotoxic injury ([164]Mukhopadhyay et al., 2004). The
over-expression of miR-144 can decrease cardiomyocyte cell death by
targeting CUGBP2 ([165]Alipour et al., 2013). miR-451 is also largely
responsible for ischemic preconditioning-mediated cardioprotection,
which also showed protective effects against simulated
ischemia/reperfusion-induced cardiomyocyte death by CUGBP2 regulation
([166]Weiss et al., 2012; [167]Chen et al., 2014; [168]Feng et al.,
2016). Subsequently, GPX7 is an endoplasmic reticulum (ER)-mitochondria
protein that plays important and emerging functional roles in T-cell
development ([169]Higashi et al., 2013). Numerous clinical studies have
found that hyperhomocysteinemia (HHcy) is an independent risk factor
for cardiovascular diseases in humans ([170]Chen et al., 2016). HHcy
accelerates atherosclerosis by affecting the immuno-inflammatory
response and repressing regulatory T-cell functions ([171]Feng et al.,
2016). Furthermore, the results of the gene co-expression network
analysis in this study showed the co-expression of BCL6, CEACAM8, and
CUGBP2. ADCK2 and AFF3 were also associated with MI in this study.
However, evidence regarding their roles in MI is scarce. Thus, ACOX1,
BCL6, CEACAM8, CUGBP2 and GPX7 may play key roles in MI pathogenesis.
Conclusion
Myocardial infarction is one of the most dangerous diseases worldwide.
This study screened for genes associated with such diseases. We
obtained gene expression datasets ([172]GSE61145, [173]GSE60993, and
[174]GSE34198) related to human MI. We searched microarray datasets
involving human MI and then investigated the DEGs between MI and normal
samples. The genes associated with MI were further screened by
identifying the disease-associated modules to construct a gene
co-expression network. ACOX1, BCL6, CEACAM8, CUGBP2, and GPX7 might be
key genes implicated in MI development. The MI-associated genes may
provide targets for novel therapy for MI. As our findings were
partially drawn by prediction, they require additional validation.
However, this study has several limitations that should be addressed in
future work. The SVM algorithm can be treated as a typical
classification model in the field of bioinformatics and computational
biology. Therefore, several classification algorithms, including random
forest, neural network, and some deep learning algorithms, can be used
to correct this issue. This study used the [175]GSE61145 dataset to
train the classification mode. Considering the generality of the
classification model, more datasets should be trained. Future work
should also utilize cross-validation methods.
Data availability statement
Publicly available datasets were analyzed in this study. The names of
the repository/repositories and accession number(s) can be found in the
article/supplementary material.
Author contributions
LY and XP designed the experiments in this study. WL, the
co-corresponding author, constructed the model and edited the
manuscript. YZ, GY, and DZ edited the figures. LW, CZ, and TL edited
the tables.
Conflict of interest
The authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a
potential conflict of interest.
Publisher’s note
All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors, and the
reviewers. Any product that may be evaluated in this article, or claim
that may be made by its manufacturer, is not guaranteed or endorsed by
the publisher.
References