Abstract

   Lung cancer has a high incidence rate and lung squamous cell carcinoma
   accounts for about 30% of lung cancer. Due to the lack of accurate
   diagnosis for early stage lung squamous cell carcinoma and the
   limitations of treatment methods, the clinical treatment effect of lung
   squamous cell carcinoma is not satisfactory.we used single-cell
   sequencing analysis, bioinformatics analysis, and Mendelian
   randomization analysisto look for risk factors that can predict the
   onset of lung squamous cell carcinoma, the prognosis of lung squamous
   cell carcinoma patients, and the “star” molecule—Fatty Acid Binding
   Protein 4 (FABP4) as a mediator of macrophage infiltration in the lung
   squamous cell carcinoma tumor microenvironment. Overall, this “star”
   molecule (FABP4) may play an important role in the future prediction
   and treatment of lung squamous cell carcinoma.We first clarified the
   role of FABP4 in the occurrence and prognosis of LUSC through
   single-cell sequencing analysis, bioinformatics analysis, and Mendelian
   randomization analysis. This provides important evidence for the
   prevention and treatment of LUSC in the future.

   Keywords: Single-cell RNA sequencing, Bioinformatics analysis,
   Mendelian random analysis, Fatty acid binding protein 4, Lung squamous
   cell carcinoma

Introduction

   Mendelian random analysis is an analytical technique that can be used
   to investigate the relationship between exposure factors and outcome
   indicators while avoiding confounding, reverse causality, and a variety
   of biases [[34]1]. Mendelian randomization is an application of
   instrumental variable analysis aimed at testing causal hypotheses in
   non experimental data. In MR analysis, genetic variations (usually
   single nucleotide polymorphisms) are used as instrumental variables to
   infer risk factors. The MR principle refers to Mendel's second law,
   which states that when DNA is passed from parents to offspring during
   gamete formation, genetic alleles are independently separated. This is
   similar to randomized allocation therapy in RCTs, aimed at generating
   groups with similar clinical characteristics to reduce the risk of
   confounding. Single-cell RNA sequencing is a popular method for
   investigating cellular heterogeneity and functional variety in
   anatomical tissues. [[35]2–[36]6]. Lung cancer is one of the leading
   causes of cancer mortality in both China and the United States, and its
   prevalence has gradually increased [[37]7, [38]8]. Squamous cell
   carcinoma of the lung is currently the second most common subtype of
   all lung cancers in terms of incidence rate [[39]9]. The most common
   cause of lung squamous cell carcinoma is smoking [[40]10]. Currently,
   the most common treatments for lung squamous cell carcinoma are
   surgery, radiation therapy, chemotherapy, immunotherapy, and others
   [[41]11]. Because lung squamous cell carcinoma develops rapidly, there
   is an urgent need for early identification and cutting-edge treatment
   options.

   One of the most prevalent members of the intracellular Fatty
   acid-binding proteins (FABPs)family, fatty acid binding protein 4
   (FABP4) may bind hydrophobic ligands to regulate lipid trafficking and
   is connected to advancement in many malignant cancers. For instance,
   lipid intake and tumor growth are inextricably linked. Exogenous FABP4
   has been shown to improve fatty acid uptake and accelerate tumor cell
   development in prostate cancer and breast cancer via activating the PI3
   K/AKT and MAPK pathways, according to research [[42]12, [43]13]. An
   essential factor in preventing tumor cell metastasis is preventing
   fatty acid absorption. In ovarian cancer, MIR-409-3P directly controls
   FABP4 and has an impact on the disease’s spread [[44]14]. When tumor
   cells absorb too many fatty acids, they can create lipid droplets in
   the cytoplasm that make them more resistant to chemotherapy and
   radiation [[45]15, [46]16]. FABP4 prevents the decrease of
   intracellular lipid droplets, which increases the sensitivity of tumor
   cells to carboplatin, according to investigations that have been
   conducted on animals [[47]17]. FABP4 also unquestionably contributes to
   the tumor microenvironment. Typically, immune cells, cancer cells,
   endothelial cells, and adipocytes make up the tumor microenvironment.
   Adipocytes are directly transferred from blood lipids to ovarian cancer
   cells as a result of the link between adipocytes and ovarian cancer
   cells in ovarian cancer, which encourages the growth of tumor cells
   both in vitro and in vivo. At the fat tumor cell interface, ovarian
   cancer cells were found to express FABP4 [[48]18]. Adipocytes are
   directly transferred from blood lipids to ovarian cancer cells as a
   result of the link between adipocytes and ovarian cancer cells in
   ovarian cancer, which encourages the growth of tumor cells both in
   vitro and in vivo. At the fat tumor cell interface, ovarian cancer
   cells were found to express FABP4 [[49]19]. By sending signals via the
   control of NFMOR-6/STT3, FABP4 expression in breast cancer contributes
   to the promotion of breast cancer development [[50]20]. Additionally,
   FABP4 inhibits NF by ubiquitinating ATPB in tumor-associated
   macrophages, which promotes the growth of neuroblastoma and B-IL1
   pathways [[51]21].

   In this study, we used FABP4 as the exposure factor and lung squamous
   cell carcinoma as the outcome measure. Through Mendelian randomization
   analysis, we found that FABP4 is a risk factor for the occurrence of
   lung squamous cell carcinoma. In addition, we used single cell RNA
   sequencing and bioinformatics analysis to evaluate the role of FABP4 in
   the incidence rate of lung squamous cell carcinoma and its impact on
   the immune microenvironment. These findings will help us understand the
   functional significance of FABP4 in lung squamous cell carcinoma and
   create more precise approaches for treating the disease.

Materials and methods

Mendelian random analysis

   In order to explore the hypothesized causal relationship between
   exposure factors and outcomes, we used the R software packages
   “TwoSampleMR” and “GSMR” to investigate the hypothesized causal
   relationship between exposure factors and outcomes, in order to obtain
   suggestive associations (P < 0.05), and conducted Mendelian random
   analysis on the causal relationship between exposure and outcomes using
   the main five MR methods (MR Egger, weighted median, inverse variance
   weighting, weighted mode, simple mode). We chose SNPs with genome-wide
   significance (P < 5 × 10 − 6) as instrumental variables. The
   F-statistic data for each tool is estimated by F = β 2/SE2, with a
   screening criterion of F-test value > 10. Heterogeneity analysis was
   evaluated through Inverse variance weighted and MR Egger tests, and
   Pvalue < 0.05 indicates the presence of heterogeneity in the study. The
   MR Egger intercept test can detect the pleiotropy of the data, and
   Pvalue < 0.05 indicates the presence of pleiotropy in the data.
   Finally, use the leave one method for sensitivity analysis (to
   determine the impact of each SNP on the Mendelian randomization
   analysis results, and if there are outliers, they need to be removed
   and reanalyzed).

Source of transcriptome and single-cell RNA sequencing data

   We downloaded RNA sequencing data and clinical information of lung
   squamous cell carcinoma from the TCGA database, and also downloaded RNA
   sequencing data and clinical information of lung squamous cell
   carcinoma from the GEO database ([52]GSE73403).

   The scRNA-seq data of LUSC tumor ([53]GSM3304009)and normal
   ([54]GSM3304010) sample of [55]GSE117570 were downloaded from the GEO
   database.

   The Cancer Genome Atlas (TCGA)database:
   ([56]https://www.cancer.gov/ccg/).

   Gene Expression Omnibus (GEO) database:
   ([57]http://www.ncbi.nlm.nih.gov/geo/).

Preprocessing of single-cell RNA sequencing datas

   We downloaded RNA sequencing data and clinical information of lung
   squamous cell carcinoma from the TCGA database, and also downloaded RNA
   sequencing data and clinical information of lung squamous cell
   carcinoma from the GEO database ([58]GSE73403). The single-cell
   sequencing data of lung squamous cell carcinoma is sourced from the
   single-cell sequencing data of [59]GSE117570 in the GEO database.
   Simultaneously use R software"Seurat package"[[60]22] to convert scRNA
   seq data into Seurat objects. The single-cell inclusion features we
   selected are: minimum gene count of 200, maximum gene count of 4500,
   minimum RNA count of 1000, maximum RNA count of 20,000, and maximum
   mitochondrial RNA percentage of 10%. Finally, we use
   the"NormalizedData"package for data normalization.

Clustering and annotation of differentially expressed genes in single-cell
RNA sequencing datas

   We use T-distributed stochastic neighbor embedding (t-SNE) [[61]23] for
   unsupervised clustering and unbiased visualization on the
   two-dimensional graph plane of cell subpopulations, and annotate cell
   subpopulations using the"SingleR"[[62]24] software package.
   The"FindAllMarkers"function is used to compare differentially expressed
   genes between different clusters, and the screening criteria for
   differentially expressed genes are: |log2 (fold change)|> 0.5 and
   adjusted P-value < 0.05.

Cell communication and cell trajectory analysis

   We use cellphoneDB software to analyze the communication interactions
   of different cell subpopulations. To infer the developmental trajectory
   of each cell type, we conducted a pseudo temporal analysis using R
   package Monocle 3, and the results were clustered and visualized using
   the UMAP method.

   cellphoneDB: ([63]https://www.cellphonedb.org/).

   Monocle 3: ([64]https://cole-trapnell-lab.github.io/monocle3).

Differential gene screening and pathway enrichment analysis

   Using the limma package for differential analysis, the screening
   conditions for differentially expressed genes are: |log2 (fold
   change)|> 1.5 and adjusted P-value < 0.05. Perform pathway enrichment
   analysis on differentially expressed genes using KEGG and GO.

   KEGG: ([65]https://www.kegg.jp/).

   GO: ([66]https://geneontology.org/).

Survival analysis

   We conducted survival analysis using the"survey"function of the R
   survival package and used Cox regression analysis to determine the
   significance of the difference groups.

Drug sensitivity analysis

   We use"oncoPredict"to predict the sensitivity of FABP4 gene high
   expression and low expression groups to drugs.

Pan cancer analysis

   We analyzed the differential expression of FABP4 gene in cancer tissue
   and normal tissue in the Timer 2.0 database. We downloaded a
   standardized pan cancer dataset from the UCSC database and analyzed the
   differential expression of FABP4 in different clinical stages of other
   tumors, as well as its impact on the survival of other cancers.

   TIMER 2.0 database: ([67]http://timer.cistrome.org/).

   UCSC database: ([68]https://xenabrowser.net/).

Statistical analysis

   Our data in this study (represented by means ± SDs) were statistically
   analyzed using GraphPad Prism 9 and tested for statistical significance
   using one-way ANOVA. A p-value < 0.05 is considered statistically
   significant. The significant statistical differences are expressed as
   *P < 0.05, **P < 0.01, ***P < 0.001, and **** P < 0.0001. Non
   significant statistical differences are expressed as"ns".

Results

Single cell subpopulation clustering and annotation

   We obtained single-cell sequencing data from [69]GSE117570 for lung
   squamous cell carcinoma and normal lung tissue. After quality control,
   standardization, and dimensionality reduction using the PCA method, all
   cells were classified into 8 subgroups using tSNE clustering and known
   marker genes, and annotated as 6 cell types: epithelial cells,
   endothelial cells, T cells, B cells, monocytes, and macrophages
   (Fig. [70]1A, B). Finally, we analyzed the characteristic genes and
   differential genes of six cell subtypes (Fig. [71]1C, D).

Fig. 1.

   [72]Fig. 1
   [73]Open in a new tab

   Cell clustering and annotation (A, B). The main expressed genes and
   differentially expressed genes (C, D) in six types of cells

Single cell subpopulation maker gene annotation

   We identified the maker genes of epithelial cells, endothelial cells, T
   cells, B cells, monocytes, and macrophages, and analyzed the expression
   of these genes in these six subtypes of cells. We found that KRT6 A,
   CAPS, BPIFA1, and SFTPC are maker genes of epithelial cells and are
   highly expressed in epithelial cells. G0S2 is the maker gene of
   monocytes and is highly expressed in monocytes. FABP4 is the maker gene
   of macrophages and is highly expressed in macrophages. CD69 is the
   maker gene of T cells and is highly expressed in T cells. MGP is the
   maker gene of endothelial cells and is highly expressed in endothelial
   cells, IGLC2 is the maker gene of B cells and is highly expressed in B
   cells (Fig. [74]2A–R).

Fig. 2.

   [75]Fig. 2
   [76]Open in a new tab

   Maker genes of six types of cells and their expression in each type of
   cell (A–R)

Single factor and multiple regression analysis of the Maker gene

   We integrated the maker gene sets KRT6 A, G0S2, CAPS, FABP4, BPIFA1,
   CD69, MGP, SFTPC, and IGLC of lung squamous cell carcinoma, and
   conducted univariate regression analysis based on the survival time and
   status of lung squamous cell carcinoma patients. Next, we conducted a
   multivariate regression analysis on the gene sets with P < 0.05 in the
   univariate regression. Combining univariate and multivariate regression
   analysis, we found that FABP4 is an independent factor affecting the
   prognosis of lung squamous cell carcinoma patients (Fig. [77]3A, B). We
   found that in lung squamous cell carcinoma patients in the [78]GSE73403
   and TCGA databases, patients with high expression of FABP4 A had
   shorter survival compared to those with low expression of FABP4
   (Fig. [79]3C, D).

Fig. 3.

   [80]Fig. 3
   [81]Open in a new tab

   Univariate and multivariate regression analysis of maker genes (A, B).
   The effect of FABP4 expression on the survival of lung squamous cell
   carcinoma (C, D)

Pathway enrichment analysis and drug sensitivity analysis

   We used the R software package limma (version 3.40.6) for differential
   analysis to obtain the differential genes between the high expression
   group of FABP4 and the low expression group of FABP4 (Fig. [82]4A).
   Here we present the top 50 differentially expressed genes through a
   heatmap (Fig. [83]4B). We conducted GO and KEGG pathway enrichment
   analysis on these differentially expressed genes and found that they
   are enriched in pathways such as cell mitosis, DNA replication,
   cytokines, growth factors, immunoglobulins, cell cycle, and ECM
   receptor interactions (Fig. [84]4C, D). Finally, we used oncoPredict to
   predict and analyze the sensitivity of the FABP4 gene high expression
   and low expression groups to drugs. We found that patients with low
   expression of FABP4 have high sensitivity to Selumetinib and Compound,
   but low sensitivity to Vincristine and Parbendazole (Fig. [85]4E–J).

Fig. 4.

   [86]Fig. 4
   [87]Open in a new tab

   Volcano plot of differentially expressed genes between high and low
   FABP4 groups of patients (A). The top 50 differentially expressed genes
   (B) in patients with high and low FABP4 levels. GO and KEGG pathway
   analysis of differentially expressed genes (C, D). Predictive Analysis
   of Drug Sensitivity in Patients with High and Low FABP4 Groups (E-J)

Immune cell infiltration analysis

   We used Cibersort to analyze the infiltration of 22 immune cells in
   lung squamous cell carcinoma and found that M2 macrophage infiltration
   was weaker in patients with high FABP4 expression than in patients with
   low FABP4 expression in TCGA and [88]GSE73403 lung squamous cell
   carcinoma (Fig. [89]5A, B).

Fig. 5.

   [90]Fig. 5
   [91]Open in a new tab

   Immunocyte Infiltration Analysis (A, B)

Cell trajectory and cell communication analysis

   To further reveal the cell differentiation process in the immune
   microenvironment, we calculated the cell differentiation process in
   lung squamous cell carcinoma patients. We found that monocytes have
   evolved to some extent towards macrophages and T cells. However, the
   evolutionary pathways of epithelial cells, endothelial cells, B cells,
   monocytes, T cells, and macrophages are far apart, and it is possible
   that there is no evolutionary relationship between these cells
   (Fig. [92]6A, B).We then analyzed the dynamic expression of FABP4 in
   cell trajectories and found that it is still highly expressed in
   macrophages during cell differentiation trajectories. As evolution
   began, FABP4 was highly expressed in macrophages and showed an upward
   trend. Then, as evolution progressed, FABP4 expression gradually
   decreased, and finally decreased in epithelial cells, monocytes, T
   cells, and B cells (Fig. [93]6C, D). Finally, we analyzed the
   communication connections between these cells and found that there is a
   close relationship between macrophages and endothelial cells, and there
   is a chemotaxis phenomenon between macrophages and monocytes
   (Fig. [94]6E). (Fig. [95]6F) shows the expression of genes regulating
   co inhibition, co stimulation, and chemokines among epithelial cells,
   endothelial cells, B cells, T cells, monocytes, and macrophages.

Fig. 6.

   [96]Fig. 6
   [97]Open in a new tab

   Analysis of cell differentiation trajectory (A, B). The expression
   levels and dynamic changes of FABP4 in six types of cells (C, D). Cell
   communication analysis between 6 types of cells (E). The expression of
   co inhibitory, co stimulatory, and chemokine genes among six types of
   cells (F)

Pan cancer analysis

   We analyzed the differential expression of FABP4 gene in cancer tissues
   and normal tissues in the Timer 2.0 database, and found that the
   expression level of FABP4 was lower in most tumors than in normal
   tissues (Fig. [98]7A). We downloaded a standardized pan cancer dataset
   from the UCSC database and used R software (version 3.6.4) to calculate
   the expression differences of genes in different clinical stage samples
   for each tumor. We observed significant differences in 10 types of
   tumors, including CESC, BRCA, STES, KIPAN, STAD, KIRC, THYM, THCA, OV,
   and BLCA. In addition, we established a Cox proportional hazards
   regression model using the Coxph function of the R software package
   survival (version 3.2–7) to analyze the relationship between gene
   expression and prognosis in each tumor. Logrank test was used for
   statistical testing to obtain prognostic significance. Finally, we
   observed that in CESC, TCGA, TCGA-STES, HNSC), COAD High expression of
   COADREAD, THCA, and READ has a poor prognosis, while low expression in
   GBMLGG, KIRC, and LIHC has a poor prognosis (Fig. [99]7B, C).

Fig. 7.

   [100]Fig. 7
   [101]Open in a new tab

   Differences in FABP4 between tumor tissue and normal tissue in pan
   cancer (A). The impact of FABP4 on stage staging and survival in pan
   cancer (B, C)

Exploring the causal relationship between FABP4 and lung squamous cell
carcinoma through dual sample Mendelian random analysis

   We conducted Mendelian random analysis on the causal relationship
   between exposure and outcome using five main MR methods (MR Egger,
   Weighted median, Inverse variance weighted, Weighted mode, Simple
   mode). Through the main method of evaluating causal effects, Inverse
   variance weighted, it was found that FABP4 is a risk factor for lung
   squamous cell carcinoma (OR [95% CI] 1.31 [1.09–1.57], p-value = 0.004)
   (Fig. [102]8A). Meanwhile, both the forest plot and scatter plot
   indicate that FABP4 is a risk factor for lung squamous cell carcinoma
   (Fig. [103]8B, C). In order to evaluate the reliability of the results,
   we used the retention method and funnel plot to evaluate our model. We
   found that in the sensitivity analysis, when a single SNP was
   systematically deleted and MR analysis was rerun, we found no
   significant bias in the observed causal relationship, and the funnel
   plot was no longer offset, indicating that there was no bias in this
   study (Fig. [104]8D, E). We use MR Egger, Inverse variance weighted for
   heterogeneity detection while also utilizing Egger_ Intercept is used
   for horizontal pleiotropy testing. We found that our causal conclusions
   do not exhibit heterogeneity or level of validity (Fig. [105]8F).
   Finally, we used lung squamous cell carcinoma as the exposure factor
   and FABP4 as the outcome factor using a two sample Mendelian analysis.
   We found that there is no causal relationship between lung squamous
   cell carcinoma and FABP4, in other words, FABP4 is a risk factor for
   lung squamous cell carcinoma. This conclusion is not affected by the
   reverse causal relationship (Fig. [106]8G).

Fig. 8.

   [107]Fig. 8
   [108]Open in a new tab

   Two sample Mendelian random analysis of FABP4 and lung squamous cell
   carcinoma (A). Forest map, scatter plot, forest map and funnel plot
   (B-E). Heterogeneity analysis and horizontal pleiotropy analysis (F).
   Reverse Mendelian random analysis (G)

Exploring the causal relationship between FABP4 and benign lung tumors and
other subtypes of lung cancer through dual sample Mendelian random analysis

   To further understand whether FABP4 has a causal relationship only with
   lung squamous cell carcinoma, we used a two sample Mendelian random
   analysis to explore the causal relationship between FABP4 and benign
   lung tumors and other subtypes of lung cancer. We found no causal
   relationship between FABP4 and benign lung tumors, lung adenocarcinoma,
   and small cell lung cancer through the main method of evaluating causal
   effects, Inverse variance weighted (Fig. [109]9A–C).

Fig. 9.

   [110]Fig. 9
   [111]Open in a new tab

   Two sample Mendelian random analysis (A–C) of FABP4 in benign lung
   tumors, lung adenocarcinoma, and lung squamous cell carcinoma

Discussion

   Using existing large-scale GWAS data for Mendelian randomization
   research has the advantages of being faster and cheaper. Mendelian
   randomization studies can understand the potential causal relationships
   between modifiable risk factors and exposed diseases, which often
   require extensive sample sizes and long-term follow-up to draw
   conclusions. However, one drawback of MR design is that it can only be
   applied to risk factors with appropriate genetic variations. Genetic
   variations typically have a significant impact on most risk factors,
   which may lead to low statistical confidence and the risk of false
   negative results in MR analysis. Therefore, we need to collect clinical
   data to validate our research conclusions.

   We reported a gene called Fatty Acid Binding Protein 4 (FABP4) that can
   serve as a risk factor for the development of lung squamous cell
   carcinoma, predict the prognosis of lung squamous cell carcinoma
   patients, and serve as a marker for macrophage infiltration in the
   immune microenvironment of lung squamous cell carcinoma. The impact of
   FABP4 on lung squamous cell carcinoma has never been explored before
   our research report. The only studies in the field of lung cancer
   include SIRT5 promoting cancer progression in non-small cell lung
   cancer by reducing FABP4 acetylation levels, and the increased
   expression of FABP3 and FABP4 being associated with poor prognosis in
   non-small cell lung cancer [[112]25, [113]26]. Therefore, exploring the
   effect of FABP4 on lung scales has great research significance.

   Fatty Acid Binding Protein 4 (FABP4) is a member of the intracellular
   lipid binding protein family, which binds hydrophobic ligands to
   regulate the transport of extracellular lipids and plays an important
   role in intracellular metabolic pathways [[114]27–[115]29]. FABP4 is
   abnormally expressed in many types of cancer and activates multiple
   oncogenic signaling pathways by promoting lipid transport [[116]30]. In
   addition, FABP4 plays an important role in the infiltration of immune
   cells in the tumor immune microenvironment, especially in the
   infiltration of macrophages [[117]21–[118]33]. In this study, we found
   that FABP4 is a marker of macrophage infiltration in the immune
   response environment of lung squamous cell carcinoma. In addition, when
   using Cibersort to analyze the infiltration of 22 immune cells in the
   high and low FABP4 groups of lung squamous cell carcinoma patients in
   TCGA and [119]GSE73403, we found that M2 macrophage infiltration was
   more pronounced in patients with high FABP4 expression. Existing
   studies have shown that the infiltration of M2 macrophages has a
   promoting effect on tumor development [[120]34–[121]36], which may be
   the reason why patients with high FABP4 expression have a shorter
   survival period than those with low FABP4 expression.

   Due to the current lack of predictive methods for the occurrence of
   lung squamous cell carcinoma, most patients are diagnosed with lung
   squamous cell carcinoma in the middle and late stages, which leads to a
   significantly poor prognosis for lung squamous cell carcinoma patients,
   with a low 5-year survival rate of only about 15% [[122]37, [123]38].
   Based on this, possible exposure factors for the occurrence of lung
   squamous cell carcinoma can be explored through Mendelian random
   analysis. At present, research using Mendelian random analysis to
   explore exposure factors for lung squamous cell carcinoma includes:
   airflow restriction may be an independent predictive factor for lung
   squamous cell carcinoma [[124]39], gastroesophageal reflux disease is
   positively correlated with the incidence of lung squamous cell
   carcinoma in patients [[125]40], idiopathic pulmonary fibrosis is an
   independent risk factor for lung squamous cell carcinoma and may
   increase the risk of lung squamous cell carcinoma [[126]41],
   Hypothyroidism has a genetic protective causal relationship with lung
   squamous cell carcinoma [[127]42]. In this study, we found that FABP4
   is a risk factor for lung squamous cell carcinoma, and there is no
   causal relationship between the occurrence of benign lung tumors, lung
   adenocarcinoma, or small cell lung cancer. This has a positive effect
   on predicting the causal relationship specific to lung squamous cell
   carcinoma. At present, research on the anti-tumor effects of FABP4
   inhibitors is still limited to some tumors. For example, the FABP4
   small molecule inhibitor BMS-309403 can enhance the sensitivity of
   ovarian cancer cells to carboplatin [[128]17], inhibit lipid transfer
   of TME with FABP4 inhibitors, reduce tumor regeneration, and inhibit
   tumor regeneration through inherited or pharmacological targeted SCD1
   [[129]43]. FABP4 inhibitors can inhibit the metastasis of
   cholangiocarcinoma cells [[130]44].

Conclusions

   In summary, in our study, we integrated Mendelian randomization,
   bioinformatics, and single-cell RNA sequencing analysis to screen for
   the"star"molecule fatty acid binding protein 4 (FABP4), which may
   provide some guidance for more specific treatment of lung squamous cell
   carcinoma in the future.

Limitation

   We only use information from lung squamous cell carcinoma patients in
   public databases to construct prognostic model, lacking our own
   collection of lung squamous cell carcinoma patients to validate the
   prognostic model. Finally, due to the limitations of the analysis model
   type and experimental conditions in our article, we did not conduct
   corresponding in vitro and in vivo validations.

Acknowledgements