Association between air pollution and dry eye: insights from network toxicology and molecular docking analysis

Abstract

Background

In recent years, air pollution has been increasingly recognized as a contributing factor to the onset of different diseases. This study seeks to investigate the potential link between air pollutants and dry eye (DE), as well as to identify key genes that may play a pivotal role in this relationship.

Methods

A comprehensive search of multiple online databases was conducted to determine target genes associated with both air pollutants and DE. Protein-protein interaction (PPI) analysis and visualization were performed for the overlapping genes. This was followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses to elucidate potential mechanisms. Subsequently, molecular docking and molecular dynamics simulation (MDS) analysis were employed to examine interactions between key genes and air pollutants.

Results

Eight ubiquitous air pollutants (benzene, SO₂, O₃, NO, CO, toluene, formaldehyde, and naphthalene) were chosen for analysis, and a total of 26 intersecting targets associated with DE were identified. Functional enrichment analyses using GO and KEGG pathways indicated that these targets are predominantly involved in biological processes including lipid metabolism, hormone regulation, oxidative stress, and inflammatory responses. This analysis underscored six pivotal genes linked to air pollutants and DE. Molecular docking analysis substantiated potential interactions between air pollutants and these core targets. Futhermore, MDS analysis revealed more stable relationships between MMP-2, EGFR, MMP-9, ESR-1 and air pollutants.

Conclusion

These findings indicate that TNF, EGFR, MMP-9, ESR-1, ICAM-1, and MMP-2 are integral to the mechanisms through which air pollutants influence DE. This research provides fresh insights about molecular mechanisms linking air pollutants and DE, emphasizing the critical need to enhance public awareness regarding the pervasive issue of air pollution.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40360-025-00994-0.

Introduction

In recent years, air pollution constitutes a significant global health threat to human populations, ranking as the second leading risk factor for mortality on a worldwide scale [1]. Air pollution not only has adverse impacts on the external environment but also is closely associated with human health disorders directly. Some investigations revealed that air pollution was strongly linked to the progression of respiratory illness, cardiovascular diseases, neurodegenerative disorders, and even cancers [1, 2]. At present, common atmospheric pollutants encompass particulate matter (PM2.5, PM10), sulfur dioxide (SO₂), ozone (O₃), nitrogen dioxide (NO₂), nitric oxide (NO), carbon monoxide (CO), volatile organic compounds (VOCs), and polycyclic aromatic hydrocarbons (PAHs) [3–5] which possess a specific representativeness of air pollution.

Specifically speaking, among them, main indoor air contamination released by home decoration and furniture, such as benzene, toluene, and formaldehyde, all belong to VOCs mentioned above. Among them, benzene is an aromatic hydrocarbon with the characteristics of being flammable and volatile, and it is toxic to the human body, and even carcinogenic. Chronic exposure of it is linked to a heightened risk of cardiovascular disorders and its metabolites can increase cardiac vascular permeability in vitro experiment [6]. In terms of eyesight, this organic solvent can lead to acquired dyschromatopsia which is a sensitive early marker of neurotoxic damage [7]. Toluene is another primary type of VOCs. Similar to benzene, toluene is prone to volatilize into the air, and their main sources include industrial emissions, traffic exhaust, and the volatilization of household paints and cleaning agents in daily life having high similarities too. Short-term exposure can irritate eyes, nose and throat, causing dizziness and headache. Relevant studies showed that inhalation of toluene with low concentrations, even for short single exposures, can cause gaze destabilization and optokinetic nystagmus, together with vertigo symptoms [8], and even damage the nervous system of vulnerable groups [9]. Formaldehyde in air pollution is a colorless, pungent VOC with high toxicity. It primarily comes from outdoor fuel combustion, vehicle emissions, etc., while indoors it roots in building materials and furniture. Inhaling formaldehyde can irritate the respiratory tract and previous research deemed that prolonged exposure to ambient formaldehyde concentrations may significantly increase the risk of respiratory mortality, leading to a substantial mortality burden for people [10]. Being classified as a carcinogen, it may pose cancer risks such as lung cancer [11]. Further on, it was observed that exposure to formaldehyde gas at environmental levels aggravates cell death and inflammation observed in dry air conditions and in-vitro model of dry eye (DE) can serve as a pertinent instrument for examining and elucidating inflammatory responses observed in patients with DE [12]. As far as PAHs is concerned, naphthalene containing two benzene rings is a frequent air toxic and its main sources are similar to those of VOCs like toluene, including industrial emissions and waste incineration, etc. This kind of substance can induce acute sensory irritative effects at the eyes and upper airways in exposed workers [13]. And an exploratory study of gaseous pollutants found that cancer risk was mainly due to naphthalene [14]. In comparison to particulate matter, SO2, O3, nitrogen oxides, and carbon oxides, these substances present even more significant risks to both environmental integrity and human health. In the meantime, it is surveyed that all of these pollutants may adopt mechanisms, including oxidative stress, inflammation, excitotoxicity, and epigenetic modifications to induce development of various diseases [1, 15].

DE is a complex ocular disease markedly characterized by tear film instability, hyperosmolarity, and ocular surface inflammation which are key factors contributing to this condition [16]. The TFOS DEWS II Epidemiology Report indicates that DE prevalence ranges from 5–50% [17], with higher rates observed in Asia compared to Europe and North America [18]. Recent estimates suggest that 10–20% of individuals over the age of 40 experience moderate to severe symptoms or search for treatment of DE [16]. Age, female sex, genetics, and ethnicity are all definite risk factors for this disease [16]. Furthermore, several studies have concluded that exposure to environmental factors such as airborne pollutants were associated with risk of DE outpatient visits [3, 19]. Since atmospheric pollutants can affect DE primarily through necroptosis, autophagy, inflammatory responses, and oxidative stress [20], causing a massive burden to individuals and society, it is imperative to focus on understanding the specific physiopathological mechanisms involved in order to effectively prevent and manage the onset and progression of this disease.

Bioinformatics technology has flourished currently which has been considered as a new feasible tool for exploring potential interactions between environmental factors and diseases. This area of study is particularly emphasized within the field of network toxicology. Network toxicology, a branch of network pharmacology, utilizes network analysis methods to substantiate the toxicity of targeted substances. It integrates multi-disciplinary knowledge such as bioinformatics, systems biology, and chemical informatics to analyze complex interactions and can be widely used in assessing certain impact of exposure to environmental contaminants, industrial chemicals, and medical drugs, thereby helping to determine the safety of chemicals through predicting toxicological properties and identifying potential molecular targets involved in disease pathogenesis. It unravels the possible methods in which chemical substances interfere with biological networks and influence cell functions, thereby leading to diseases. In addition, molecular docking is a pivotal molecular simulation technique used to study intermolecular interactions at the atomic level, such as the intercorrelations between ligands, small molecules and receptor macromolecules. Its core mechanism is built based on simulating spatial and energy matching between molecules, predicting binding patterns and affinity. Molecular dynamics simulation (MDS) is a computational technique grounded in classical mechanics, employed to simulate the trajectories of atoms and molecules within molecular systems and to analyze their temporal evolution. This method operates on the principle of characterizing motion state of target particles by solving Newton’s equations of motion.

However, the comprehensive elucidation about the effects of air pollutants on DE remains incomplete, and precise molecular mechanisms involved are still not fully known. This study represents a novel investigation into the potential impacts of eight prevalent air pollutants (benzene, SO₂, O₃, NO, CO, toluene, formaldehyde, and naphthalene) on DE. Employing network toxicology, molecular docking and molecular dynamics simulation methodologies, this research aims to offer new insights into these interactions for the first time. A comprehensive methodology is shown in Fig. 1, with schematic diagram of this research is plotted vividly in Fig. 2.

Fig. 1.

A complete workflow of the methodology

Fig. 2.

Schematic diagram of the research

Methods and materials

Identification of air pollutants’ toxicity

Chemical structures and associated molecular information for eight air pollutants were retrieved from the PubChem database (https://pubchem.ncbi.nlm.nih.gov). Potential for eye corrosion and irritation of these pollutants, used as criteria for assessing toxicity, was evaluated on the ADMETLAB 3.0 website (https://admetlab3.scbdd.com).

Collection of target genes associated with air pollutants

Potential human target genes for the aforementioned air pollutants were sourced from the TargetNet database (http://targetnet.scbdd.com), the Swiss Target Prediction database (http://www.swisstargetprediction.ch/), and the SEA database (https://sea.bkslab.org/). Data from these three databases were consolidated, the duplicate data were deleted, and a comprehensive and accurate set of target genes for these air pollutants was compiled.

Collection of genes associated with DE

Genes linked to DE were gathered from the GeneCards database (https://www.genecards.org/) and OMIM database (https://omim.org/) using the keyword “dry eye”. After merging the genes in the two databases and removing duplicates, a gene dataset related to DE was got.

Construction of a PPI network and core target screening for air pollutants and DE

Intersection processing of air pollutants and DE targets was conducted on the Venny 2.1.0 website (https://bioinfogp.cnb.csic.es/tools/venny/). And because PPI network can analyze direct interactions between the input target genes and generate a network of interactions where genes are represented as nodes, its analysis of all obtained intersecting genes between air pollutants and DE was then performed utilizing the STRING database (https://cn.string-db.org/), applied with a confidence threshold of ≥ 0.4 to identify interactions of biological significance, specifically focusing on the human species. The final data were subsequently visualized through construction of a protein interaction network using Cytoscape version 3.10.3 (https://cytoscape.org/).

GO/KEGG enrichment analysis

GO functional annotation and KEGG pathway enrichment analysis of these overlapping genes associated with air pollutants and DE were conducted using the Metascape database (https://metascape.org/). This database further revealed potential roles of key genes in biological processes (BP), molecular function (MF) and cellular components (CC), as well as main signaling pathways in which they were involved. Specifically, BP refers to the biological targets associated with genes, MF pertains to the functional activities of specific genes, and CC denotes cellular locations where genes are situated. So as to acquire a more intuitive visualization of results, platform for online data visualization (https://www.bioinformatics.com.cn) which was a convenient tool for data presentation was provided, enhancing visualization and readability of these results.

Molecular docking analysis

PDB format files of key genes were procured from the RCSB database (https://www.rcsb.org/), while molecular structure files of these eight air pollutants were downloaded from the PubChem database. Structure-based molecular docking analysis was performed using the CB-Dock2 platform (https://cadd.labshare.cn/cb-dock2/php/blinddock.php#job_list_load) and the Autodock Vina software for the interactions between genes and the structures of air pollutants. The binding energy was employed as a metric to assess specific interactions between ligands and receptors: a binding energy below 0 kcal/mol signifies spontaneous binding, whereas a binding energy below − 5 kcal/mol indicates stable binding capacity. Ultimately, molecular docking results were visualized on the online visualization platform (https://www.bioinformatics.com.cn) and a heatmap based on CB-Dock2 platform was generated to show different interaction intensities between these critical genes and atmospheric pollutants.

Molecular dynamics simulation

Genes and air pollutants, also called protein and small molecule complex systems, were performed for 100 ns MDS using Gromacs 2022.2 software after molecular docking analyses. By hydrogenating, repairing, and solvating proteins, this dynamic characterization simulation was performed under constant temperature, constant pressure, and periodic boundary conditions. The optimal conformations of these small molecules were generated by the ATB version 3.0 platform (https://atb.uq.edu.au/) to acquire the corresponding topology files. The small molecules adopted the GAFF force field, while the proteins used the AMBER14SB force field and TIP3P water model. All hydrogen bonds involved were constrained using the linear constraint solver (LINCS) algorithm, with an integration step size of 2 fs. Electrostatic interactions were calculated using the particle-mesh ewald (PME) method, with a cutoff value set at 1.2 nm. Cutoff value for non-bonded interactions was set at 10 Å (1 Å = 0.1 nm), and the simulation was updated every 10 steps. Simulated temperature was controlled by the V-rescale temperature coupling method at 298 K, and the pressure was controlled using the Berendsen method at 1 bar (1 bar = 100 kPa). At 298 K, 100 ps constant atoms, volume, temperature (NVT) and constant atoms, pressure, and temperature (NPT) equilibrium simulation were conducted, and 100 ns MDS of the complex systems were performed. Conformations were saved every 10 ps.

After the simulation was completed, simulation trajectories were analyzed using visual molecular dynamics (VMD) and Pymol 3.1 software, and the molecular mechanics Poisson-Boltzmann surface area (MMPBSA) binding free energy analyses between key genes and air pollutants, were conducted using the g_mmpbsa program. Various structural parameters such as root mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration (Rg), buried solvent accessible surface area (Buried SASA), principal component analysis (PCA) and free energy landscape (FEL) were also calculated [21, 22].

Results

Toxicity assessment of air pollutants

Toxicity assessments for these eight air pollutants were conducted on a toxicity prediction platform. According to the selection criteria, if the eye corrosion and irritation of these pollutants were labeled in red on the ADMETLAB 3.0 platform, this kind of air pollutant was considered toxic to eyes. Finally, all eight air pollutants met the criteria and passed the toxicity tests (Table 1).

Table 1.

Molecular weight, SMILES structure and ocular toxicity of air pollutants

Name	Molecular weights(g/mol)	SMILES structures	Eye Corrosion	Eye Irritation
Benzene	78.11	C1 = CC = CC = C1	1.0	1.0
SO2	64.069	O = S = O	0.875	0.99
O3	47.998	[O-][O+] = O	0.973	1.0
NO	30.006	[N] = O	0.966	0.992
CO	28.01	[C-]#[O+]	1.0	1.0
Toluene	92.14	CC1 = CC = CC = C1	1.0	1.0
Formaldehyde	30.026	C = O	0.991	0.997
Naphthalene	128.169	C1 = CC = C2C = CC = CC2 = C1	0.935	0.995

Collection of target genes associated with air pollutants

After integrating target prediction data from the TargetNet, Swiss Target Prediction, and SEA databases, potential target gene information for each air pollutant were gained. In the TargetNet database, the possibility was set to be greater than zero, and all included models had area under the curve (AUC) value not less than 0.7. Fingerprint type was ECFP4 fingerprints by default. In the end, there were 249 genes for benzene, 219 for SO₂, 209 for O₃, 223 for NO, 230 for CO, 247 for toluene, 244 for formaldehyde, and 243 for naphthalene which were included. In the Swiss Target Prediction database, study species were selected as “Homo sapiens”, filtering possibility was set to be above zero. Molecules with less than 5 heavy atoms, included SO2, O3, NO, CO, and formaldehyde, cannot run through Swiss Target Prediction. Finally, 3 genes for benzene, one for toluene and 3 for naphthalene were acquired sparsely. In the SEA database, the target ID was set to be human similarly. And because benzene, SO2, O3, NO, CO, and formaldehyde didn’t meet the standards, they were all excluded. Meanwhile, 16 genes for toluene and 9 genes for naphthalene were got. After removing duplicate genes completely, 330 air pollutants-related target genes were ultimately identified in these three databases (Supplementary Table 1).

Collection of genes associated with DE

After searching specific keyword “dry eye”, 672 related target genes from the GeneCards database and 26 genes from the OMIM database were obtained credibly. The OMIM database was updated on April 11th, 2025. After duplicate entries were eliminated, a final set of 680 genes related with DE were identified (Supplementary Table 2).

PPI network construction and core target screening for air pollutants and DE

Intersection of air pollutants and DE target genes was conducted on the online Venny 2.1.0 website (https://bioinfogp.cnb.csic.es/tools/venny/) as depicted in the form of a Venn diagram in Fig. 3. 26 intersecting target genes associated with air pollutants and DE were acquired and then imported into the STRING database for further PPI analysis. Category of organism was “Homo sapiens”. Required interaction score was no less than 0.4. After filtering out isolated genes in this network, 23 air pollutants-DE-related targets remained (Fig. 4A). Meanwhile, because the interaction between nodes is represented as edges in PPI, the degree value of a node is determined by the number of edges connected to it. The more edges there are, the higher the degree value, indicating that the node is more prominent. There were 100 edges and average node degree was 7.69 through PPI analysis. Cytoscape version 3.10.3 was used to visualize this PPI network then. In this new network, all targets were ranked by degree, darker colors and larger circles indicated stronger interactions with other proteins (Fig. 4B). There were 23 critical targets with 100 connective edges, and the average node connectivity was 8.696 in this PPI core network map. It provides an unambiguous summarization of interrelations between key targets, paving the way for a more in-depth investigation into the molecular mechanisms connecting air pollutants with DE.

Fig. 3.

Intersection of targets between the eight air pollutants and DE-related genes

Fig. 4.

Establishment, visualization of PPI network and GO enrichment and KEGG pathway enrichment analysis. (A) The PPI network was constructed utilizing the STRING database with a confidence score threshold of ≥ 0.4. Nodes indicate proteins, while edges represent interactions between them. The network elucidates functional associations among target genes related to air pollutants and DE. (B) The PPI network was visualized and analyzed using Cytoscape version 3.10.3. Nodes are colored and sized according to their degree values, with darker colors and larger circles signifying stronger interactions. (C) The histogram of GO enrichment analysis of targets associated with air pollutants and DE. (D) Sankey dot map of KEGG pathway enrichment analysis of targets associated with air pollutants and DE

GO/KEGG enrichment analysis

Through cross-over analysis of genes associated with air pollutants and DE, 23 overlapping genes were recognized. GO functional enrichment analysis revealed that these genes were predominantly involved in various biological processes such as cellular response to lipid, response to amyloid-beta, cellular response to hormone stimulus, and cellular response to nitrogen compound (Fig. 4C). In terms of cellular component, these genes were mainly related to external side of plasma membrane, extracellular matrix and membrane raft. Nuclear receptor activity, serine-type peptidase activity, and kinase binding were primary molecular functions linked with these genes. What’s more, KEGG pathway analysis revealed significant enrichment in several critical pathways, including fluid shear stress and atherosclerosis, endocrine resistance, proteoglycans in cancer, IL-17 signaling pathway, and ras signaling pathway according to Gene Ratio (Fig. 4D). Top terms of GO enrichment results were visualized in the bubble diagram (Fig. 5A). These results demonstrate that environmental factors like air pollutants may influence the occurrence and development of DE by lipid metabolism modulation, hormone regulation, oxidative stress, and inflammatory responses, showing their huge potential by regulating various biological processes and cellular signaling pathways.

Fig. 5.

GO enrichment analysis and KEGG pathway enrichment analysis. (A) The bubble diagram illustrating the top terms of GO enrichment analysis. The size of each bubble is proportional to the count, and the color of the bubble denotes the P value. (B) The bubble diagram of KEGG enrichment analysis of 23 intersection targets, highlighting the top 5 significant enrichment terms. The X axis represents the gene ratio; the Y axis denotes the enrichment pathway; the size of each bubble reflects the number of genes; the color of the bubble indicates the P value. (C) The top 5 terms of KEGG pathways enrichment analysis. The length of each bar corresponds to the gene counts, with taller bars indicating larger counts and higher enrichment levels

Further, the bubble diagram exhibited the top 5 enriched KEGG signaling pathways, arranged in ascending order of p-values (Fig. 5B). Each bubble portrayed a specific pathway mentioned above, with the area of the bubble corresponding to the total number of enriched genes within that pathway. Color intensity of each bubble indicated the importance of enrichment result. The redder the color was, the higher significance it indicated. It has been proved that the core targets Tumor Necrosis Factor (TNF), Epidermal Growth Factor Receptor (EGFR), Matrix Metallopeptidase 9 (MMP-9), Estrogen Receptor 1 (ESR-1), Intercellular Adhesion Molecule 1 (ICAM-1), and Matrix Metallopeptidase 2 (MMP-2) are chiefly linked with fluid shear stress and atherosclerosis, and endocrine resistance pathways, which are important mechanisms to account for correlations between ambient air pollutants and DE. In short, these data meant that air pollutants could promote DE by inducing signaling pathways as mentioned above.

As KEGG pathway enrichment analysis resulted in 5 primary enriched signaling pathways in Fig. 5C, it is worthy of delving into these screened-in pathways. Fluid shear stress (FSS) and atherosclerosis, and endocrine resistance have been well studied in human cardiovascular diseases before [23]. While shear stress refers to a kind of friction force formed by constant exposure of blood flow on the surface of endothelial cells, cardiovascular diseases such as atherosclerosis can damage vascular health and affect blood vessels in eyes. Moreover, endocrine resistance implies insufficient response of cells to endocrine hormones, which may occur in numerous diseases. Taking eye diseases into consideration, endocrine anomaly such as androgen resistance could cause DE undoubtedly [24]. On the contrary, an animal experiment revealed that estrogen receptor-1 (ER-1) deficiency can induce inflammation and lipid deposition to meibomian gland (MG) and lacrimal gland (LG), culminating in DE-like manifestations on the ocular surface [25]. A study showed that estrogen treatment in females further amplified ocular hyperalgesia and neuroimmune responses in aqueous deficient DE model [26]. On the other hand, proteoglycans (PGs)-one of the main components for corneal extracellular matrix, as well as glycolipids play a crucial role in cell signaling, and are capable of mediating interactions with pathogens at the ocular surface, thereby influencing a range of ocular conditions, including DE disease [27]. Menon et al. [28] demonstrated that recombinant human proteoglycan 4 (rhPRG4), a kind of PGs, was verified to be a clinically effective therapeutic agent for DE, exhibiting anti-inflammatory properties for key inflammatory pathways and cytokines in corneal epithelial cells. Besides, the IL-17 signaling pathway was closely correlated with acute and chronic inflammatory responses. IL-17, named interleukin (IL)-17, a inflammatory signaling protein acting in immune response, which can be secreted by T helper 17 (Th17) cells was confirmed to activate corneal epithelial barrier disruption in DE [29]. Multiple animal experiments showed IL-17 level increased in mice models with DE [30, 31]. Furthermore, the Ras signaling pathway begins with receptors on the cell surface, which receives signals from the outside like mitogens and transmits signals through a series of related proteins. The rat sarcoma virus (RAS)/rapidly accelerated fibrosarcoma(Raf)/mitogen-activated protein kinase kinase (MEK)/extracellular signal-regulated kinase (ERK) pathway, also known as the mitogen-activated protein kinase (MAPK)/ERK pathway, is part of the Ras signaling pathway which transmits signals within cells as a protein chain. Research showed that activated nuclear factor-B (NF-κB) and MAPK pathways can induce inflammation, stimulate metalloproteinase production, disrupt epithelial junction structures, and affect corneal barrier function, leading to DE ultimately [32]. A study utilizing a DE rat model found that a specific hormone mechanistically activates the MEK-ERK pathway in corneas and conjunctivas of DE rats, thereby ameliorating ocular surface dysfunctions and lesions [33]. Similarly, p38 MAPK signaling pathway was activated in limbal corneal epithelium of severe DE rats. Abnormal differentiation and stemness loss in corneal epithelium can be reversed with a p38 inhibitor treatment in vivo model of severe DE and in vitro corneal epithelium model [34]. Considering that the MAPK/ERK pathway can result in proliferation, migration, and differentiation of corneal epithelial cells, the signaling pathway of Ras might play a significant role in ocular surface disorders therein.

Molecular docking

Among these 23 targets, top six core targets identified by network topological analyses according to degree value in this mapped PPI network were employed for follow-up molecular docking. Their specific topological parameters containing degree, betweenness centrality (BC), clustering coefficient, closeness centrality (CC), and average shortest pathway (ASPL) were depicted in Table 2. To be specific, the degree of a node quantifies the number of connections it has with other nodes within the network. And BC serves as an indicator of a node’s capacity to facilitate connections with other nodes in the network. Meanwhile, the CC value evaluates the degree to which a node’s immediate neighbors are interconnected, with a high CC value signifying the node’s involvement in a closely-knit cluster. Additionally, the ASPL measure represents the mean shortest path distance between a given node and all other nodes within a particular network [35]. The eccentricity centrality can measure the maximum distance from a node to other nodes. Nodes with a lower eccentricity are closer to the center in the network. Neighborhood connectivity can be used to measure degree of connection between direct neighbors of a node. It reflects the clustering of the local neighborhood of a node, that is, whether nodes tend to connect with each other. Topological coefficient is calculated to measure the overlap of the neighbors of a node, showing the degree of sharing of a node’s neighbors with the neighbors of other nodes. It can mainly identify functional similarity or structural redundancy of nodes. The closer the value is to zero, the higher the uniqueness of the node’s neighbors, and the greater the possibility that it is a key bridge node. Radiality measures path efficiency of a node to other nodes in the network. The higher the value, the more likely the node is to be in a core position and have strong centralization.

Table 2.

Topological parameters of core targets

Core targets	Degree	BC	Clustering coefficient	CC	ASPL	Eccentricity	Neighborhood connectivity	Topological coefficient	Radiality
TNF	18	0.217	0.477	0.846	1.182	2	9.5	0.432	0.990
EGFR	18	0.187	0.464	0.846	1.182	2	9.555	0.434	0.990
MMP-9	17	0.149	0.522	0.786	1.273	3	9.941	0.494	0.985
ESR-1	16	0.044	0.608	0.759	1.318	3	10.688	0.509	0.982
ICAM-1	14	0.028	0.670	0.710	1.409	3	11.286	0.537	0.977
MMP-2	12	0.013	0.742	0.667	1.5	3	11.917	0.567	0.972

Statistical results of molecular docking demonstrated that all six key genes can spontaneously bind with these eight air pollutants, as evidenced by binding energies of less than 0 kcal/mol (Fig. 6). Furthermore, KEGG pathways that these six genes were mainly enriched in were also calculated, with p-values < 0.01, all shown in Supplementary Table 3. Notably, MMP-9, MMP-2 bound to toluene, and MMP-2, TNF, EGFR, MMP-9, ESR-1 bound to naphthalene had binding energies lower than − 5 kcal/mol, showing more effective and stable binding abilities between these genes and air pollutants (Figs. 7 and 8). This signifies that the selected air pollutants were verified to exhibit robust structure-activity relationships with these key genes, directly impacting DE-related biological processes and potentially contributing to etiopathogenesis. To ensure the reliability of the docking results, the detailed results of binding energies obtained from the docking processes using the Autodock Vina software were listed in Supplementary Table 4.

Fig. 6.

Heatmap based on the molecular docking results of the CB-Dock2 platform. The heatmap illustrates binding energies (kcal/mol) obtained from molecular docking analysis involving 6 key genes (TNF, EGFR, MMP-9, ESR-1, ICAM-1, and MMP-2) and 8 air pollutants (benzene, SO₂, O₃, NO, CO, toluene, formaldehyde, and naphthalene)

Fig. 7.

Visual analysis of molecular docking. (A) Molecular docking visualization of MMP-9 with toluene. (B) Molecular docking visualization of MMP-2 with toluene. (C) Molecular docking visualization of MMP-2 with naphthalene

Fig. 8.

Visual analysis of molecular docking. (A) Molecular docking visualization of TNF with naphthalene. (B) Molecular docking visualization of EGFR with naphthalene. (C) Molecular docking visualization of MMP-9 with naphthalene. (D) Molecular docking visualization of ESR-1 with naphthalene

Molecular dynamics simulation analysis

Stability analysis

Four pairs of complex systems with the stablest binding abilities including MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from the molecular docking results were selected for subsequent MDS analysis. It was found that these pairs of small molecule complexes still maintained stable binding during MDS with a total duration of 100 ns. RMSD results of the simulation system were plotted in Fig. 9A, and the RMSD value describes the degree of deviation from the initial structure to the next structure during simulation process, which is an important index to evaluate the overall conformational stability of the system. The smaller the deviation, the better the structural stability.

Fig. 9.

Analysis of MDS. (A) RMSD analysis of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right. (B) RMSF analysis of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right. (C) Rg analysis of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right. (D) Distance between small molecules and protein center in MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right. (E) Buried SASA analysis of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right

To be specific, the RMSD value of the complex formed by MMP-2 and naphthalene gradually stabilized within the range of 0–50 ns as the simulation proceeded, indicating that the structure of this complex was gradually stabilizing, and the naphthalene’s binding to the new binding site and the MMP-2’s binding were also gradually stabilizing. As the simulation progressed, there was a significant fluctuation in the binding site of naphthalene within 50–100 ns. The reason might be that the binding site of naphthalene changed at this time. RMSD values of the constituent complexes formed by EGFR, MMP-9, ESR-1 and naphthalene also gradually stabilized and reached equilibrium as the simulations progressed, showing that the structures of these complexes were increasingly maintaining stable. The average RMSD for MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene was 0.28, 0.25, 0.33, and 0.28 nm, respectively (Table 3). The average RMSD after ligand binding for all the complexes got the stability as compare to single protein or ligand.

Table 3.

Average MDS value for all the complexes

Complex	RMSD (nm)	RMSF (nm)	Rg (nm)	Distance(nm)	Buried SASA (nm2)	Binding energy (kJ·mol− 1)
MMP2-naphthalene	0.28	0.13	2.27	1.94	167.54	-65.178 ± 0.793
EGFR-naphthalene	0.25	0.13	1.91	1.06	136.34	-75.958 ± 1.596
MMP9-naphthalene	0.33	0.16	2.12	0.64	159.55	-74.195 ± 1.013
ESR1-naphthalene	0.28	0.14	2.29	1.87	217.47	-71.594 ± 1.004

Residue mobility analysis

RMSF can represent the flexibility of amino acid residues in a protein, reflecting local fluctuation of each atom or residue in entire molecular dynamics trajectory within the molecule. It is often used to measure the displacement amplitude of each atom or residue relative to its average position, helping to identify flexible and rigid regions within the molecule. The smaller the RMSF value, the lower the flexibility of protein residue.

RMSF results of the simulation system were shown in Fig. 9B and residue mobilities can be seen. The average RMSF for MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene was 0.13, 0.13, 0.16, and 0.14 nm, respectively (Table 3). A high fluctuation can be observed in the four complexes. However, the flexibility of the amino acids around naphthalene in the four complexes was relatively low, so the binding of these small molecules to the proteins was correspondingly stable.

Compactness and centroid evolution analysis

Compactness of the selected complexes has been analyzed using the Rg analysis. Overall, Rg represents the root mean square distance of all atoms in the molecule relative to the center of mass, reflecting distribution range of the molecule’s atoms relative to the center of mass. It is a crucial parameter employed to measure the overall structural compactness or tightness of protein-small molecule complexes. The smaller the rotation radius, that is, the smaller the Rg value, the more compact the molecule is; the larger the Rg value, the more loose the molecular structure is, possibly in a fluffy state.

As shown in Fig. 9C, the Rg values of the MMP-2 and naphthalene complex gradually stabilized within 0–50 ns, meaning that the structure of this complex was also gradually stabilizing. Meanwhile, the Rg values of the complexes formed by EGFR, MMP-9, ESR-1 and naphthalene also gradually stabilized as the simulation progressed, suggesting that the structures of these complexes were also gradually maintaining stability. It can be found that EGFR-naphthalene complex showed very least value as compare to all other complexes in Fig. 9C. The other three complexes had a similar Rg value. The average Rg value for MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene was 2.27, 1.91, 2.12, and 2.29 nm, respectively (Table 3).

To analyze the state of small molecules on the protein surface, after obtaining the initial docking sites of small molecules, distances between the centroids of the initial docking site residues and the centroids of the small molecules were analysed. We also calculated the distance between the small molecules and the centroids of the proteins. By analyzing these distances, the binding state of small molecules and the proteins can be understood fully. As presented in Fig. 9D, in the complex of MMP-2 and naphthalene, the distance between naphthalene and the center of MMP-2, and the distance between naphthalene and the binding site gradually tended to stabilize within 0–50 ns, indicating that naphthalene was stably bound to the binding site of MMP-2, and the binding kept stable. The other three pairs of complexes also became more and more stable in combination as the simulation progressed. The average distance value for MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene was 1.94, 1.06, 0.64, and 1.87 nm, showing in Table 3.

Buried SASA analysis

Buried SASA is a metric used to assess which parts of a molecule or molecular complex are buried within the interior and can’t be directly accessed by the solvent. The higher the value of buried SASA, the stronger the interaction between molecules and the larger the contact area.

As shown in Fig. 9E, the Buried SASA values of the MMP-2 and naphthalene complex gradually stabilized within 0–50 ns, indicating that the contact area between MMP-2 and naphthalene stabilized, and the binding between the two gradually became stable. While the buried SASA values of the EGFR, MMP9, ESR-1 and naphthalene complexes remained relatively stable during the 100 ns simulation period, it meant that the binding of these three complexes was relatively stable. The average buried SASA values for all the systems were got. Average value for MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene was 167.54, 136.34, 159.55, and 217.47 nm² (Table 3).

Principal component analysis

PCA is an analytical method carried out to reduce the dimensionality of data and extract main movement patterns of molecules after ligand binding. Through this analysis, complex high-dimensional molecular motion data can be simplified into a few principal components (PCs), and the most significant movement patterns of molecules during simulation process can be captured. PCA analyses on the trajectories of small molecules in these four complexes were conducted, and it was observed that in PCA graphs of these four complexes in Fig. 10A, there was always a red spot with the highest frequency, which represented the relatively well-stable conformation where the small molecule was bound to corresponding protein.

Fig. 10.

Analysis of MDS. (A) PCA analysis of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right. (B) Energy of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right. (C) Binding energy of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right. (D) FEL analysis of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene from left to right

Energy analysis

Without considering the effect of solvation, the van der Waals forces and electrostatic interactions between the complexes were calculated, and changes of the binding forces during the simulation process were analyzed. Here, VDW represents the van der Waals forces and hydrophobic interactions, ELE represents the electrostatic interactions, and Binding represents the sum of VDW and ELE, which can represent binding energy between the small molecules and protein without considering effect of solvation. As shown in Fig. 10B, within the range of 0–50 ns, VDW and ELE in these complexes all gradually became more stable, indicating that the binding between these pairs of small molecules and protein remained stabilized also.

MM-PBSA and residue contribution analysis

Considering the solvation energy, RMSD, Rg, Distance, Buried SASA and interaction energy, the composite trajectories in the steady state were selected and MM-PBSA method was used to calculate the binding energy-related energy term, which were shown in Table 3. ΔEele is the electrostatic interaction between the small molecule and protein, ΔEvdw is the van der Waals’s energy. ΔEpol is the polar solvation free energy, which can represent the electrostatic potential energy. ΔEnonpol is the non-polar solvation free energy, which can reflect the hydrophobic interaction. ΔEele and ΔEnonpol contribute to the mutual binding of protein ligands. The affinity between the ligand and protein is smaller, the higher the affinity. ΔEMMPBSA is the sum of ΔEele, ΔEvdw, ΔEpol, and ΔEnonpol, while Gibbs Binding Energy binding energy ΔGbind is equal to the sum of ΔEMMPBSA and -TΔS, representing binding free energy. Respectively, ΔGbind of MMP-2, EGFR, MMP-9 and ESR-1 to naphthalene were − 65.178 ± 0.793, -75.958 ± 1.596, -74.195 ± 1.013, and − 71.594 ± 1.004 kJ·mol^− 1. These results further verified the stability of the binding of proteins to small molecules from the perspective of energy.

In Table 4, it can be known that among the four complexes, the ΔEvdw was higher than the ΔEnonpol, and both were higher than the ΔEele. Therefore, in the composition of the binding energy of these four complexes, van der Waals force interaction plays a major role, hydrophobic interaction plays a secondary role, and electrostatic interaction plays a complementary role. The ΔEMMPBSA values of the four pairs of small molecules and proteins were − 72.471 ± 1.982 kJ·mol^− 1, -81.948 ± 0.958 kJ·mol^− 1, -80.844 ± 1.211 kJ·mol^− 1, and − 78.181 ± 0.883 kJ·mol^− 1, respectively, indicating that binding energy and affinity of the two were high.

Table 4.

Binding energy and its composition at steady state (unit: kJ·mol^− 1)

Complex	ΔEvdw	ΔEele	ΔEpol	ΔEnonpol	ΔEMMPBSA	-TΔS	ΔGbind
MMP2-naphthalene	-72.429 ± 1.127	-1.572 ± 0.405	12.646 ± 1.194	-11.116 ± 0.008	-72.471 ± 1.982	7.293 ± 1.538	-65.178 ± 0.793
EGFR-naphthalene	-106.489 ± 1.157	-1.437 ± 0.284	37.041 ± 0.93	-11.062 ± 0.038	-81.948 ± 0.958	5.99 ± 0.862	-75.958 ± 1.596
MMP9-naphthalene	-90.351 ± 1.053	-6.558 ± 0.378	28.157 ± 2.251	-12.092 ± 0.068	-80.844 ± 1.211	6.649 ± 0.29	-74.195 ± 1.013
ESR1-naphthalene	-87.035 ± 0.369	-1.506 ± 0.137	22.218 ± 0.623	-11.858 ± 0.041	-78.181 ± 0.883	6.588 ± 1.241	-71.594 ± 1.004

ΔEvdw represents van der Waals energy, ΔEele represents electrostatic energy, ΔEpol represents polar solvation energy, ΔEnonpol represents non-polar solvation energy, ΔEMMPBSA = ΔEele + ΔEvdw + ΔEpol + ΔEnonpol. ΔGbind = ΔEvdw + ΔEele + ΔEpol + ΔEnonpol-TΔS

Meanwhile, the binding energy ΔEMMPBSA was decomposed to obtain concrete contribution of each amino acid to the overall binding energy, and important amino acids in the protein were evaluated, residues in the protein that contributed more to their respective binding energy were shown in Fig. 10C, from which it can be seen that key amino acids of MMP2-naphthalene, EGFR-naphthalene, MMP9-naphthalene, and ESR1-naphthalene were ASP-72 and ASP-96; LEU-777 and MET-766; LEU-188 and MET-247; LEU-505 and MET-421 in order.

FEL analysis

FEL is used to describe the free energy distribution of molecular systems in different conformational states, and it can display various possible conformations of a molecule or system and its relative stability. The results were shown in Fig. 10D, with RMSD and Rg of the complex used as the free energy morphology. The deep blue color in the plot represents the lowest energy conformation while the red color is on behalf of the highest energy conformation. The lower energy transition states are presented by the deep valley with the blue color. It was revealed that one state of low energy existed in all four complexes, indicating that overall structures of these complexes were relatively stable.

Compound structure analysis

For the complex conformations of molecular docking, their structures and interactions were further analyzed by MDS. As shown in Fig. 11A, in the complex formed by MMP-2 and naphthalene, GLY-66, ARG-67 and naphthalene formed Pi-Sigma, Pi-Alkyl hydrophobic interactions, and amino acids such as HIE-70, ASN-64, and HIE-70 formed van der Waals force interactions with naphthalene. As plotted in Fig. 11B, in the complex system formed by EGFR and naphthalene, MET-766, PHE-856, LEU-777, and LEU-788 formed Pi-Sulfur, Pi-Pi T-shaped, and Pi-Alkyl hydrophobic interactions with naphthalene, and amino acids such as CYS-775 and LYS-745 formed van der Waals force interactions with naphthalene. As shown in Fig. 11C, in the MMP9-naphthalene compound, MET-247, HIE-226, VAL-223 and naphthalene formed Pi-Pi Stacked, Amide-Pi Stacked, Pi-Alkyl hydrophobic interactions, and amino acids such as LEU-188, TYR-245, and TYR-248 formed van der Waals force interactions with naphthalene. Finally in Fig. 11D, in the ESR1-naphthalene compound, PHE-404, LEU-346, LEU-349, LEU-391, LEU-387, ALA-350, MET-388 formed Pi-Pi T-shaped and Pi-Alkyl hydrophobic interactions with naphthalene, and amino acids such as LEU-384 and GLU-353 formed van der Waals force interactions with naphthalene.

Fig. 11.

Analysis of the complex structure. (A) Interaction between MMP-2 and naphthalene. (B) Interaction between EGFR and naphthalene. (C) Interaction between MMP-9 and naphthalene. (D) Interaction between ESR-1 and naphthalene

Discussion

Air pollution constitutes a significant environmental hazard in contemporary society, originating from a multitude of sources. The primary sources of these ubiquitous pollutants include emissions from on-road vehicles, industrial activities, and the combustion of coal or other fuels. These pollutants have been shown to have direct negative impacts on the respiratory and cardiovascular systems, contribute to neurodegenerative diseases [36–38], and are closely associated with the development of various chronic diseases [39]. Some studies elucidated that air pollutants can disrupt normal physiological functions within the body through oxidative stress and inflammation, thereby facilitating disease progression [40]. Furthermore, these pathogenic processes can lead to DNA damage, which is considered a significant mechanism underlying the detrimental effects of air pollution [41, 42].

DE is a highly prevalent chronic ocular surface disease influenced by multiple etiological factors in clinical practice nowadays. While age, female sex, genetic predisposition, and ethnicity are established risk factors for DE, there is an increasing acknowledgment of the significant impact that environmental exposure factors have on the onset and progression of certain diseases. They have garnered significant interest in scientific research. In terms of epidemiological data on DE, a clinical survey has found that the prevalence of DE among ophthalmic outpatients in China was quite high. Air pollutants such as O3, PM2.5, and SO2 were identified as risk factors for DE [43]. A retrospective cohort study suggested that higher PM2.5, O3, and NO2 levels were associated with higher DE symptom assessment and tear osmolality of individuals [44]. Recent findings assemble evidence that exposure to environment, including air pollutants, can promote DE by inducing inflammatory responses, allergic reactions, oxidative stress, DNA damage, mitochondrial impairment, toxicity, and suppress proliferation and migration of cells on the ocular surface [45, 46]. Nevertheless, the precise molecular pathogenic mechanisms that connect these environmental factors to DE remain incompletely elucidated. In this investigation, a mixed-method approach was employed, incorporating network toxicology, molecular docking, and molecular dynamics simulation to seek underlying associations between air pollutants and DE. In addition, this research ascertains key genes and their interaction networks, providing in-depth understanding about the part of environmental factors in DE occurrence and contributing to the broader discourse on environmental health and disease prevention and therapy.

To be specific, 26 genes were found to have relationships between air pollutants and DE. GO and KEGG functional enrichment analyses resulted that air pollutants may cause DE by modulating lipid metabolism, hormone regulation, oxidative stress, and inflammatory responses. Among these, TNF, EGFR, MMP-9, ESR-1, ICAM-1, and MMP-2 were recognized as pivotal genes within this interaction network between air pollutants and DE. These findings provide a novel theoretical foundation for exploring possible molecular mechanisms by which air pollutants affect DE, and offering potential molecular targets for early prevention and prognostic evaluation of disease progression.

TNF is a pleiotropic cytokine that can cause inflammation produced by immune system [30]. Due to the association of DE with inflammation, there is a significant correlation between DE and TNF. TNF is involved with DE through multiple mechanisms. The member of the tumor necrosis factor family such as TNF-α is highly expressed in DE which is known to stimulate the activation and maturation of antigen-presenting cells (APCs) with relation to innate immunity [47] and enhance endoplasmic reticulum stress, reduce Ca²⁺ concentration, and stimulate cholinergic agonist secretion to reduce mucin secretion by goblet cells [32]. All of these can exacerbate chronic inflammatory responses on the ocular surface, potentially initiating and advancing the progression of DE. Previous researches have corroborated that TNF-α was notably elevated in clinic patients or animal models with DE, which can be served as a crucial biomarker for clinical diagnosis and severity assessment [48, 49]. Meanwhile, TNF-related inhibitors have been proven to be safe and feasible for DE treatment and improve disease syndromes [50, 51].

EGFR is a member of the ERBB family of tyrosine kinase receptors. The signaling cascade initiated by EGFR can regulate cellular processes such as proliferation, differentiation, division, survival, and the development of cancer [52]. Ligands of EGFR have the capacity to activate the PI3K/AKT signaling pathway through receptor binding, thereby enhancing anti-apoptotic, anti-inflammatory, proliferative, and migratory functions, as well as facilitating wound healing in the corneal epithelium. But this facilitated pathway will be inhibited in corneal epithelial cells from individuals with DE [53]. α-melanocyte stimulating hormone has been shown to ameliorate lesions and enhance ocular surface functions by upregulating EGFR expression in DE [54]. Similarly, eye drops such as diquafosol are effective in facilitating the healing of corneal epithelial wounds, potentially through the stimulation of cell proliferation and migration via the EGFR/ERK signaling pathway [55]. Conversely, systemic EGFR inhibitors have the potential to exacerbate dry eye symptoms in oncology patients and impede epithelial cell proliferation and stratification during corneal regeneration in cases of dry eye disease [56, 57].

MMP-9 is a class of enzymes that belongs to the zinc-metalloproteinases family primarily involved in the degradation of extracellular matrix (ECM) and tissue remodeling. Numerous studies have expounded that levels of MMP-9 increased in DE conditions [47, 58]. In response to hyperosmolar environments, the production of MMP-9 contributes to disruption of corneal barrier, and rises with the severity of DE [58]. Therefore, related researches have validated that both direct and indirect inhibition of MMP-9 can serve as therapeutic strategies for DE. And these strategies are designed to enhance corneal barrier function, mitigate inflammation on the ocular surface, and restore tear production [59]. A highly soluble MMP9 inhibitor has been identified as a promising candidate for the potential treatment of DE, as corroborated by recent studies [60].

ESR-1 belongs to the NR3 subfamily of the nuclear receptor family. It plays an increasingly important role in proliferation, differentiation and homeostasis of cells. ESR-1 functions by binding to estrogen and activating the estrogen response element. And it mainly plays an important role in diseases related to sex hormones such as ovarian cancer and hormone receptor-positive (HR⁺) breast cancer [25, 61]. Recently it has been discovered that ESR-1 also plays an indispensable role in ocular surface diseases. Existing research indicates that the detection of ER alpha reinforces the hypothesis of a receptor-mediated mechanism of estrogen in these tissues, thus contributing to DE [62]. An animal experiment elucidated that ESR-1 is pivotal for maintaining ocular surface homeostasis, revealing that ESR-1 deficiency can induce inflammation and lipid accumulation in the meibomian and lacrimal glands, ultimately resulting in DE-like symptoms on the ocular surface [25].

ICAM-1 is a kind of cell surface glycoprotein which is thought as an adhesion molecule and it is part of the immunoglobulin superfamily. It is integral to cell adhesion and immune responses, facilitating interactions among immune cells. Current research suggests that ICAM-1 interacting with lymphocyte function-associated antigen-1 can facilitate the homing of dendritic cells to the lymph nodes, interaction of dendritic cells with T cells and subsequent activation or differentiation of T cells, migration of activated CD4⁺ T cells from lymph nodes to the ocular surface, reactivation of T cells by resident APCs at the ocular surface, playing a crucial role in cell-mediated immune response and inflammation associated with DE [63]. An in vitro experiment demonstrated that overexpression of miR-146a reduced TNF-α-induced ICAM-1 expression, thereby modulating inflammatory response via the associated pathway in DE [64]. On the contrary, the upregulation of ICAM-1, both locally and systemically, can enhance lymphocyte activation and facilitate their migration to the ocular surface, which is associated with DE [65].

MMP-2, a zinc-dependent enzyme within the matrix metalloproteinase family, is integral to the degradation and remodeling of ECM. It is also implicated in numerous physiological processes, encompassing tissue repair, wound healing, angiogenesis, and embryogenesis [66]. The dysregulation of MMP-2 is related to the occurrence of various diseases, notably ocular conditions. Several studies have documented increased activity and elevated levels of MMP-2 in individuals with DE [67, 68].

Results of molecular docking analyses confirmed that all six key genes exhibit effective binding with these eight air pollutants, with MMP-9, MMP-2 binding to toluene, and MMP-2, TNF, EGFR, MMP-9, and ESR-1 binding to naphthalene, showing enhanced stability and improved binding energies than other combinations. Besides, through MDS analysis, it was found that MMP-2, EGFR, MMP-9, ESR-1 and naphthalene could bind more stably. These findings provide additional evidence for the hypothesis that air pollutants may directly interact with these key genes to modulate their biological functions, possibly posing significant health risks associated with DE.

This study has several strengths. Firstly, the extensive implementation of a methodology integrating network toxicology and molecular docking to investigate the relationship between air pollutants and DE has enhanced the scientific rigor of this research. Secondly, by focusing on the intrinsic molecular mechanism of air pollutants and DE, the connection at the molecular level between air pollutants and DE has been further clarified. But this study also has some limitations. To our best knowledge, these genes collected from commonly used databases may lead to selection bias of the results inevitably. And the subjective selection of ambient air pollutants for DE raises concerns regarding bias similarly. Further, all included target information was obtained from predictions generated by network databases, such as the network toxicology platform. These predictions may be affected by the underlying algorithms and the quality of the original data, potentially leading to deviations from the actual situation. Meanwhile, this study lacks direct experimental evidences to confirm that air pollutants exert their influence on DE through the modulation of these target genes. Consequently, molecular experiments should gain approval to verify these findings.

Conclusion

This present study systematically reveals the potential association between air pollutants and DE through specific application of network toxicology, molecular docking and molecular dynamics simulation techniques. In real life, after taking into account the relationship between clinical patients and public environmental pollution, individuals in highly polluted areas are actually at increased risk for DE. It is suggested wearing wrap-around sunglasses to protect eyes, using indoor air purifiers, and considering preservative-free artificial tears during pollution spikes. Ophthalmologists should ask about air-quality exposure in routine check-ups for DE symptoms. Additionally, DE awareness should be included in pollution-alert systems to provide timely warnings and eye-care advice to at-risk groups.

Returning to the results of this study, it was indicated that six key genes encompassing TNF, EGFR, MMP-9, ESR-1, ICAM-1, and MMP-2 may serve as critical ingredients in mediating the relationship between air pollutants and the onset and progression of DE. This research offers novel perspectives on the underlying molecular mechanisms by which air pollutants influence DE progression in theory, and deepens individual understanding of partial responsibility of air pollution on DE. In addition, these key genes represent promising molecular targets for early prevention, timely assessment of prognosis, and targeted intervention methods for DE. In other words, it is recognized that these findings can provide a scientific basis for reducing air pollution release and alleviating ocular surface diseases possibly, which has important public health significance. It is expected that future studies are warranted to explore functional mechanisms of these genes and their potential clinical applications in addressing DE, thereby contributing to a more robust scientific foundation for understanding the internal effects of air pollution on DE for people.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Y.W. wrote the main manuscript text and conceptualized the thesis, Z.J. and Y.L. confirmed data curation and original analysis, Y.Z. managed technical review and material support, X.C. and H.L. ensured data methodology, and G.L. guaranteed review and editing of the manuscript. All authors reviewed the manuscript and approved the submitted version.

Funding

This work was supported by a grant from National Natural Science Foundation of China (grant number 81774369), Natural Science Foundation of Fujian Province, China (grant number 2023J01852), Fujian Major Research Grants for Young and Middle-aged Health Professionals (grant number 2021ZQNZD012), and Special Project of Fujian University of Traditional Chinese Medicine (X2022018-SP).

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Ethnic and consent to participate declarations

Not applicable.

Competing interests

The authors declare no competing interests.