Mechanistic insights into DEHP-induced progression of non-small cell lung cancer based on network toxicology and molecular docking Authors: Zhe Chen, Chi Wang, Yixing Li, Xiao Xiao, Heng Zhao, Bin He, Haotian Bai, Rui Zhao, Guangjian Zhang, Jinteng Feng Published: 2025 ABSTRACT AbstractDi(2-ethylhexyl) phthalate (DEHP), a widely used plasticizer, has been implicated in various health risks, including tumorigenesis. Non-small cell lung cancer (NSCLC), accounting for over 80% of lung cancer cases, remains a leading cause of cancer-related mortality. This study employed network toxicology and molecular docking to explore the molecular mechanisms underlying DEHP’s toxic effects on NSCLC. DEHP and NSCLC targets were retrieved from CTD, SwissTargetPrediction, and GeneCards, yielding 225 overlapping genes. Protein-protein interaction (PPI) network analysis identified five core targets: TP53, JUN, SRC, AKT1, and ESR1. Gene Ontology (GO) and KEGG pathway enrichment analyses revealed significant involvement of the PI3K/AKT signaling pathway and regulation of apoptotic signaling in DEHP-induced NSCLC pathogenesis. Molecular docking confirmed strong binding affinities between DEHP and the core targets, with binding energies ranging from − 4.611 to -7.535 kcal/mol. These findings suggest that DEHP promotes NSCLC progression, metastasis, and chemoresistance through PI3K/AKT signaling and apoptotic pathway dysregulation. This study provides mechanistic insights into DEHP’s role in NSCLC and highlights the need for public health interventions to mitigate DEHP exposure. Further experimental validation is warranted to strengthen these findings and guide the development of targeted therapies.Keywords: DEHP, NSCLC, Network toxicology, Molecular docking, PI3K/AKT pathway, ApoptosisSubject terms: Cancer genomics, Lung cancer Introduction Di(2-ethylhexyl) phthalate (DEHP), a member of the phthalate ester family1, has become a critical component in the plastics industry due to its cost-effectiveness and efficiency2. However, the widespread use of plasticizers, recognized as carcinogens and/or endocrine disruptors, has raised concerns about DEHP’s potential health risks, including its association with tumorigenesis, which has emerged as a research focus3. Lung cancer, one of the most prevalent malignancies with the highest global incidence and mortality rates, is categorized into small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC)4. NSCLC accounts for over 80% of all lung cancer cases, yet the 5-year survival rate for NSCLC patients remains below 15%5. Epidemiological studies suggest that environmental pollutants may contribute to elevated lung cancer incidence and metastasis. DEHP’s endocrine-disrupting properties, which interfere with hormonal regulation—a critical factor in NSCLC progression—have been well-documented. Experimental evidence confirms that DEHP promotes lung cancer cell initiation6, proliferation7, migration, and invasion8. Network toxicology, an interdisciplinary field integrating bioinformatics, big data analytics, genomics, and related technologies9, enables the exploration of toxic pathways and molecular mechanisms underlying diseases. This approach shifts the traditional “one-target-one-drug” paradigm to a “multi-target” strategy10, elucidating complex interactions among genes, proteins, and metabolites within disease and toxicant networks. Molecular docking, one of the most widely used approaches in drug design11, predicts binding conformations between small-molecule ligands (e.g., toxicants) and target protein binding sites, validating mechanisms of toxic action. This study aims to investigate DEHP’s toxicological effects on NSCLC at the molecular level by identifying DEHP-targeted and NSCLC-associated proteins, constructing interaction networks, and determining hub targets. Subsequent analyses include biological functions, pathway enrichment, and disease ontology exploration, with molecular docking techniques employed to validate DEHP-core target interactions. Work flow of this study is showed in Fig. 1. Work flow of this study. Methods DEHP target identification DEHP targets were identified through the following workflow: Compound Retrieval: The chemical structure of DEHP was retrieved from the PubChem database(https://pubchem.ncbi.nlm.nih.gov/, Version: 2025Feb, access date: 10 Feb 2025) using the keyword “DEHP”. Molecular identity was verified by matching the systematic name and molecular formula. Target Prediction: The SMILES file of DEHP (CCCCC(CC)COC(= O)C1 = CC = CC = C1C(= O)OCC(CC)CCCC) was uploaded to Swiss Target Prediction (http://www.swisstargetprediction.ch/, Version: 2023Oct, access date: 10 Feb 2025), the prediction condition probability > 0 was set, and it was submitted to download the prediction results. Meanwhile, DEHP-related targets were retrieved from the Comparative Toxicogenomics Database (CTD, http://ctdbase.org/ Version: January 31, 2025, access date: 10 Feb 2025) by searching the ‘Chemicals’ module with ‘DEHP’ as the query term. Predicted targets from both platforms were merged, and duplicates were removed to establish the DEHP target library. NSCLC disease target collection NSCLC-associated targets were curated from GeneCards (http://www.genecards.org/, Version 5.23, access date: 10 Feb 2025) using the keyword “NSCLC”. Targets with a relevance score > 1 were selected as NSCLC-related proteins. Target intersection analysis The overlapping targets between DEHP and NSCLC were identified using the online Venn diagram tool (http://bioinformatics.psb.ugent.be/webtools/Venn/, Version 2.0, access date: 10 Feb 2025). DEHP and NSCLC target libraries were uploaded to generate the intersection dataset. PPI network construction STRING Database Analysis: Overlapping targets were submitted to the STRING database(https://string-db.org/, Version 12.0, access date: 10 Feb 2025) to construct the PPI network. In order to ensure the high confidence of information, the isolated proteins in the figure were hidden, and the selected target proteins were limited to “Homo sapiens”. To make the data have a good authenticity and accuracy, we selected the confidence level of 0.9 as the analysis condition. The TSV file of interaction data was downloaded. STRING Database Analysis: Overlapping targets were submitted to the STRING database(https://string-db.org/, Version 12.0, access date: 10 Feb 2025) to construct the PPI network. In order to ensure the high confidence of information, the isolated proteins in the figure were hidden, and the selected target proteins were limited to “Homo sapiens”. To make the data have a good authenticity and accuracy, we selected the confidence level of 0.9 as the analysis condition. The TSV file of interaction data was downloaded. Hub Target Identification: The PPI network was imported into Cytoscape (v3.9.1). Using the Network Analyzer plugin, network topology parameters (e.g., degree centrality) were calculated. Targets with degree values ≥ 2× the median were defined as hub targets. (The centrality threshold (≥ 2×median) was selected to mitigate outlier influence while retaining biologically relevant signals, consistent with robustness criteria in skewed distributions) Hub Target Identification: The PPI network was imported into Cytoscape (v3.9.1). Using the Network Analyzer plugin, network topology parameters (e.g., degree centrality) were calculated. Targets with degree values ≥ 2× the median were defined as hub targets. (The centrality threshold (≥ 2×median) was selected to mitigate outlier influence while retaining biologically relevant signals, consistent with robustness criteria in skewed distributions) Molecular docking validation The molecular docking technique was employed to investigate the potential binding interactions between di(2-ethylhexyl) phthalate (DEHP) and core target proteins associated with non-small cell lung cancer (NSCLC). Protein structure files (PDB format) for the targets were retrieved from the RCSB PDB database: TP53 (PDBID:4KVP), JUN (PDBID:5T01), SRC (PDBID:8JN8), ESR1 (PDBID:5DXE), and AKT1 (PDBID:6S9X), while the structural file (SDF format) for DEHP was obtained from PubChem. Using MGLTools 1.5.7, the protein structures underwent preprocessing, including removal of water molecules, removal of native ligands, and addition of polar hydrogen atoms, before being designated as the receptor and exported in PDBQT format. Concurrently, hydrogen atoms were added to the DEHP ligand molecule and partial atomic charges were calculated. Molecular docking was performed using AutoDock Vina 1.2.5 within a predefined docking grid box, with specified parameters, executing 15 independent docking runs. The conformation exhibiting the lowest binding free energy (ΔG) was selected. Subsequent analysis in PyMOL 3.1.4 utilized the align command to calculate the Root Mean Square Deviation (RMSD) between this lowest-energy conformation and the others. Docking results with an RMSD value below 2.0 Å were retained. Finally, the optimal docking complex was visualized in PyMOL: peptide chains containing the binding pocket were colored Cyan, peptide chains without the binding pocket were colored Marine, and the DEHP ligand was colored Deepsalmon. The resultant visualization was then output as an image for subsequent analysis. GO and KEGG pathway enrichment Differentially expressed genes (DEGs) underwent functional enrichment analysis using the cluster Profiler package (v4.0) in R. Enriched terms were identified for: Gene Ontology (GO): Biological Process (BP), Molecular Function (MF), Cellular Component (CC) and KEGG Pathways(Version Release 113.0, access date: 10 Feb 2025). Statistical significance was set at p < 0.05 with false discovery rate (FDR) correction. Results Retrieval of DEHP and NSCLC targets After removing duplicates from CTD and SwissTargetPrediction, 592 DEHP target genes were obtained. A total of 1,720 NSCLC target genes were identified from GeneCards. Intersection analysis of DEHP and NSCLC targets yielded 225 common genes (Table 1; Fig. 2), representing potential key genes mediating DEHP’s toxicological effects on NSCLC. Table 1225 common genes of DEHP and NSCLC.1AR51PIK3R1101SQSTM1151DIABLO201GNAS2PRKCA52TP53102CCND2152JAG1202GSTP13MAPK1453IGF1103CDH1153JAK3203HPGD4FLT154IRS1104FOS154JUN204HSPB15PGR55NQO1105FOXO3155KIT205ITGA26HSD17B156CDKN1A106HSPA5156LAMC2206ITGB17CTSB57CYP1B1107METTL3157LDHA207MAP2K28CCR158PCNA108PIK3CG158NOTCH3208METTL149PTGS259PRKN109XBP1159PTGER2209MIR15B10CCND360FABP4110CDK4160PTGER4210NAMPT11PDGFRB61MYC111DNMT1161PTGES211NANOG12EGFR62CASP8112DNMT3A162TLR4212PDGFRA13CTSL63CTNNB1113EGR1163TUBB3213PKM14NR3C164INSR114HSP90AA1164UCP2214POLD115PARP165PTEN115STAT5B165ABCC3215PTGS116CYP24A166RELA116TIMP1166ACTB216RPS6KB117ABCG267IL4117VDAC1167ADAMTS1217S100A818CCND168SIRT1118VEGFA168AIFM1218SERPINE119CDK169SLC7A11119ACTA2169ATM219SOX220CCNE170TGFB1120ALDH1A1170BAD220SPP121MAPK871CASP1121BCL2L1171CASP7221STAT122BCL272CDH2122BDNF172CAV1222TAP123PPARG73FAS123CCL2173CDKN2A223WT124CASP374GJA1124DNMT3B174CXCL1224YAP125BAX75TIMP2125GCLC175CXCL10225YTHDF126ESR176TJP1126KEAP1176CXCL827AKT177FASN127LPL177DUSP128HMOX178LEP128NOTCH2178FGF929NFE2L279PPARGC1A129POU5F1179HSP90AB130TNF80SREBF1130PPARD180ID131CYP19A181CASP9131SLC2A1181IGF2R32IL1B82HDAC6132SOX9182IL1833AHR83PDK4133THBS1183LCN234SOD184GSK3B134BECN1184MIR106B35SOD285MAP1LC3B135DDIT3185PIK3CA36MAPK186TSHR136ERBB2186PTCH137MAPK387ADIPOQ137ERBB4187SNAI138IL688CREB1138FOSB188SRC39MMP989HIF1A139GADD45A189TPO40CYP1A190NOTCH1140GSTM1190TXNRD141CYP3A491CSF2141IGF1R191ATF442NLRP392CYCS142NOS2192BCL2L1143GPX493H2AX143PRDX5193CCL544MTOR94IL10144TFRC194CD16345MMP295MKI67145ATF3195CD946NRG196EIF2AK3146BCL6196CYP2E147ESR297EPHX1147BHMT197DCN48IFNG98FOXO1148CCNA1198DNAH849PINK199MIR93149CCNA2199EGLN350VIM100SIRT3150CCNB1200FOXA1 225 common genes of DEHP and NSCLC. Venn diagram illustrating the intersection of DEHP and NSCLC, identifying a total of 225 overlapping targets. PPI network construction and identification of core targets A protein-protein interaction (PPI) network of the 225 shared targets was constructed using the STRING database. The TSV file exported from STRING was imported into Cytoscape for topological network analysis (Fig. 3). Nodes closer to the center, with darker colors and larger sizes, indicated higher degree values, suggesting their potential as core targets. The network comprised 208 nodes and 992 edges, with degree values ranging from 1 to 57 and an average neighbor count of 9.849, reflecting its complexity. Using the CytoHubba plugin in Cytoscape, five core targets—TP53, JUN, SRC, AKT1, and ESR1, were identified based on degree, closeness, and betweenness centrality (Table 2). Protein-protein interaction (PPI) network constructed using the overlapping targets of DEHP and NSCLC. Node color gradient (light to dark) and node size (small to large) represent the increasing degree value. Table 2Degree, closeness, and betweenness centrality of five core genes (TP53, JUN, SRC, AKT1, ESR1).DegreeClosenessBetweenessTP5357123.08337803.817JUN371123404.301SRC35107.66673383.494AKT1411152724.609ESR136111.83332793.035 Degree, closeness, and betweenness centrality of five core genes (TP53, JUN, SRC, AKT1, ESR1). GO and KEGG pathway enrichment analysis To elucidate the biological mechanisms underlying DEHP’s toxicity in NSCLC, GO and KEGG12–14 pathway enrichment analyses were performed on the 225 targets using the R package clusterProfiler. The top three enriched GO-BP terms included regulation of apoptotic signaling pathway, gland development, and response to xenobiotic stimulus. The top three GO-CC terms were membrane raft, membrane microdomain, and apical part of cell. For GO-MF, the top terms were DNA-binding transcription factor binding, RNA polymerase II-specific DNA-binding transcription factor binding, and ubiquitin-like protein ligase binding (Fig. 4). KEGG pathway analysis (Fig. 5) revealed the top five pathways after excluding infection- and nervous system-related terms: PI3K-Akt signaling pathway, MicroRNAs in cancer, Proteoglycans in cancer, Chemical carcinogenesis - receptor activation, and MAPK signaling pathway. Top 10 GO terms of DEHP-NSCLC intersection genes. From top to bottom are Biological Process, Cellular Component, and Molecular Function. Top 30 pathways of DEHP-NSCLC intersection genes. Pathways (Y-axis), gene ratio (X-axis), and -log10(p value) (color change). Bubble size indicates the number of genes enriched in the pathway. Molecular docking of DEHP with NSCLC core targets Molecular docking was performed to investigate interactions between DEHP and five core targets: TP53 (PDBID:4KVP), JUN (PDBID:5T01), SRC (PDBID:8JN8), ESR1 (PDBID:5DXE), and AKT1 (PDBID:6S9X). To ensure data reliability, each molecular docking experiment was repeated 15 times, and the root mean square deviation (RMSD) values for all docking results were consistently within 2 Å, confirming successful docking. Results from AutoDock showed binding energies ranging from − 4.611 to -7.535 kcal/mol. Binding energies < 0 indicate binding activity, with values < -3.0 kcal/mol suggesting strong interactions (Table 3). DEHP exhibited robust binding affinities with all five targets, supporting their critical roles in DEHP-induced NSCLC pathogenesis. Docking results were visualized in 3D using PyMOL (Fig. 6). Table 3Binding affinities between DEHP and five core targets of NSCLC on DEHP.GeneAKT1TP53JUNESR1SRCDEHP-7.535 kcal/mol-5.058 kcal/mol-4.611 kcal/mol-5.291 kcal/mol-6.304 kcal/mol Binding affinities between DEHP and five core targets of NSCLC on DEHP. Molecular docking results between DEHP and five core genes (TP53, JUN, SRC, AKT1, ESR1) mediating DEHP’s toxic effects on NSCLC. (A) DEHP- AKT1; (B) DEHP - TP53; (C) DEHP-JUN; (D) DEHP-SRC; (E) DEHP-ESR1; (F) Heatmap of the absolute binding energy values, where darker colors indicate higher absolute binding energy values. Discussion This study employed network toxicology to integrate data from CTD, PubChem, SwissTargetPrediction, and GeneCards, identifying 225 potential targets linking DEHP to NSCLC toxicity. Through PPI network analysis, five core targets—TP53, JUN, SRC, AKT1, and ESR1—were identified as central mediators. KEGG and GO analyses highlighted key mechanisms, while molecular docking confirmed strong binding affinities between DEHP and these targets. TP53, a tumor suppressor gene mutated in ~ 65% of NSCLC cases15, is activated by DEHP metabolites such as MEHP, triggering mitochondrial apoptosis16. Our findings position TP53 as a pivotal target in DEHP-driven NSCLC pathogenesis. AKT1, a critical node in the PI3K pathway, regulates cell survival and proliferation. DEHP’s activation of PI3K/AKT signaling in neuroblastoma17 and hepatocellular carcinoma18 aligns with its role in NSCLC chemoresistance19 and metastasis20. Our results further validate AKT1 as a key mediator of DEHP’s effects. SRC, a tyrosine kinase, promotes tumor progression via STAT3, FAK, and AKT signaling21–23. Elevated SRC activity in NSCLC adenocarcinomas24 and its role in EGFR-TKI resistance25 underscore its importance, supported by DEHP’s strong binding affinity (-6.15 kcal/mol). KEGG analysis prioritized the PI3K/AKT pathway, corroborated by DEHP’s high binding affinities with AKT1 (-7.535 kcal/mol), SRC (-6.304 kcal/mol), and TP53 (-5.058 kcal/mol). DEHP’s activation of PI3K/AKT in reproductive26 and neuronal toxicity models27 aligns with its proposed role in NSCLC progression, metastasis, and drug resistance. GO enrichment analysis revealed that the most significant biological process (BP) through which DEHP exerts its toxicological effects on NSCLC is the regulation of the apoptotic signaling pathway28. Apoptosis, or programmed cell death, is a physiological process that provides an effective, non-inflammatory means of removing excess or damaged cells from tissues, thereby ensuring tissue homeostasis. Defects in the apoptotic signaling pathway contribute to tumorigenesis and therapeutic resistance, ultimately leading to chemotherapy failure29. Evasion of apoptosis can be considered one of the hallmarks of tumor progression. Based on the GO enrichment analysis results, DEHP may induce the development of NSCLC and chemotherapy resistance by downregulating the apoptotic pathway. In previous studies, DEHP has been demonstrated to promote the progression, migration, and invasion of NSCLC through the Wnt/β-catenin7 signaling pathway and the NF-κB pathway8,30. Regarding other cancer-associated signaling pathways, DEHP31 may promote the progression of hepatocellular carcinoma through activation of the JAK2/STAT3 signaling pathway. DEHP and its primary metabolite mono(2-ethylhexyl) phthalate (MEHP) have been shown to promote the migration of colorectal cancer cells through activation of the AKT − β-catenin–IL-8 signaling axis32. Furthermore, DEHP may mediate chemoresistance in colorectal cancer via metabolic reprogramming mechanisms32. Evidence suggests that chronic exposure to DEHP/MEHP potentially enhances cancer stemness properties in colorectal malignancies33, which appears to be associated with specific glycosylation alterations in tumor cells. This study further elucidates the mechanisms by which DEHP acts on NSCLC, providing preliminary evidence through network toxicology and molecular docking techniques that DEHP promotes the occurrence, growth, migration, invasion, and drug resistance of NSCLC via the PI3K/AKT pathway and apoptotic pathways. The PI3K/AKT pathway has been extensively studied in lung cancer. Its activation promotes NSCLC cell proliferation and inhibits apoptosis, driving tumor progression34,35. This pathway enhances cancer cell migration and invasion by regulating cytoskeletal remodeling and cell adhesion molecules34. Additionally, it stimulates VEGF expression to promote tumor angiogenesis, supporting nutrient supply. Aberrant PI3K/AKT activation contributes to therapy resistance in NSCLC, including resistance to EGFR inhibitors34. The pathway also modulates the tumor immune microenvironment by influencing immune cell functions, particularly T cell activation/suppression, thereby facilitating immune evasion. However, our KEGG and GO enrichment analyses did not identify NSCLC-specific signaling pathways (e.g., EGFR, ALK), indicating limitations of this approach in detecting context-dependent mechanisms. Future studies should employ transcriptomic profiling to delineate DEHP-responsive pathways uniquely operative in NSCLC pathogenesis. This study establishes a computational framework prioritizing molecular mechanisms linking DEHP exposure to NSCLC progression, while computational limitations—including database gaps in dose-response thresholds and acute/chronic exposure dynamics, compounded by network centrality biases intrinsically prioritizing highly connected hub genes over context-specific mediators like DEHP—highlight actionable refinement pathways. Additionally, molecular docking serves as a tool to simulate target interactions under specific conditions but cannot replicate the actual in vivo environment, including dynamic protein conformations, cellular microenvironments, and competitive endogenous ligand binding. Therefore, systematic experimental validation utilizing physiologically relevant animal models and cell or PDO models in vitro is warranted to address these mechanistic gaps. Such investigations enable simultaneous interrogation of dose-dependent responses across acute and chronic exposure paradigms while recapitulating the multicellular dynamics and metabolic gradients inherent to pulmonary tissue, thereby bridging computational predictions with pathophysiological relevance. Our findings underscore DEHP’s potential as an underrecognized risk factor for NSCLC in humans, particularly given its pervasive exposure through food packaging, medical plastics, and airborne microplastics. Mass (or weight)-based estimates of annual MP(Microplastics) ingestion were reported36 to be 15–287 g/person. Epidemiological evidence elevates median concentrations of pre- and post-shift urinary samples in the exposed workers37 were 16.1 and 55.9 µg/l for MEHP, 37.6 and 103.7 µg/l for MCEPP (MEHP and MCEPP are in vivo metabolites of DEHP and are commonly used as biomarkers to assess DEHP exposure levels in human biomonitoring studies). Children and pregnant women exhibit 3-5fold38 higher DEHP metabolite burdens due to developmental susceptibility, raising concerns about early-life exposure. Current analysis has not established a quantitative model linking external exposure doses to endogenous compound concentrations within tumor tissues, nor does it provide direct evidence to elucidate the dynamic relationship between the expression levels of target proteins in the tumor microenvironment and localized ligand concentrations. Therefore, uncertainties remain in translating target-binding predictions into actual biological interference risks. To mitigate these threats, we urge integrating pathway-specific toxicity data into regulatory frameworks, replacing outdated high-dose animal thresholds. Conduct in-depth investigations into the relationship between intratissue concentrations of DEHP and associated protein targets pre- and post-exposure to elucidate potential mechanistic links. Prioritizing biomonitoring in high-risk cohorts (e.g., industrial workers, urban residents with PM2.5 co-exposure) and establishing DEHP dysregulation as a biomarker-driven risk metric could enable precision prevention strategies, directly addressing DEHP’s emerging role in environment-linked lung carcinogenesis. Conclusion By integrating network toxicology and molecular docking, this study identified 225 targets and five core genes (TP53, JUN, SRC, AKT1, ESR1) mediating DEHP’s toxic effects on NSCLC. KEGG analysis highlighted the PI3K/AKT pathway as the primary mechanism, validated by strong DEHP-target binding affinities. These results demonstrate that DEHP exacerbates NSCLC pathogenesis via PI3K/AKT signaling, promoting tumorigenesis, metastasis, and chemoresistance. This mechanistic insight supports public health interventions to reduce DEHP exposure and guides future research into PI3K/AKT-targeted therapies. Acknowledgements We thank National Natural Science Foundation of China and Shanxi Province Department of Science and Technology, for their funding support. Author contributions ZC and CW wrote the main manuscript text. ZC prepared Figs. 1, 2, 3, 4 and 5. CW, YL and XX prepared Fig. 6. HZ and BH prepared Tables 1, 2 and 3.HB, RZ, GZ and JF reviewed and edited the manuscript. Funding This work was supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project (SQ2023AAA031457), the Project of Traditional Chinese Medicine Administration of Shaanxi Province (2023-ZDYJSL-004), Noncommunicable Chronic Diseases-National Science and Technology Major Project(2024ZD0528600) and Key Research and Development Program of Shaanxi Province (2024SF-ZDCYL-02-09). Data availability The datasets generated and/or analysed during the current study are available in the [Pubchem, CTD, SwissTatgetPrediction, Genecards] repository, [https://pubchem.ncbi.nlm.nih.gov/,https://ctdbase.org/,http://swisstargetprediction.ch/, https://www.genecards.org/] Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Guangjian Zhang, Email: michael8039@xjtu.edu.cn. Jinteng Feng, Email: fengjinteng@xjtu.edu.cn. References Associated Data Data Availability Statement The datasets generated and/or analysed during the current study are available in the [Pubchem, CTD, SwissTatgetPrediction, Genecards] repository, [https://pubchem.ncbi.nlm.nih.gov/,https://ctdbase.org/,http://swisstargetprediction.ch/, https://www.genecards.org/]