Transcriptomic data were obtained from ACRG, TCGA and CPTAC cohorts and proteomics data were obtained from PKU and CPTAC cohorts. Phosphoproteome were mainly derived from CPTAC cohort. A total of 142 samples from ACRG cohort, 68 samples from TCGA cohort, 74 samples from CPTAC cohort and 82 samples from PKU cohort were included in this study. DGC patients were divided into ITGB1 high expression and low expression subgroup based on the median expression (RNA or protein level) value as cutoff. Somatic mutations (MuTect2), CNV data (SNP array) and clinical data of the selected DGC cohort were extracted from TCGA cohort (https://portal.gdc.cancer.gov). The GISTIC score and gene copy number amplification and deletion data for each sample were analyzed by GISTIC 2.0 software and plotted by maftools package (54, 55). All extracted DNA and RNA for sequencing were obtained from primary untreated tumor tissues. Association of ITGB1 subgroup with clinical information, including age, gender, clinical stage, overall survival, and molecular subtype, was also collected from these studies and is presented in Table S1 . For proteomics and phosphoproteomics, iTRAQ-labeld LC-MS/MS analysis was performed, and the relative protein abundance of each gene was log2 transformed and centered at zero to obtain the final relative abundance values, which were constructed into an expression matrix. Differential analysis and enrichment analysis were subsequently performed according to different subgroups of ITGB1.

Mutational signatures exacted from aggregated DGC samples (n = 68) genomic data by using the previous framework (56). The method based on variant non-negative matrix factorization can automatically find the appropriate number of mutational signatures. The mutation portrait matrix A was divided into two nonnegative matrices W and H. The matrices W represented mutational processes and the matrices H represented the corresponding mutational activities. The rows of matrix A represents the 96 mutational contexts which rooted in combinations of 6 mutational types (i.e., C > A, C > G, C > T, T > A, T > C, and T > G) and their 5′- and 3′-adjacent bases, and the columns represents the DGC samples. The extracted mutational portrait of DGC was compared and annotated by cosine similarity analysis against the COSMIC (57, 58).

The R package limma (59) was used to evaluate the differential expressed genes in DGC samples with ITGB1-low and ITGB1-high groups. Briefly, we inputted the normalized expression data into lmFit and eBayes functions, the differential statistics were calculated by the limma package. Afterwards the logFC produced by limma performed GSEA referring to the KEGG reference gene set (download from MSigDB database v7.1). The enrichment plot obtained from the fast gene set enrichment analysis algorithm was implemented in the Bioconductor R package fgsea.

The variation in biological processes among different ITGB1 expression subgroups were investigated site-centric PTM Signature Enrichment Analysis (PTM-SEA) on phosphoproteomics data sets with the PTM signatures database (PTMsigDB) (60). Kinase-substrate enrichment analysis (KSEA) were performed by KESA App website (https://casecpb.shinyapps.io/ksea/) using phosphosite data according to its manual with the cutoff of p < 0.05 and substrate count more than 1.

The 64 immune and stromal cell types (spanning multiple adaptive and innate immunity cells, hematopoietic progenitors, epithelial cells, and ECM cells) were inferred by the gene signature–based xCell algorithm (61). Gene expression profiles were prepared using standard annotation files, and data were uploaded to the xCell web portal (https://xcell.ucsf.edu/), with the algorithm run using the xCell signature.

Available clinical annotation and expression profile of human gastric cancer cell lines (N=40) were obtained from the Broad Institute Cancer Cell Line Encyclopedia (CCLE) project (https://portals.broadinstitute.org/ccle/). Gene dependency screening system (CERES and RNAi) and drug sensitivity database (GDSC1 and PRISM) were accessed from the dependency map (DepMap) portal (https://depmap.org/portal/).

MKN-45 cells were kindly provided by Key Laboratory for Experimental Teratology of the Ministry of Education, Department of Pathology, School of Basic Medical Sciences, Shandong University. MKN-45 cells were maintained in 1640 medium (Gibco) supplemented with fetal bovine serum (FBS, 10% PAN), penicillin (100U/mL, Thermo Fisher), and streptomycin (100U/mL, Thermo Fisher) and cultured in 95% air and 5% CO₂ at 37°C. The 1% O₂ stimulation was maintaining MKN-45 cells by 1640 medium (Gibco) supplemented with fetal bovine serum (FBS, 10% PAN), penicillin (100U/mL, Thermo Fisher), and streptomycin (100U/mL, Thermo Fisher) and cultured in 94% N₂, 5% CO_2, and 1% O₂ at 37°C.

Lentivirus particles of shNC (Negative control), shITGB1-1, shITGB1-2 (ITGB1 knockdown), Vector (vector control), or ITGB1 (ITGB1 overexpression) for humans were purchased from Genomeditech. MKN-45 cells were infected by 1640 medium (Gibco) supplemented with fetal bovine serum (FBS, 10% PAN), penicillin (100U/mL, Thermo Fisher), lentivirus, polybrene (0.1%) and streptomycin (100U/mL, Thermo Fisher) for 48h and selected with puromycin (0.5μg/ml, MedChemExpress) for 7 days.

Total RNA from cells was extracted by Trizol reagent (Vazyme, China). RNA was reversely transcribed into cDNA by HiScript III RT SuperMix for qPCR (Vazyme, China) following the manufacturer’s instructions. The quantitative real-time polymerase chain reaction (qRT-PCR) method was used to detect mRNA expression levels by the ChamQ Universal SYBR qPCR Master Mix (Vazyme, China) protocol on Applied Biosystems QuantStudio 1 Real Time PCR system (Applied Biosystems, ThermoFish). The primer sequences were used for qRT-PCR as follows: for ITGB1, 5′- GTCGTGTGTGTGAGTGCAAC-3′ (forward), 5′- GCTGGGGTAATTTGTCCCGA′ (reverse). GAPDH was used as reference for mRNA. The relative expression levels of mRNA were calculated by using the 2−ΔΔCt method in which higher 2−ΔΔCt reflects higher expression.

In brief, total cell lysates were prepared with cell lysis buffer. After denaturing via boiling, total protein was quantified using a BCA protein assay kit (Solarbio). Equivalent amounts of protein were separated by SDS-PAGE at 80 V for 2.5h and transfected to PVDF membranes for 1.5h. The membranes were washed using 1% TBST by three cycles of 5 minute after incubation with Primary antibodies targeting ITGB1 (Proteintech, 12594-1-AP), FDX1 (Proteintech, 12592-1-AP), LIAS (Proteintech, 11577-1-AP), DLD (Proteintech, 16431-1-AP), DLST (abcam, ab177934), DLAT (Proteintech, 13426-1-AP), PDHA1 (Proteintech, 18068-1-AP), PDHB (Proteintech, 14744-1-AP), GLS (Proteintech, 29519-1-AP) and β-actin (Proteintech, 20536-1-AP) at 4°C overnight. Then membranes were treated with secondary antibodies (Proteintech, SA00001-2).

Cell proliferation assays were performed with Cell Counting Kit-8 (DojinDo, Japan) according to the manufacturer’s protocol and detected at 0h, 24h, 48h, 72h and 96 h.

Knockdown or forced overexpression of ITGB1 in MKN-45 cells (2×103cells/well) were seeded into 6-well plates and cultured for 1 weeks at 37°C, and the culture medium was replaced with fresh medium every 4 days. Then, cells were washed using PBS, fixed with 4% paraformaldehyde for 30 min at room temperature, and stained with 0.5% crystal violet for 30 min at room temperature. The number of colonies (containing >50 cells) was observed and counted using an optical microscope.

For migration and invasion analysis, cell (migration: 3 × 105/mL, invasion: 5 × 105/mL) suspension (200 μl of serum-free medium) were seeded onto 8-mm Pore Transwell Inserts (Corning) coated with Matrigel for invasion assay, or without Matrigel for migration assay. Lower chambers were filled with complete medium (600 μl). Cells on the Transwell Inserts were then fixed with paraformaldehyde/PBS (4%) for 30 min. Next, fixed cells were stained with hematoxylin solution (Sigma-Aldrich) for 30min. Then microphotograms of the cells migrated onto the lower side of the filter were imaged using a microscope. From the microphotograms, cells that migrated or invaded onto the lower side of the filter were manually counted. Cell numbers were quantified from ten randomly selected fields with the same area.

Knockdown or forced overexpression of ITGB1 in MKN-45 cells were seeded into 6-well plates and allowed to grow until > 95% confluence. And then the cell layer was gently scratched through the central axis using a sterile plastic tip and loose cells were washed away by PBS and the media replaced with serum free media. Quantification of cell motility by measuring the distance between the invading fronts of cells in three random selected microscopic fields (×100) for each condition and time point (0h, 24h, 48h).

Statistical analyses in this study were generated by R-4.0.1. For quantitative data, statistical significance for normally distributed variables was estimated by Student’s 2-tailed t-tests, and non-normally distributed variables were analyzed by the Wilcoxon’s rank-sum test. The χ2 test and Fisher’s exact test were used to analyze contingency tables depending on specific grouping condition. Correlations between two quantitative variables were then analyzed and described using Spearman’s correlation coefficient. Kaplan Meier survival analysis were used to analyze the survival with R survival package (survminer 2.40-1). Cox regression analysis was used to test the association between ITGB1 and survival outcomes, controlling for age, gender, stage. Experiment statistical analyses were carried out using Prism 8 (GraphPad). Data were obtained from at least three independent experiments and for statistical significance were analyzed using Student’s t-test or Wilcoxon’s rank-sum test. P < 0.05 was considered statistically significant.

DGC patients were divided into ITGB1 high and low expression subgroup based on the median expression (RNA or protein level) value as cutoff. In Kaplan-Meier survival analysis, we found that the ITGB1-low subgroup has a better survival outcomes compared with the ITGB1-high subgroup among the DGC from transcriptomic dataset (ACRG cohort: P = 0.00044, Figure 1A ; TCGA cohort: P = 0.048, Figure 1C ; log-rank test) and proteomic dataset (PKU cohort: P = 0.016, Figure 1E ; log-rank test). When controlled for age, gender, and stage in multivariate Cox proportional hazards regression analysis, we found that the ITGB1 remains significantly associated with better prognosis in the three cohorts (ACRG cohort: HR, 1.79 [95%CI, 1.15–2.78], P = 0.010, Figure 1B ; TCGA cohort: HR, 2.64 [95%CI, 1.05–6.60], P = 0.04. Figure 1D ; PKU cohort: HR, 4.98 [95%CI, 1.85–13.41], P = 0.001, Figure 1F ). However, the prognosis between the two ITGB1 subgroups was not significant in the intestinal-type gastric cancer of ACRG and TCGA cohort (ACRG cohort: P = 0.5, Figure S1A ; TCGA cohort: P = 0.18, Figure S1B ; log-rank test, IGC samples was not existed in PKU cohort). In short, ITGB1 could be a potential prognostic indicator for patients with DGC,

To determine the previous findings on the association of ITGB1 with DGC, we further explored the biological mechanism from a cellular perspective. We generated MKN-45 cells with ITGB1 knockdown and overexpression by shRNA (sh-ITGB1) ( Figure 2A ). According to colony formation assay ( Figure 2B ) and cell counting kit 8 (CCK8) assay ( Figure 2C ), knockdown of ITGB1 in MKN-45 cells markedly decreased cell proliferation. While overexpression of ITGB1 markedly increased cell proliferation. In addition, transwell assay ( Figures 2D–G ) and wound healing assay ( Figures 2H, I ) showed that ITGB1 knockdown also significantly decreased the migration and invasion. Overexpression of ITGB1 markedly increased the migration and invasion. These results suggested that ITGB1 played a protumorigenic role in DGC.

Somatic mutational profiles of DGC patients from the ITGB1-low group and the ITGB1-high group studies were analyzed in TCGA cohort. We found that the ITGB1-low group had significantly higher tumor non-silent mutation load, compared with the ITGB1-high group (Wilcoxon rank-sum test, P = 0.0089, Figure S2A ). We next evaluated mutations of individual genes (such as common oncogenic driver mutations, TP53, KRAS, and PTEN; EMT associated genes, CDH1, and CTNNB1) that may be associated with ITGB1 ( Figure 3A ). Samples with ARID1A, MUC6, and COL11A1 mutations were significantly more frequent in ITGB1 low subgroup than high subgroup (Fisher exact test, P < 0.05). Interestingly, mutations in COL11A1, RASA1, PTEN, MCF1, PLB1, and KRAS were only found in ITGB1-low group. Calculating the number of single nucleotide variants in the matrix of 96 possible mutations with trinucleotide background found that predominant mutations in DGC were featured by the C>T transitions at ApCpN trinucleotide sites. Specifically, the C>T transition at ApCpA were highlighted in low ITGB1 subgroup, whereas the T>G transition at GpTpC were elevated in high ITGB1 subgroup ( Figure 3B ), suggested the specific mutational processes operative in ITGB1 subgroup heterogeneity. Subsequently, we analyzed the gene copy number variation of DGC in different ITGB1 expression subgroup. In general, the chromosomal copy number variation (both gain and loss) of the ITGB1-low group was relatively higher than the ITGB1-high group ( Figure 3C ). Focal level SCNAs revealed that the specific cytobands (FDR<0.01) in each ITGB1 subgroup. As shown by genome plot, the cytobands in 7q21.2, 15q26.1 in low ITGB1 score subgroup, and 8p23.1, 8q21.11, 8q24.13, 8q24.21, 19q12 in high ITGB1 subgroup contained the markedly amplified focal regions; cytobands in 3p14.2, 5q12.1, 6p25.3, 9q21.3, 9p23 in low ITGB1 subgroup contained the frequently deleted regions ( Figures 3D , S2B ). To further explore the mutational processes operative in patients with DGC, we extracted the mutation signatures from the mutational profile ( Figure S2C ). The identified four mutational signatures (eg. signatures 1, 2, 3, 4) were re-annotated them against the Catalogue of Somatic Mutations in Cancer (COSMIC-v3) signature nomenclature by using cosine similarity analysis ( Figures S2D , 3E ). We observed that the SBS1 signature consisted largely of C>T transition that was associated with spontaneous or enzymatic deamination of 5-methylcytosine in most cancers. The SBS17b signature was associated with T>G for some unknown reasons. The SBS15 and SBS6 signature was associated with C>T transition due to defective DNA mismatch repair. Meanwhile, we observed that the percentage of mutations of SBS6 signature and SBS15 signature were markedly higher in the ITGB1-low group, while the percentage of mutations of SBS17b was markedly higher in the ITGB1-high group ( Figure 3F ).

To get closer to the most primitive manifestation of the DGC, we analyzed it from the phosphoproteome level. It was well known that phosphorylation was one of the most common and important modification patterns of proteins. In CPTAC database we performed a differential analysis and identify the phosphorylation sites associated with ITGB1 (such as: upregulated: TNS1_s1477, TNS1_s1164, FLNA_s966 and FLNA_t2167; downregulated: GSTA1_s202, GSTA3_s202 and KCNQ1_s27) in DGC ( Figure 4A , Table S2 ). By analysis of ITGB1 combining proteomics and phosphorylation proteomics in the CPTAC and PKU databases (CPTAC: P < 0.01; PKU: P < 0.05), we functional annotated the biological processes associated with ITGB1. In order to further explore the biological significance of ITGB1, we conducted GO enrichment and Metascape analysis in the protein level and phosphorylated protein level. Enriched biological processes summarized that in the phosphorylated protein level, the ITGB1 was concentrated on Actin filament-based process, Signaling by Rho GTPases, Focal adhesion, mRNA processing, and so on ( Figures 4B–D ), and in the protein level, the ITGB1 was characterized by Metabolism of RNA, Translation, Vesicle-mediated transport, Hemostasis, and others ( Figures S3A–C ). Biological processes associated with metabolism raking in the top 50 were Metabolism of RNA, Peptide metabolic process, Cellular amide metabolic process, Selenoamino acid metabolism, and ncRNA metabolic process. By Metascape analysis, we performed protein-protein interaction enrichment analysis (phosphorylated protein level: Figures 4D, E ; protein level: Figures S3D, E ). The network generated by enrichment analysis consisted of a series of protein clusters. Therefore, we divided the set of proteins that physically interacted into 10 sub-clusters based on the MCODE method, and proteins with the same clusters were characterized by the same GO terms and KEGG pathways. Cluster MCODE1 associated with Smooth Muscle Contraction (R-HSA-445355), Structural molecule activity (GO:0005198), Muscle contraction (R-HSA-397014), et.al, which consisted of SRPRA, ITGA1, ALB, et al. Cluster MCODE2 associated with Translation (R-HSA-72766), ECM-receptor interaction (HSA04512), which consisted of SPCS2, EIF3E, DDOST, et al. Cluster MCODE3 associated with Regulation of expression of SLITs and ROBOs (R-HSA-901055), Signaling by ROBO receptors (R-HSA-376176), Cellular responses to stress (R-HSA-2262752), et.al, which consisted of PSMD2, STT3B et al. Then according to the enrichment analysis, we drew a heat map of ITGB1 related biological processes ( Figure S3F ) and genes which targeted by and regulated ITGB1 ( Figures S3G, H ). We found lots of biological processes, target genes, and regulated genes associated with ITGB1 (such as target genes: DLX6, FOXG1, HES4, et.al; regulated genes: IRF9, HIF1A, ERG, et.al).

Meanwhile, in order to find out the association of ITGB1 with phosphorylation-related pathways, we applied the differential phosphorylated protein to perform ssGSEA/post-transcriptional modification (PTM) analysis and composed a heatmap to visualize the relative abundance. We found that the phosphorylation processes of kinase CK2A1/CSNK2A1, CDK2, and U0126 (MEK inhibitor) were upregulated in the ITGB1-low group, and ROCK1, PKCA/PRKCA, PLK1, and Lepin were upregulated in the ITGB1-high group ( Figure 4F ). Kinase-Substrate Enrichment Analysis (KSEA) revealed that some kinases (like ROCK2, CDK5, DMPK, GSK3A, MYLK, CIT, ROCK1, PRKCA, GSK3B, MAPK8, MAPK1, TGFBR2, MAP2K4, PRKG1, AURKB, and DAPK3) expression up-regulated and some kinases (like CLK1 CSNK2A1, CDK7, and CDK2) expression down-regulated ( Figure 4G , Table S3 ). These results further revealed the protein phosphorylation profile underlying ITGB1 dysregulation and provided the comprehensive insights on ITGB1-mediated transcriptional modification.

Previous studies have identified different molecular subtypes nomenclature of DGCs on the basis of transcriptomic and genomic analysis. Here, we also investigated the association of ITGB1 expression subgroup with previous identified clinical and molecular characteristics ( Figures 5A , S4A ). Interestingly, the MSS/TP53+ subtype (ACRG-defined), GI.HM-indel and Immune-C2 (TCGA-defined), RNA2 and metabolism subtype (CPTAC-defined), PX1-cell cycle (PKU-defined) were predominantly enriched in ITGB1 low expression subgroup, whereas EMT subtype (ACRG-defined), GI.GS subtype and Immune-C3 (TCGA-defined), RNA1 and invasion subtype (CPTAC-defined), PX2-EMT subtype (PKU-defined) was strongly enriched in ITGB1 high expression subgroup ( Figure 5A upper panel, Figure S4A ). The differentially expressed RNA in CPTAC, ACRG, and TCGA cohorts were also illustrated ( Figure 5A lower panel and Figure 4B ). The mRNA levels (such as, RAB31, NRP2, ANTXR1, and CLIC4) were upregulated and the mRNA levels (such as, PGC, TFF2, and GKN1) were downregulated in CPTAC, ACRG, and TCGA cohorts ( Table S4 ). To explore the effects of ITGB1 on biological process, we performed GSEA analysis with KEGG database on RNA levels from ACRG, TCGA, and CPTAC cohorts; and protein levels from CPTAC cohorts. GSEA with ring heatmap indicated that ITGB1 high expression was significantly enriched in immune inflammation, cell adhesion and migration, tumorigenesis pathway, whereas the ITGB1 low expression was predominately enriched in cell cycle and DNA repair, metabolism ( Figure 5B ). Interestingly, metabolism pathway related to aerobic respiration (citrate cycle TCA cycle, oxidative phosphorylation, drug metabolism other enzymes) and glycan-related circuits were negatively associated with ITGB1 expression. We further performed the ssGSEA analysis by using the identified immune-oncology signatures curated form Zeng et al. studies (62). ITGB1-low group had a significantly higher enrichment score compared with the ITGB1-high group in metabolism (citric acid cycle, pentose phosphate, pyruvate metabolism, glyoxylate and dicarboxylate metabolism, gluconeogenesis, oxidative phosphorylation, and purine biosynthesis) and DNA Repair (Base Excision Repair). In contrast, the cell migration and cell matrix (EMT, CAF, Pan F-TBRs) and tumor inflammation (Macrophages Bind, Hypoxia, MDSC) had significantly more enrichment in the ITGB1-high group (Wilcoxon rank-sum test, Figure 5C ). These findings were also verified in the ACRG cohort (Wilcoxon rank-sum test, Figure S4C ).

Furthermore, we evaluated (with the xCell algorithm) the abundance of cell subpopulations in the DGC microenvironment using gene expression data. We found that tumor inflammation cells, fibroblasts cells, and endothelial cells (Astrocytes, cDC, Chondrocytes, Fibroblasts, Hematopoietic stem cells (HSC), Megakaryocytes, Mesangial cells, Monocytes, Neurons, Tregs, Smooth muscle) had a more enrichment in the ITGB1-high group. While CD4+ memory T-cells, common lymphoid progenitors (CLP), Epithelial cells, Keratinocytes, Megakaryocyte–erythroid progenitors (MEP), MSC, plasmacytoid dendritic cells (pDC), Plasma cells, pro B-cells, Sebocytes, and Th1 cells had a better enrichment in the ITGB1-low group ( Figure 5D ).

Previous enrichment analysis revealed that ITGB1 was negatively associated with mitochondria TCA metabolism. As one of the metabolism of the most important substance, glucose metabolism is vital for organisms to maintain homeostasis. Recent studies showed that cuproptosis, a novel form of cell death, was based on glucose metabolism, which was characterized by mitochondrial tricarboxylic acid (TCA) cycle and protein lipoylation (47, 48). In consideration of previous findings that the citric acid cycle was higher in the ITGB1-low group, we further explored the association of ITGB1 and cuproptosis signature. Further analysis indicated that cuproptosis signature score was markedly higher in the ITGB1-low group rather than in the ITGB1-high group (Wilcoxon rank-sum test, p = 1.6e-5, Figure 6A ). In Kaplan-Meier survival analysis, we found that the cuproptosis-high group demonstrated a better survival compared with the cuproptosis-low group among the patients with DGC (TCGA cohort: P = 0.036, Figure S5A ; ACRG cohort: P = 0.048, Figure S5B ; log-rank test). We examined the relationship between known cuproptosis-related-genes and ITGB1 through Spearman analysis. A heatmap of the correlation matrix demonstrated that ITGB1 was negatively correlated with DLST, DLAT, FDX1, ATP7B, and PDHA1, but positively correlated with PDHB and GLS ( Figure 6B , Table S5 ). Correlation between FDX1 and ITGB1 was shown in dot plot among ACRG, TCGA and CPTAC cohorts (ACRG cohort: Spearman r = -0.31, P < 0.001, Figure 6C ; TCGA cohort: Spearman r = -0.36, P < 0.001, Figure S5C ; CPTAC cohort: Spearman r = -0.39, P < 0.001; Figure S5D ).

Subsequently, we verified the findings via molecular experiments. Western blotting analysis indicated that the knockdown of ITGB1 in MKN-45 cells increased the protein levels of LIAS, DLD, DLST, DLAT, FDX1, PDHA1 and reduced the protein level of PDHB by stimulating 1% O₂ for 8h and 1uM CuCl₂, 200 nM Elesclomol for 7h ( Figure 6D ). On the contrary, forced overexpression of ITGB1 decreased the protein levels of these genes and increased the protein level of PDHB ( Figure 6E ). These findings suggested that ITGB1 may be involved in cuproptosis in DGC.

We further investigated the cell viability and drug sensitivity in relation to ITGB1 expression with GC cell model. The cancer-dependent score was analyzed using genetic dependency of RNAi and CRISPR screening dataset from DepMap database (https://depmap.org/portal/download/). Median CERES score in these cell lines was -0.27 (CERES below 0.2 means the gene is an essential gene), and most of the RNAi score were below 0 (RNAi approach to 0 means the gene is not an essential gene, Figure S6 ). These findings indicated that ITGB1 could be regarded as an essential gene in GC.

We also explored the potential compounds in treating the ITGB1 high expressed DGC tumors by using the Genomics of Drug Sensitivity in Cancer (GDSC) database ( Figure 7A ) and Profiling Relative Inhibition Simultaneously in Mixtures (PRISM) database ( Figure 7B ). Agents with significant negative correlation between ITGB1 and drugs IC50 were screened across the GC cell lines (Devimistat: Spearman r = -0.706, P = 0.006; CCT.018159: Spearman r = -0.584, P = 0.036; AS605240: Spearman r = -0.557, P = 0.047; Dabrafenib: Spearman r = -0.553, P = 0.049; Telatinib: Spearman r = -0.580, P = 0.018; Fluvastatin: Spearman r = -0.543, P = 0.029; GSK429286A: Spearman r = -0.524, P = 0.037; NVP-BEZ235: Spearman r = -0.523, P = 0.037; Figure 7B , Table S6 ). These potential therapeutic agents were highly negatively correlated with the ITGB1 and may have potential therapeutic implications for patients with high ITGB1.