Incorporating deep learning and multi-omics autoencoding for analysis of lung adenocarcinoma prognostication

Lung cancer is the most occurring cancer type, and its mortality rate is also the highest, among them lung adenocarcinoma (LUAD) accounts for about 40 % of lung cancer. There is an urgent need to develop a prognosis prediction model for lung adenocarcinoma. Previous LUAD prognosis studies only took single-omics data, such as mRNA or miRNA, into consideration. To this end, we proposed a deep learning-based autoencoding approach for combination of four-omics data, mRNA, miRNA, DNA methylation and copy number variations, to construct an autoencoder model, which learned representative features to differentiate the two optimal patient subgroups with a significant difference in survival (P = 4.08e-09) and good consistency index (C-index = 0.65). The multi-omics model was validated though four independent datasets, i.e. GSE81089 for mRNA (n = 198, P = 0.0083), GSE63805 for miRNA (n = 32, P = 0.018), GSE63384 for DNA methylation (n = 35, P = 0.009), and TCGA independent samples for copy number variations (n = 94, P = 0.0052). Finally, a functional analysis was performed on two survival subgroups to discover genes involved in biological processes and pathways. This is the first study incorporating deep autoencoding and four-omics data to construct a robust survival prediction model, and results show the approach is useful at predicting LUAD prognostication.     

Clinical implications of proliferation activity in T1 or T2 male gastric cancer patients

Proliferation activity has already been established as a prognostic marker or as a marker for anticancer drug sensitivity. In gastric cancer, however, the prognostic significance of proliferation activity is still being debated. Several studies evaluating proliferation activity using Ki-67 have shown controversial results in terms of the relationship between proliferation activity and overall survival (OS) or drug sensitivity in gastric cancer patients. Because cytoskeleton-associated protein 2 (CKAP2) staining has recently been introduced as a marker of proliferation activity, we analyzed 437 gastric cancer tissues through CKAP2 immunohistochemistry, and we evaluated the chromatin CKAP2-positive cell count (CPCC) for proliferation activity. Although the CPCC did not show any significant correlation with OS in the male, female or total number of cases, it did show a significant correlation in the T1 or T2 male patient subgroup, according to log-rank tests (P=0.001) and univariate analysis (P=0.045). Additionally, multivariate analysis with the Cox proportional hazard regression model showed a significant correlation between the CPCC and OS (P=0.039) for the co-variables of age, gender, T stage, N stage, histology, tumor location, tumor size and adjuvant chemotherapy. In male gastric cancer cell lines, faster-growing cancer cells showed higher sensitivity to cisplatin than slow-growing cells. Thus our study indicates that CPCC-measured proliferation activity demonstrates a significantly worse prognosis in T1 or T2 male gastric cancer patients. The CPCC will help to more precisely classify gastric cancer patients and to select excellent candidates for adjuvant chemotherapy, which in turn will facilitate further clinical chemotherapeutic trials.     

High prevalence of TP53 loss and whole-genome doubling in early-onset colorectal cancer

The global incidence of early-onset colorectal cancer (EO-CRC) is rapidly rising. However, the reason for this rise in incidence as well as the genomic characteristics of EO-CRC remain largely unknown. We performed whole-exome sequencing in 47 cases of EO-CRC and targeted deep sequencing in 833 cases of CRC. Mutational profiles of EO-CRC were compared with previously published large-scale studies. EO-CRC and The Cancer Genome Atlas (TCGA) data were further investigated according to copy number profiles and mutation timing. We classified colorectal cancer into three subgroups: the hypermutated group consisted of mutations in POLE and mismatch repair genes; the whole-genome doubling group had early functional loss of TP53 that led to whole-genome doubling and focal oncogene amplification; the genome-stable group had mutations in APC and KRAS, similar to conventional colon cancer. Among non-hypermutated samples, whole-genome doubling was more prevalent in early-onset than in late-onset disease (54% vs 38%, Fisher’s exact P = 0.04). More than half of non-hypermutated EO-CRC cases involved early TP53 mutation and whole-genome doubling, which led to notable differences in mutation frequencies between age groups. Alternative carcinogenesis involving genomic instability via loss of TP53 may be related to the rise in EO-CRC.Subject terms: Medical genomics, Colon cancer     

Zinc oxide nanoparticle attenuates chemotherapy resistance by inducing cell stemness progression of colorectal cancer via miR-1321/HIF-2α axis

BackgroundColorectal cancer (CRC) is the most malignant cancer type with high morbidity and mortality worldwide. Developed drug resistance severely affected the prognosis of CRC patients. This work aimed to explore the effects of zinc oxide nanoparticles (ZONs) in chemo-resistant CRC.MethodsWe established Oxaliplatin (Oxa)-resistant CRC cells, and in vitro and in vivo model to evaluate the function effect of ZONs. Cell counting kit-8 (CCK-8), colony formation, and 5-ethynyl-2′-deoxyuridine (EDU) assay were performed to detect CRC cell viability and proliferation. CRC cell apoptosis was checked by flow cytometry. Tumor cell proliferation was checked by immunohistochemistry (IHC). Cell stemness was determined by sphere formation assay. Luciferase reporter gene assay was conducted to assess the binding of miR-1321 and HIF-2α 3′UTR region.ResultsZONs suppressed the viability and proliferation of Oxa-resistant CRC cells both in vitro and in vivo. ZONs suppressed CRC cell stemness and enhanced the sensitivity of CRC cells to chemo-therapy, along with decreased expression of stem cell biomarkers. ZONs elevated level of miR-1321 and reduced level of HIF-2α in CRC cells. MiR-1321 targeted the 3′UTR region of HIF-2α to suppress its expression. ZONs repressed HIF-2α expression by inducing miR-1321 in CRC cells.ConclusionZONs treatment remarkably converted the drug resistance and stemness of CRC cells, via upregulating miR-1321 to repress the expression of HIF-2α. Our findings suggested that ZONs are potential and effective agent for chemo-resistant CRC patients.     

Clinicopathological values of NBS1 and DNA damage response genes in epithelial ovarian cancers

Epithelial ovarian cancers (EOCs) are highly lethal gynecological malignancies with a high recurrence rate. Therefore, developing prognostic markers for recurrence after chemotherapy is crucial for the treatment of ovarian cancers. As ovarian cancers frequently respond to DNA-damaging agents, we assessed the clinicopathological significance of key double-strand DNA break (DSB) repair genes, including BRCA1, BRCA2, BARD1, ATM, RAD51 and NBS1 in EOC cell lines and paraffin-embedded tissue sections from 140 EOC patients treated with cytoreductive surgery, followed by platinum-based chemotherapy. These samples were analyzed for the clinicopathological impact of DSB genes by western blot analysis, immunohistochemistry and quantitative real-time PCR. Of the DSB repair genes, BRCA1, ATM and NBS1, which are involved in the homologous recombination-mediated repair pathway, were related to aggressive parameters in EOC. When survival analysis was performed, NBS1 expression exhibited an association with EOC recurrence. Specifically, increased NBS1 expression was found in 107 out of 140 cases (76.0%) and correlated with advanced stage (P=0.001), high grade (P=0.001) and serous histology (P=0.008). The median recurrence-free survival in patients with positive and negative expression of NBS1 was 30 and 78 months, respectively (P=0.0068). In multivariate analysis, NBS1 was an independent prognostic factor for the recurrence of EOC. Together, these results suggest that NBS1 is a marker of poor prognosis for the recurrence of EOC and is associated with aggressive clinicopathological parameters.     

PAK1 as a Potential Therapeutic Target in Male Smokers with EGFR-Mutant Non-Small Cell Lung Cancer

P21-activated kinases (PAKs) are serine/threonine protein kinases that contribute to several cellular processes. Here, we aimed to determine the prognostic value of PAK1 and its correlation with the clinicopathological characteristics and five-year survival rates in patients with non-small cell lung cancer (NSCLC). We evaluated PAK1 mRNA and protein expression in NSCLC cells and resected tumor specimens, as well as in healthy human bronchial epithelial cells and adjacent healthy lung tissues, respectively, for effective comparison. Immunohistochemical tissue microarray analysis of 201 NSCLC specimens showed the correlation of PAK1 expression with clinicopathological characteristics. The mRNA and protein expression of PAK1 were 2.9- and 4.3-fold higher in six of seven NSCLC cell types and human tumors (both, p < 0.001) than in healthy human bronchial epithelial BEAS-2B cells and adjacent healthy lung tissues, respectively. Decreased survival was significantly associated with PAK1 overexpression in the entire cohort (χ² = 8.48, p = 0.0036), men (χ² = 17.1, p < 0.0001), and current and former smokers (χ² = 19.2, p < 0.0001). Notably, epidermal growth factor receptor (EGFR) mutation-positive lung cancer patients with high PAK1 expression showed higher mortality rates than those with low PAK1 expression (91.3% vs. 62.5%, p = 0.02). Therefore, PAK1 overexpression could serve as a molecular target for the treatment of EGFR mutation-positive lung cancer, especially among male patients and current/former smokers.     

Bipartite network analysis reveals metabolic gene expression profiles that are highly associated with the clinical outcomes of acute myeloid leukemia

Dysregulated and reprogrammed metabolism are one of the most important characteristics of cancer, and exploiting cancer cell metabolism can aid in understanding the diverse clinical outcomes for patients. To investigate the differences in metabolic pathways among patients with acute myeloid leukemia (AML) and differential survival outcomes, we systematically conducted microarray data analysis of the metabolic gene expression profiles from 384 patients available from the Gene Expression Omnibus and Cancer Genome Atlas databases. Pathway enrichment analysis of differentially expressed genes (DEGs) showed that the metabolic differences between low-risk and high-risk patients mainly existed in two pathways: biosynthesis of unsaturated fatty acids and oxidative phosphorylation. Using the gene-pathway bipartite network, 62 metabolic genes were identified from 272 DEGs involved in 88 metabolic pathways. Based on the expression patterns of the 62 genes, patients with shorter overall survival (OS) durations in the training set (hazard ratio (HR) = 1.58, p = 0.038) and in two test sets (HR = 1.69 and 1.56 and p = 0.089 and 0.029, respectively) were well discriminated by hierarchical clustering analysis. Notably, the expression profiles of ALAS2, BCAT1, BLVRB, and HK3 showed distinct differences between the low-risk and high-risk patients. In addition, models for predicting the OS outcome of AML from the 62 gene signatures achieved improved performance compared with previous studies. In conclusion, our findings reveal significant differences in metabolic processes of patients with AML with diverse survival durations and provide valuable information for clinical translation.     

Noninvasive detection of colorectal cancer by analysis of exhaled breath

There has been growing interest in exhaled breath analysis for cancer screening and disease monitoring; however, limited breath biomarker information exists regarding colorectal cancer (CRC). The objective of this study was to screen for breath biomarkers of CRC. Exhaled breath was collected from 20 CRC patients and 20 healthy controls; subsequently, solid-phase microextraction–gas chromatography/mass spectrometry (SPME-GC/MS) was used to assess the exhaled volatile organic compounds (VOCs) of the study participants. The statistical methods of principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were performed to process the final data. The VOCs in the exhalations of CRC patients exhibited significant differences from the VOCs in the exhalations of healthy controls; in particular, relative to the latter exhalations, the former exhalations contain significantly higher levels of cyclohexanone, 2,2-dimethyldecane, dodecane, 4-ethyl-1-octyn-3-ol, ethylaniline, cyclooctylmethanol, trans-2-dodecen-1-ol, and 3-hydroxy-2,4,4-trimethylpentyl 2-methylpropanoate but significantly lower levels of 6-t-butyl-2,2,9,9-tetramethyl-3,5-decadien-7-yne (P?<?0.05). Analyses of breath VOCs provide a related model of CRC exhalation that could represent an effective and convenient screening method for this disease.

Practical prediction model of the clinical response to programmed death-ligand 1 inhibitors in advanced gastric cancer

The identification of predictive biomarkers or models is necessary for the selection of patients who might benefit the most from immunotherapy. Seven histological features (signet ring cell [SRC], fibrous stroma, myxoid stroma, tumor-infiltrating lymphocytes [TILs], necrosis, tertiary lymphoid follicles, and ulceration) detected in surgically resected tissues (N = 44) were used to train a model. The presence of SRC became an optimal decision parameter for pathology alone (AUC = 0.78). Analysis of differentially expressed genes (DEGs) for the prediction of genomic markers showed that C-X-C motif chemokine ligand 11 (CXCL11) was high in responders (P < 0.001). Immunohistochemistry (IHC) was performed to verify its potential as a biomarker. IHC revealed that the expression of CXCL11 was associated with responsiveness (P = 0.003). The response prediction model was trained by integrating the results of the analysis of pathological factors and RNA sequencing (RNA-seq). When trained with the C5.0 decision tree model, the categorical level of the expression of CXCL11, a single variable, was shown to be the best model (AUC = 0.812). The AUC of the model trained with the random forest was 0.944. Survival analysis revealed that the C5.0-trained model (log-rank P = 0.01 for progression-free survival [PFS]; log-rank P = 0.012 for overall survival [OS]) and the random forest-trained model (log-rank P < 0.001 for PFS; log-rank P = 0.001 for OS) predicted prognosis more accurately than the PD-L1 test (log-rank P = 0.031 for PFS; log-rank P = 0.107 for OS).Subject terms: Predictive markers, Predictive markers, Translational research, Predictive medicine     

MiR-330-3p and miR-485-5p as biomarkers for glioblastoma: An integrated bioinformatics and experimental study

Glioblastoma Multiforme (GBM) is the most common, invasive, and malignant primary brain tumor with a poor prognosis and a median survival of 12–15 months. This study tried to identify the most significant miRNA biomarkers in both tissue and serum samples of GBM. GSE25632 was employed from gene expression omnibus and using WGCNA package, association of miRNA networks and clinical data was explored and brown and green modules identified as the most relevant modules. Independently, Limma package was utilized to identify differentially expressed miRNAs (DEMs) in GSE25632 by cutoff $\left|\text{l}\text{o}\text{g}\text{F}\text{C}\right|$ > 2 and P.value < 0.05. By merging the results of Limma and WGCNA, the miRNAs that were in brown and green modules and had mentioned cutoff were selected as hub miRNAs. Performing enrichment analysis, Pathways in cancer, Prostate cancer, Glioma, p53 signaling pathway, and Focal adhesion were identified as the most important signaling pathways. Based on miRNA- target genes, has-mir-330−3p and has-mir-485−5p were identified as core miRNAs. The expression level of core miRNAs was validated by GSE90604, GSE42657, and GSE93850. We evaluated the expression level of common target genes of two detected core genes based on GSE77043, GSE42656, GSE22891, GSE15824, and GSE122498. The ability of detected miRNAs to discriminate GBM from healthy controls was assessed by area under the curve (AUC) using the ROC curve analysis. Based on TCGA database, we tested the prognostic significance of miRNAs using overall survival analysis. We evaluated the expression level of the miRNAs in tissue of 83 GBM patients and also non-tumoral adjacent (as control) tissues. We used serum samples of 34 GBM patients to evaluate the expression levels of the hub miRNAs compare to the controls. Our results showed that has-mir-330−3p and has-mir-485−5p could be potential biomarkers in GBM.     

Identification of key genes and expression profiles in osteoarthritis by co-expressed network analysis

BackgroundThe underlying molecular characteristics of osteoarthritis (OA), a common age-related joint disease, remains elusive. Here, we aimed to identify potential early diagnostic biomarkers and elucidate underlying mechanisms of OA using weighted gene co-expression network analysis (WGCNA).Material and methodsWe obtained the gene expression profile dataset GSE55235, GSE55457, and GSE55584, from the Gene Expression Omnibus. WGCNA was used to investigate the changes in co-expressed genes between normal and OA synovial membrane samples. Modules that were highly correlated to OA were subjected to functional enrichment analysis using the R clusterProfiler package. Differentially expressed genes (DEGs) between the two samples were screened using the “limma” package in R. A Venn diagram was constructed to intersect the genes in significant modules and DEGs. RT -PCR was used to further verify the hub gene expression levels between normal and OA samples.ResultsThe preserved significant module was found to be highly associated with OA development and progression (P < 1e-200, correlation = 0.92). Functional enrichment analysis suggested that the antiquewhite4 module was highly correlated to FoxO signaling pathway, and the metabolism of fatty acids and 2-oxocarboxylic acid. A total of 13 hub genes were identified based on significant module network topology and DEG analysis, and RT-PCR confirmed that these genes were significantly increased in OA samples compared with that in normal samples.ConclusionsWe identified 13 hub genes correlated to the development and progression of OA, which may provide new biomarkers and drug targets for OA.     

Gastric cancer in proximal site exerts poorer survival outcome with divergent genetic features than distal site

BackgroundAnatomical subsites always harbor specific biological features in carcinogenesis. The divergent prognosis of proximal gastric cancer (PGC) and distal gastric cancer (DGC) has been reported. The current study aimed to comprehensively interpret anatomic subsite-specific genomic profiles, which may improve the effectiveness of personalized management.MethodsSurvival and genomic data from the online Surveillance, Epidemiology, and End Results (SEER) and The Cancer Genome Atlas (TCGA) databases were queried for prognostic and genetic analysis, respectively. Propensity score matching (PSM) analysis was performed to balance patient epidemiological factors. Differentially expressed genes (DEGs) were analyzed using the DESeq algorithm. Functional enrichment was performed by the clusterProfiler package. The protein-protein interaction network of DEGs was predicted by the online STRING database.ResultsA total of 3,955 patient pairs were assembled by PSM in SEER data with even background characteristics. Prognostic analysis indicated worse overall survival of PGC than DGC (17 vs 20 months, p = 0.0002). Genetic analysis of TCGA database identified 280 DEGs, 90 of which were upregulated in the DGC group and the remaining 190 were upregulated in the PGC group. Functional enrichment analysis indicated that kallikrein serine protease activity, ion channel (Na⁺/Cl⁻) activity, and cytoskeleton constituent might be attributed to the poor prognosis observed in PGC. Furthermore, alcohol, retinol, and lipoprotein metabolism were the features for DGC malignancy.ConclusionThe current study first demonstrated that PGC exerts poorer survival outcome than DGC based on the SEER database. Further bioinformatic investigation depicts the specific genetic features for PGC and DGC, which may generate differences in tumor malignancy. Our findings provide promising genetic targets for future specific and individualized gastric cancer therapy.     

Effect of Extracellular Tyrosinase on the Expression Level of P450, Fpr,and Fdx and Ortho-hydroxylation of Daidzein in <Emphasis Type="Italic">Streptomyces avermitilis</Emphasis>

We have reported that the expression of CYP105D7 in Streptomyces avermitilis produces 112.5 mg L^?1 of 7,3′,4′-trihydroxyisoflavone (3'ODI) in 15 h of the reaction time, when 7,4′-dihydroxyisoflavone (daidzein) is used as a substrate. Although production is significant, rapid degradation of 3'ODI after 15 h was observed in a whole-cell biotransformation system, suggesting the further modification of 3'ODI by endogenous enzymes. In this present study, the effect of deletion of extracellular tyrosinase (melC2) in S. avermitilis for 3'ODI production as well as the expressions of CYP105D7, ferredoxin (Fdx), and ferredoxin reductase (Fpr) were investigated. The result revealed that daidzein hydroxylation activity in the ?melC2 mutant decreased by 40% compared with wild-type S. avermitilis. Further, melC2 deletion significantly affects the messenger RNA (mRNA) expression profile of CYP105D7 and its electron transfer counterparts. Real-time PCR analysis of 9 Fdx, 6 Fpr, and CYP105D7 revealed a significant decrease in mRNA expression level compared to wild-type S. avermitilis. The result clearly shows that the decrease in daidzein hydroxylation activity is due to the lower expression level of CYP105D7 and its electron transfer counterpart in the ?melC2 mutant. Furthermore, melC2 deletion prevents the degradation of 3'ODI.     

A comprehensive profile of DNA copy number variations in a Korean population: identification of copy number invariant regions among Koreans

To examine copy number variations among the Korean population, we compared individual genomes with the Korean reference genome assembly using the publicly available Korean HapMap SNP 50 k chip data from 90 individuals. Korean individuals exhibited 123 copy number variation regions (CNVRs) covering 27.2 mb, equivalent to 1.0% of the genome in the copy number variation (CNV) analysis using the combined criteria of P value (P < 0.01) and standard deviation of copy numbers (SD ≥ 0.25) among study subjects. In contrast, when compared to the Affymetrix reference genome assembly from multiple ethnic groups, considerably more CNVRs (n = 643) were detected in larger proportions (5.0%) of the genome covering 135.1 mb even by more stringent criteria (P < 0.001 and SD ≥ 0.25), reflecting ethnic diversity of structural variations between Korean and other populations. Some CNVRs were validated by the quantitative multiplex PCR of short fluorescent fragment (QMPSF) method, and then copy number invariant regions were detected among the study subjects. These copy number invariant regions would be used as good internal controls for further CNV studies. Lastly, we demonstrated that the CNV information could stratify even a single ethnic population with a proper reference genome assembly from multiple heterogeneous populations.     

Toll-like Receptors Gene Expression in the Gastrointestinal Tract of Salmonella Serovar Pullorum-Infected Broiler Chicken

Salmonella enterica serovar Pullorum causes substantial mortality in chicks as well as results in persistent infection and vertical transmission in layer birds. An effective innate immune response in the early stages of infection could reduce bacterial colonization and mortality in chicks and persistency of infection in later stages. Toll-like receptors (TLRs), important components of innate immune response, plays a pivotal role in early recognition of pathogen as well as in the initiation of robust and specific adaptive immune response. In the present study, we quantified the expression levels of chicken TLRs (1LA, 1LB, 2A, 2B, 3, 4, 5, 7, 15, and 21) mRNA by quantitative real-time PCR in the gastrointestinal (GI) tissues (duodenum, jejunum, ileum, and cecum) of 3-day-old broiler chicks after 24 h of oral infection with S. enterica serovar Pullorum. We found significant upregulation of TLRs (TLR2, TLR4, TLR21) mRNA expressions in GI tract tissues after S. Pullorum infection. The exceptions were for TLR3 and TLR15 with decrease in the expression levels in the jejunum after infection. TLR4 gene expression was significantly (P?<?0.05) upregulated in the duodenum and ileum of infected chicks. Gene expression for some of the TLRs (TLR1LA, ILB, 2B, and TLR5) remained unchanged after infection with S. Pullorum in all the GI tissues studied. Most substantial change in gene expression was found for TLR21, being significantly (P?<?0.05) upregulated in all the tissues investigated. The differential expression levels of TLRs shed light on tailored innate immune response induced by S. Pullorum during the early stages of infection in chicks.     

Overproduction of Polygalacturonase by Penicillium griseoroseum Recombinant Strains and Functional Analysis by Targeted Disruption of the pgg2 Gene

Inactivation of the pgg2 gene, a polygalacturonase-encoding gene from Penicillium griseoroseum, reduced the total activity of polygalacturonase (PG) by 90 % in wild-type P. griseoroseum, which indicates that the pgg2 gene is the major gene responsible for PG production in this species. To increase PG production, the coding region of the pgg2 gene was cloned under the control of the glyceraldehyde 3-phosphate dehydrogenase (gpd) promoter and the terminator region of the tryptophan synthase (trpC) gene from Aspergillus nidulans (pAN52pgg2 vector). This vector was then used to transform P. griseoroseum. The transformed strains were characterized according to PG production using glucose, sucrose, or sugar cane juice as the carbon sources. The recombinant P. griseoroseum T146 strain contained an additional copy of the pgg2 gene, which resulted in a 12-fold increase in PG activity when compared with that detected in the supernatant of the control PG63 strain. The proteins secreted by the recombinant strain T146 showed a strong band at 38 kDa, which corresponds to the molecular weight of PG of the P. griseoroseum. The results demonstrate the significant biotechnological potential of recombinant P. griseoroseum T146 for use in PG production.     

Kallikrein-11, in Association with Coiled-Coil Domain Containing 25, as a Potential Prognostic Marker for Cholangiocarcinoma with Lymph Node Metastasis

Cholangiocarcinoma (CCA) is a malignancy arising from cholangiocytes. Currently, the treatment and prognosis for CCA are mostly poor. Recently, we have reported that coiled-coil domain containing 25 (CCDC25) protein level in the sera may be a diagnostic marker for CCA. Subsequently, we identified three binding proteins of CCDC25 and found that kallikrein-11 (KLK11) expression was highest among those binding proteins. In this study, we investigated CCDC25 and KLK11 expression in CCA and adjacent normal tissues (n = 18) using immunohistochemistry. The results demonstrated that the expressions of CCDC25 and KLK11 in CCA tissues were both significantly higher than the adjacent tissues (p < 0.001 and p = 0.001, respectively). Then, using GEPIA bioinformatics analysis, KLK11 mRNA was significantly overexpressed in CCA tumor tissues compared with normal tissues (p < 0.05). Moreover, CCDC25 expression was positively correlated with KLK11 expression in CCA with lymph node metastasis (p = 0.028, r = 0.593). An analysis for the interaction of KLK11 with CCDC25 and other proteins, using STRING version 11.0, revealed that CCDC25 and KLK11 correlated with metastasis-related proteins. In addition, Kaplan-Meier survival curve analysis revealed that a high expression of KLK11 was associated with the poor prognosis of CCA. In conclusion, KLK11 is, as a binding protein for CCDC25, possibly involved in the metastatic process of CCA. KLK11 may be used as a prognostic marker for CCA.