PathExpSurv: pathway expansion for explainable survival analysis and disease gene discovery.

BACKGROUND: In the field of biology and medicine, the interpretability and accuracy are both important when designing predictive models. The interpretability of many machine learning models such as neural networks is still a challenge. Recently, many researchers utilized prior information such as biological pathways to develop neural networks-based methods, so as to provide some insights and interpretability for the models. However, the prior biological knowledge may be incomplete and there still exists some unknown information to be explored. RESULTS: We proposed a novel method, named PathExpSurv, to gain an insight into the black-box model of neural network for cancer survival analysis. We demonstrated that PathExpSurv could not only incorporate the known prior information into the model, but also explore the unknown possible expansion to the existing pathways. We performed downstream analyses based on the expanded pathways and successfully identified some key genes associated with the diseases and original pathways. CONCLUSIONS: Our proposed PathExpSurv is a novel, effective and interpretable method for survival analysis. It has great utility and value in medical diagnosis and offers a promising framework for biological research.

ROBO1 deficiency impairs HSPC homeostasis and erythropoiesis via CDC42 and predicts poor prognosis in MDS.

Myelodysplastic syndrome (MDS) is a group of clonal hematopoietic neoplasms originating from hematopoietic stem progenitor cells (HSPCs). We previously identified frequent roundabout guidance receptor 1 (ROBO1) mutations in patients with MDS, while the exact role of ROBO1 in hematopoiesis remains poorly delineated. Here, we report that ROBO1 deficiency confers MDS-like disease with anemia and multilineage dysplasia in mice and predicts poor prognosis in patients with MDS. More specifically, Robo1 deficiency impairs HSPC homeostasis and disrupts HSPC pool, especially the reduction of megakaryocyte erythroid progenitors, which causes a blockage in the early stages of erythropoiesis in mice. Mechanistically, transcriptional profiling indicates that Cdc42, a member of the Rho-guanosine triphosphatase family, acts as a downstream target gene for Robo1 in HSPCs. Overexpression of Cdc42 partially restores the self-renewal and erythropoiesis of HSPCs in Robo1-deficient mice. Collectively, our result implicates the essential role of ROBO1 in maintaining HSPC homeostasis and erythropoiesis via CDC42.

The bcl6 corepressor mutation regulates the progression and transformation of myelodysplastic syndromes by repressing the autophagy flux.

The occurrence of autophagy dysregulation is vital in the development of myelodysplastic syndrome and its transformation to acute myeloid leukemia. However, the mechanisms are largely unknown. Here, we have investigated the mechanism of the bcl6 corepressor mutation in myelodysplastic syndrome development and its transformation to acute myeloid leukemia. We identified a novel pathway involving histone deacetylase 6 and forkhead box protein O1, which leads to autophagy defects following the bcl6 corepressor mutation. And this further causes apoptosis and cell cycle arrest. The bcl6 corepressor-mutation-repressed autophagy resulted in the accumulation of damaged mitochondria, DNA, and reactive oxygen species in myelodysplastic syndrome cells, which could then lead to genomic instability and spontaneous mutation. Our results suggest that the bcl6 corepressor inactivating mutations exert pro-carcinogenic effects through survival strike, which is only an intermediate process. These findings provide mechanistic insights into the role of the bcl6 corepressor gene in myelodysplastic syndrome.

The non-cell-autonomous function of ID1 promotes AML progression via ANGPTL7 from the microenvironment.

The bone marrow microenvironment (BMM) can regulate leukemia stem cells (LSCs) via secreted factors. Increasing evidence suggests that dissecting the mechanisms by which the BMM maintains LSCs may lead to the development of effective therapies for the eradication of leukemia. Inhibitor of DNA binding 1 (ID1), a key transcriptional regulator in LSCs, previously identified by us, controls cytokine production in the BMM, but the role of ID1 in acute myeloid leukemia (AML) BMM remains obscure. Here, we report that ID1 is highly expressed in the BMM of patients with AML, especially in BM mesenchymal stem cells, and that the high expression of ID1 in the AML BMM is induced by BMP6, secreted from AML cells. Knocking out ID1 in mesenchymal cells significantly suppresses the proliferation of cocultured AML cells. Loss of Id1 in the BMM results in impaired AML progression in AML mouse models. Mechanistically, we found that Id1 deficiency significantly reduces SP1 protein levels in mesenchymal cells cocultured with AML cells. Using ID1-interactome analysis, we found that ID1 interacts with RNF4, an E3 ubiquitin ligase, and causes a decrease in SP1 ubiquitination. Disrupting the ID1-RNF4 interaction via truncation in mesenchymal cells significantly reduces SP1 protein levels and delays AML cell proliferation. We identify that the target of Sp1, Angptl7, is the primary differentially expression protein factor in Id1-deficient BM supernatant fluid to regulate AML progression in mice. Our study highlights the critical role of ID1 in the AML BMM and aids the development of therapeutic strategies for AML.

Supervised learning of high-confidence phenotypic subpopulations from single-cell data.

Accurately identifying phenotype-relevant cell subsets from heterogeneous cell populations is crucial for delineating the underlying mechanisms driving biological or clinical phenotypes. Here, by deploying a learning with rejection strategy, we developed a novel supervised learning framework called PENCIL to identify subpopulations associated with categorical or continuous phenotypes from single-cell data. By embedding a feature selection function into this flexible framework, for the first time, we were able to select informative features and identify cell subpopulations simultaneously, which enables the accurate identification of phenotypic subpopulations otherwise missed by methods incapable of concurrent gene selection. Furthermore, the regression mode of PENCIL presents a novel ability for supervised phenotypic trajectory learning of subpopulations from single-cell data. We conducted comprehensive simulations to evaluate PENCILs versatility in simultaneous gene selection, subpopulation identification and phenotypic trajectory prediction. PENCIL is fast and scalable to analyze 1 million cells within 1 hour. Using the classification mode, PENCIL detected T-cell subpopulations associated with melanoma immunotherapy outcomes. Moreover, when applied to scRNA-seq of a mantle cell lymphoma patient with drug treatment across multiple time points, the regression mode of PENCIL revealed a transcriptional treatment response trajectory. Collectively, our work introduces a scalable and flexible infrastructure to accurately identify phenotype-associated subpopulations from single-cell data.

Targeting UHRF1-SAP30-MXD4 axis for leukemia initiating cell eradication in myeloid leukemia.

Aberrant self-renewal of leukemia initiation cells (LICs) drives aggressive acute myeloid leukemia (AML). Here, we report that UHRF1, an epigenetic regulator that recruits DNMT1 to methylate DNA, is highly expressed in AML and predicts poor prognosis. UHRF1 is required for myeloid leukemogenesis by maintaining self-renewal of LICs. Mechanistically, UHRF1 directly interacts with Sin3A-associated protein 30 (SAP30) through two critical amino acids, G572 and F573 in its SRA domain, to repress gene expression. Depletion of UHRF1 or SAP30 derepresses an important target gene, MXD4, which encodes a MYC antagonist, and leads to suppression of leukemogenesis. Further knockdown of MXD4 can rescue the leukemogenesis by activating the MYC pathway. Lastly, we identified a UHRF1 inhibitor, UF146, and demonstrated its significant therapeutic efficacy in the myeloid leukemia PDX model. Taken together, our study reveals the mechanisms for altered epigenetic programs in AML and provides a promising targeted therapeutic strategy against AML.

Interaction-based transcriptome analysis via differential network inference.

Gene-based transcriptome analysis, such as differential expression analysis, can identify the key factors causing disease production, cell differentiation and other biological processes. However, this is not enough because basic life activities are mainly driven by the interactions between genes. Although there have been already many differential network inference methods for identifying the differential gene interactions, currently, most studies still only use the information of nodes in the network for downstream analyses. To investigate the insight into differential gene interactions, we should perform interaction-based transcriptome analysis (IBTA) instead of gene-based analysis after obtaining the differential networks. In this paper, we illustrated a workflow of IBTA by developing a Co-hub Differential Network inference (CDN) algorithm, and a novel interaction-based metric, pivot APC2. We confirmed the superior performance of CDN through simulation experiments compared with other popular differential network inference algorithms. Furthermore, three case studies are given using colorectal cancer, COVID-19 and triple-negative breast cancer datasets to demonstrate the ability of our interaction-based analytical process to uncover causative mechanisms.

Effect of prior malignancy on the prognosis of gastric cancer and somatic mutation.

BACKGROUND: Cancer survivors have a higher risk of developing secondary cancer, with previous studies showing heterogeneous effects of prior cancer on cancer survivors. AIM: To describe the features and clinical significance of a prior malignancy in patients with gastric cancer (GC). METHODS: We identified eligible patients from the Surveillance, Epidemiology, and End Results (SEER) database, and compared the clinical features of GC patients with/without prior cancer. Kaplan-Meier curves and Cox analyses were used to assess the prognostic impact of prior cancer on overall survival (OS) and cancer-specific survival (CSS) outcomes. We also validated our results in The Cancer Genome Atlas (TCGA) cohort and compared mutation patterns. RESULTS: In the SEER dataset, of the 35492 patients newly diagnosed with GC between 2004 and 2011, 4,001 (11.3%) had at least one prior cancer, including 576 (1.62%) patients with multiple cancers. Patients with a prior cancer history tended to be elderly, with a more localized stage and less positive lymph nodes. The prostate (32%) was the most common initial cancer site. The median interval from initial cancer diagnosis to secondary GC was 68 mo. By using multivariable Cox analyses, we found that a prior cancer history was not significantly associated with OS (hazard ratio [HR]: 1.01, 95% confidence interval [CI]: 0.97-1.05). However, a prior cancer history was significantly associated with better GC-specific survival (HR: 0.82, 95% CI: 0.78-0.85). In TCGA cohort, no significant difference in OS was observed for GC patients with or without prior cancer. Also, no significant differences in somatic mutations were observed between groups. CONCLUSION: The prognosis of GC patients with previous diagnosis of cancer was not inferior to that of primary GC patients.

Identifying phenotype-associated subpopulations by integrating bulk and single-cell sequencing data.

Single-cell RNA sequencing (scRNA-seq) distinguishes cell types, states and lineages within the context of heterogeneous tissues. However, current single-cell data cannot directly link cell clusters with specific phenotypes. Here we present Scissor, a method that identifies cell subpopulations from single-cell data that are associated with a given phenotype. Scissor integrates phenotype-associated bulk expression data and single-cell data by first quantifying the similarity between each single cell and each bulk sample. It then optimizes a regression model on the correlation matrix with the sample phenotype to identify relevant subpopulations. Applied to a lung cancer scRNA-seq dataset, Scissor identified subsets of cells associated with worse survival and with TP53 mutations. In melanoma, Scissor discerned a T cell subpopulation with low PDCD1/CTLA4 and high TCF7 expression associated with an immunotherapy response. Beyond cancer, Scissor was effective in interpreting facioscapulohumeral muscular dystrophy and Alzheimer's disease datasets. Scissor identifies biologically and clinically relevant cell subpopulations from single-cell assays by leveraging phenotype and bulk-omics datasets.

Importance-Penalized Joint Graphical Lasso (IPJGL): differential network inference via GGMs.

MOTIVATION: Differential network inference is a fundamental and challenging problem to reveal gene interactions and regulation relationships under different conditions. Many algorithms have been developed for this problem; however, they do not consider the differences between the importance of genes, which may not fit the real-world situation. Different genes have different mutation probabilities, and the vital genes associated with basic life activities have less fault tolerance to mutation. Equally treating all genes may bias the results of differential network inference. Thus, it is necessary to consider the importance of genes in the models of differential network inference. RESULTS: Based on the Gaussian graphical model with adaptive gene importance regularization, we develop a novel Importance-Penalized Joint Graphical Lasso method (IPJGL) for differential network inference. The presented method is validated by the simulation experiments as well as the real datasets. Furthermore, to precisely evaluate the results of differential network inference, we propose a new metric named APC2 for the differential levels of gene pairs. We apply IPJGL to analyze the TCGA colorectal and breast cancer datasets and find some candidate cancer genes with significant survival analysis results, including SOST for colorectal cancer and RBBP8 for breast cancer. We also conduct further analysis based on the interactions in the Reactome database and confirm the utility of our method. AVAILABILITY AND IMPLEMENTATION: R source code of Importance-Penalized Joint Graphical Lasso is freely available at https://github.com/Wu-Lab/IPJGL. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

[Effect of U2AF1 Mutation to Inflammatory Cytokine Expression in SKM-1 Cells through FOXO3a-Bim Signaling Pathway].

OBJECTIVE: To investigate the effect of U2AF1 gene mutation to inflammatory cytokine in SKM-1 cell of human myelodysplastic syndromes (MDS), and whether the above effects were mediated by FOXO3a-Bim signaling pathway. METHODS: Wide-type U2AF1 and mutant U2AF1 (the serine residue 34 was replaced by phenylalanine, and named as S34F) recombinant expression plasmids were constructed. Lentiviruses were packaged and transfected into SKM-1 cells. The expression of FOXO3a was up-regulated by lentiviruses, and its transfection rate was investigated. The cell proliferation was detected by CCK-8 method. Flow cytometry was used to detect the apoptosis and cycle of the cells. The expression pro-inflammatory cytokine IL-1ß, IL-6, TNF-α and anti-inflammatory cytokine IL-4 were detected by qRT-PCR. FOXO3a, Bim, Bcl-2 and Bax protein expression levels were detected by Western blot. RESULTS: Compared with the control group, the cell apoptosis rate, pro-inflammatory cytokine IL-1ß and TNF-α transcription levels were significantly increased in the S34F group (P<0.05); cell cycle was blocked at the G2 phase; cell proliferation and the anti-inflammatory cytokine IL-4 transcription level were significantly decreased; the expression levels of FOXO3a, Bim and Bax protein were significantly increased (P<0.05); while the expression level of Bcl-2 protein was significantly decreased (P<0.05). The up-regulation of FOXO3a could significantly inhibited the proliferation and increased cell apoptosis of SKM-1 cells with U2AF1 S34F mutation; cell cycle was blocked at the S and G2 phases; the pro-inflammatory cytokine IL-1ß and TNF-α transcription levels were significantly decreased (P<0.05), and the transcription level of anti-inflammatory cytokine IL-4 showed no statistically significant as compared with control group (P>0.05). CONCLUSION: U2AF1 S34F mutation can regulate inflammatory phenotype in SKM-1 cells, which may be mediated through FOXO3a-Bim signaling pathway.

Identification of Common Driver Gene Modules and Associations between Cancers through Integrated Network Analysis.

High-throughput biological data has created an unprecedented opportunity for illuminating the mechanisms of tumor emergence and evolution. An important and challenging problem in deciphering cancers is to investigate the commonalities of driver genes and pathways and the associations between cancers. Aiming at this problem, a tool ComCovEx is developed to identify common cancer driver gene modules between two cancers by searching for the candidates in local signaling networks using an exclusivity-coverage iteration strategy and outputting those with significant coverage and exclusivity for both cancers. The associations of the cancer pairs are further evaluated by Fisher's exact test. Being applied to 11 TCGA cancer datasets, ComCovEx identifies 13 significantly associated cancer pairs with plenty of biologically significant common gene modules. The novel results of cancer relationship and common gene modules reveal the relevant pathological basis of different cancer types and provide new clues to diagnosis and drug treatment in associated cancers.

[Research Progress in the Role of Tim3/galectin-9 in Hematological Malignancy--Review].

The incidence of hematological malignant tumor is increasing year by year, and seriously affecting the human health. In addition to the traditional radiation and chemotherapy, immunotherapy has achieved a certain effect in the treatment of blood tumor, but it is limited by exhaustion of CD8+ T cell. The exhaustion of CD8+ T cells is mainly related to the activation of some immune checkpoint inhibitors, such as (Tim3), programmed death 1 (PD-1), cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), etc. Among them, Tim3 is getting more and more attention. The combination of Tim3 and its ligand galectin-9 produced a strong cellular immunosuppressive effect. Tim3 has been proved to be associated with CD8+ T cell exhaustion in many tumors. In this review, the application of Tim3/galectin-9 in hematologic tumors is briefly summarized so as to provide theoretical basis for clinical diagnosis and treatment of these diseases.

Dynamics of epigenetic regulator gene BCOR mutation and response predictive value for hypomethylating agents in patients with myelodysplastic syndrome.

BACKGROUND: BCOR (BCL6 corepressor) is an epigenetic regulator gene involved in the specification of cell differentiation and body structure development. Recurrent somatic BCOR mutations have been identified in myelodysplastic syndrome (MDS). However, the clinical impact of BCOR mutations on MDS prognosis is controversial and the response of hypomethylating agents in MDS with BCOR mutations (BCORMUT) remains unknown. RESULTS: Among 676 MDS patients, 43 patients (6.4%) harbored BCOR mutations. A higher frequency of BCOR mutations (8.7%) was investigated in patients with normal chromosome, compared to 4.2% in patients with abnormal karyotype (p = 0.040). Compared to the BCORWT patients, the BCORMUT patients showed a higher ratio of refractory anemia with excess blasts subset (p = 0.008). The most common comutations with BCOR genes were ASXL1 (p = 0.002), DNMT3A (p = 0.114) and TET2 (p = 0.148). When the hierarchy of somatic mutations was analyzed, BCOR mutations were below the known initial mutations (ASXL1 or TET2) but were above U2AF1 mutations. Transformation-free survival was significantly shorter in BCORMUT patients than that in BCORWT patients (16 vs. 35 months; p = 0.035). RNA-sequencing was performed in bone marrow mononuclear cells from BCORMUT and BCORWT patients and revealed 2030 upregulated and 772 downregulated genes. Importantly, HOXA6, HOXB7, and HOXB9 were significantly over-expressed in BCORMUT patients, compared to BCORWT patients. Eight of 14 BCORMUT patients (57.1%) achieved complete remission (CR) with decitabine treatment, which was much higher than that in BCORWT patients (28.7%, p = 0.036). Paired sequencing results (before and after decitabine) showed three of 6 CR patients lost the mutated BCOR. The median survival of CR patients with a BCORMUT was 40 months, which was significantly longer than that in patients with BCORWT (20 months, p = 0.036). Notably, prolonged survival was observed in three BCORMUT CR patients even without any subsequent therapies. CONCLUSIONS: BCOR mutations occur more frequently in CN MDS patients, predicting higher risk of leukemia transformation. BCORMUT patients showed a better response to decitabine and achieved longer post-CR survival.

A genetic development route analysis on MDS subset carrying initial epigenetic gene mutations.

MDS development is a dynamic process during which the accumulation of somatic mutations leads to specific malignant evolution. To elucidate the differential roles of gene mutations in typical MDS, we used targeted sequencing to investigate clonal patterns from 563 patients and focused on cases (199/563 cases) with initial mutations (ASXL1, DNMT3A and TET2) at MDS diagnosis. The consistency of frequency and distribution in patients with or without aberrant chromosomes suggested early events of these initial mutations. Some additional driver mutations (SF3B1, U2AF1 or RUNX1) played roles to keep the basic disease features, or give rise to different phenotypes (BCOR, EZH2 or TP53) in individual patients. Notably, analysis in paired samples before and after MDS progression showed that the mutations identified as last events (involving active signaling, myeloid transcription or tumor suppressor) seemed necessary for MDS development to be AML. Last mutations can exist at MDS diagnosis, or emerge at AML transformation, and involve a small group of genes. Single-allele CEBPA mutations and diverse TP53 mutations were checked as the most common last event mutations. Considering the necessity of last event mutations and limited gene involvement in AML transformations, it is possible to validate a small group of last events involved mutations to develop some new strategies to block MDS progression.

Recombinant OX40 attenuates neuronal apoptosis through OX40-OX40L/PI3K/AKT signaling pathway following subarachnoid hemorrhage in rats.

Subarachnoid hemorrhage (SAH) is the most devastating form of stroke. Reducing neuronal apoptosis is an important countermeasure against early brain injury (EBI) after SAH. Recent evidence indicates that OX40-OX40L coupling is critical for cell survival and proliferation. Current study was performed to detect the role of recombinant OX40 (ReOX40) against neuronal apoptosis after SAH. The endovascular perforation model of SAH was performed on Sprague-Dawley (SD) rats. ReOX40 was injected intracerebroventricularly (i.c.v) 1 h after SAH induction and the following methods were employed: neurological function evaluation, immunofluorescence staining, fluoro-Jade C staining, and western blot. To study the underlying precise molecular mechanism, small interfering ribonucleic acid (siRNA) for OX40L and a specific inhibitor of PI3K, LY294002, were injected i.c.v. into SAH + ReOX40 rats before induction of SAH. When compared with sham rats, the expression of OX40 and OX40L was seen to decrease in the brain at 24 h after SAH induction. Administration of ReOX40 (5 µg/kg) increased expression of the OX40L, reduced the neuronal apoptosis, and improved short and long-term neurological function deficits. Furthermore, ReOx40 heightened activation of OX40L/PI3K/AKT axis, increased the downstream anti-apoptotic protein (Bcl2, Bcl-XL), and depressed the apoptotic protein (cleaved caspase 3, Bax). However, the protective effects of ReOX40 were abolished by the administration of OX40L siRNA and LY294002, respectively. These results demonstrate that ReOX40 attenuates neuronal apoptosis through OX40-OX40L/PI3K/AKT pathway in EBI after SAH.

Integration Analysis of JAK2 or RUNX1 Mutation With Bone Marrow Blast Can Improve Risk Stratification in the Patients With Lower Risk Myelodysplastic Syndrome.

Despite the improvements in prognostication of the revised International Prognostic Scoring System (IPSS-R) in myelodysplastic syndrome (MDS), there remain a portion of patients with lower risk (low/intermediate risk, LR) but poor prognostics. This study aimed to evaluate the relative contribution of mutational status when added to the IPSS-R, for estimating overall survival (OS) and progression-free survival (PFS) in patients with LR-MDS. We retrospectively analyzed clinical and laboratory variables of 328 patients diagnosed with MDS according to the FAB criteria. Twenty-nine-gene NGS assay was applied to bone marrow samples obtained at diagnosis. 233 (71.04%) patients were classified as LR-MDS. Univariate analysis showed association between inferior outcome (OS and PFS) and presence of JAK2 (p = 0.0177, p = 0.0002), RUNX1 (p = 0.0250, p = 0.0387), and U2AF1 (p = 0.0227, p = 0.7995) mutations. Multivariable survival analysis revealed JAK2 (p < 0.0001) and RUNX1 (p = 0.0215) mutations were independently prognostic for PFS in LR-MDS. Interestingly, bone marrow blast >1.5% could further predict disease progression of patients with LR-MDS (HR 8.06, 95%CI 2.95-22.04, p < 0.0001). Incorporation of JAK2, RUNX1 mutation and bone marrow blast in the IPSS-R can improve risk stratification in patients with LR-MDS. In summary, our result provided new risk factors for LR-MDS prognostics to identify candidates for early therapeutic intervention.

Osteopontin attenuates early brain injury through regulating autophagy-apoptosis interaction after subarachnoid hemorrhage in rats.

AIM: To determine the effect of osteopontin (OPN) on autophagy and autophagy-apoptosis interactions after SAH. METHODS: The endovascular perforation model of SAH or sham surgery was performed in a total of 86 Sprague-Dawley male rats. The temporal expressions of endogenous OPN and autophagy-related proteins (Beclin 1, ATG5, LC3 II to I ratio) were measured in sham and SAH rats at different time points (3, 6, 12, 24, and 72 hours). Rats were randomly divided into three groups: Sham, SAH + Vehicle (PBS, phosphate-buffered saline), and SAH + rOPN (5 µg/rat recombinant OPN). Neurobehavioral tests were performed 24 hours after SAH, followed by the collection of brain samples for assessment of autophagy and apoptosis proteins. These tests assessed whether an autophagy-apoptosis relationship existed on the histological level in the brain. RESULTS: Endogenous OPN and autophagy-related proteins all increased after SAH. rOPN administration improved neurological dysfunction, increased the expression of autophagy-related proteins (Beclin 1, ATG5, LC3 II to I ratio) and antiapoptotic protein Bcl-2, while decreasing the expression of proapoptotic proteins (cleaved Caspase-3 and Bax). rOPN also regulated autophagy-apoptosis interactions 24 hours after SAH. CONCLUSION: rOPN attenuates early brain injury and inhibits neuronal apoptosis by activating autophagy and regulating autophagy-apoptosis interactions.

Discovering cooperative biomarkers for heterogeneous complex disease diagnoses.

Biomarkers with high reproducibility and accurate prediction performance can contribute to comprehending the underlying pathogenesis of related complex diseases and further facilitate disease diagnosis and therapy. Techniques integrating gene expression profiles and biological networks for the identification of network-based disease biomarkers are receiving increasing interest. The biomarkers for heterogeneous diseases often exhibit strong cooperative effects, which implies that a set of genes may achieve more accurate outcome prediction than any single gene. In this study, we evaluated various biomarker identification methods that consider gene cooperative effects implicitly or explicitly, and proposed the gene cooperation network to explicitly model the cooperative effects of gene combinations. The gene cooperation network-enhanced method, named as MarkRank, achieves superior performance compared with traditional biomarker identification methods in both simulation studies and real data sets. The biomarkers identified by MarkRank not only have a better prediction accuracy but also have stronger topological relationships in the biological network and exhibit high specificity associated with the related diseases. Furthermore, the top genes identified by MarkRank involve crucial biological processes of related diseases and give a good prioritization for known disease genes. In conclusion, MarkRank suggests that explicit modeling of gene cooperative effects can greatly improve biomarker identification for complex diseases, especially for diseases with high heterogeneity.