Entropy-based consensus clustering for patient stratification

Consensus clustering methods are motivated by the success of combining multiple classifiers in many areas. In this paper, graph-based consensus clustering is used to improve the quality of chemical compound clustering by enhancing the robustness, novelty, consistency and stability of individual clusterings. For this purpose, Hyper-Graph Partitioning Algorithm (HGPA) [1], was applied. The clustering is evaluated based on the ability to separate actives from inactives molecules in each cluster and the results were compared with the Ward's clustering method. The chemical dataset MDL Drug Data Report (MDDR) database has been used for experiments. The MDL Drug Data Report (MDDR) database consists of 102516 molecules. For the experiments, the dataset DS1 was chosen from the MDDR database. This dataset has been used for many virtual screening experiments [2-4]. The dataset DS1contains 10 heterogeneous activity classes (8568 molecules). For the clustering experiments, two 2D fingerprint descriptors will be used which are developed by Scitegic's Pipeline Pilot [5]. These are 120-bit ALOGP and 1024-bit extended connectivity fingerprints (ECFP_4). The results were evaluated based on the effectiveness of the methods to separate actives from non-actives molecules using QPI- (for quality partition index) measure, which was devised by Varin et al. [6]. As defined by [7], an active cluster as a non-singleton cluster for which the percentage of active molecules in the cluster is greater than the percentage of active molecules in the dataset as a whole. Let p be the number of actives in active clusters, q the number of inactives in active clusters, r the number of actives in inactive clusters (i.e., clusters that are not active clusters) and s the number of singleton actives. The high value occurs when the actives are clustered tightly together and separated from the inactive molecules. Then the quality partition index, QPI, is defined to be: QPI=pp+q+r+s (1) Then, the results will be compared with Ward's individual clustering method, the standard clustering method for chemoinformatics applications. The generation process has been done by multiple run of K-means algorithms, each with random initialization of cluster centroids. The number of partitions generated in this step is ranged between n = 5 to n = 50, with 5-times step. Then, all the generated partitions were combined using HGPA to obtain the consensus partition. This process is done for each fingerprint (ALOGP and ECFP_4). The mean of QPI values are averaged over the ten activity classes of the datasets. Tables ​Tables1,1, ​,22 show the effectiveness of MDDR dataset clustering using ALOGP and ECFP_4 fingerprints. The best PQI value of consensus clustering methods for each column has been bold-faced for ease of reference. Table 1 Effectivenss of clustering of high diverse MDDR dataset: ALOGP Fingerprint. Table 2 Effectivenss of clustering of high diverse MDDR dataset: ECFP_4 Fingerprint. Visual inspection of the results enables comparisons to be made between the effectiveness of clustering of MDDR datasets and Ward's method, the best of choice clustering method for chemoinformatics applications. In addition, ten times of consensus clustering, for each fingerprint were observed in order to study the effectiveness of consensus clustering with different ensemble sizes. The results show that HGPA consensus clustering gives robust and novel result when K-means algorithm is run 20-50 times using ALOGP. The performance of consensus clustering outperforms the Wards' method. For consensus of dataset which represented by ECFP_4 fingerprint, the best QPI values of consensus clustering are obtained from ensembles of size n = 20-50. The performance of consensus clustering gives robust results which are better than overall performance of individual clusterings. The values of QPI in both datasets for consensus clustering are close to the Wards method. The consensus clustering, HGPA, provide stable clusters by decreasing the sensitivity to noise and outliers. The average percentages of singleton clusters of individual clusterings compared with consensus clustering for both fingerprints. The results show that consensus clustering partition the datasets with average percentage of singleton equal to zero, which is much better than individual clusterings and Wards' method. For example, 16.72% of molecules of DS1 are clustered as singletons when Wards method is applied on ALOGP fingerprint with number of clusters equal to 1000 clusters. Finally we conclude that graph-based consensus clustering can improve the effectiveness of chemical compounds clustering. The performance of consensus clustering is more robust, novel, stable, consistent, and out-perform Wards' method in case of using ALOGP fingerprint. By using ECFP_4 fingerprint, consensus clustering methods provide more robust, stable, and consistent clustering and close to the Wards clustering results. The experiments reported here suggest that graph-based consensus clustering can improve the quality of individual clustering by using the efficient algorithm, K-means algorithm, to generate the ensemble with size (20-50) for both structurally diverse chemical datasets.

Consensus clustering provides a framework to ensemble multiple clustering results to obtain a consensus and robust result. Most existing consensus clustering methods usually apply all data to ensemble learning, whereas ignoring the side effects caused by some difficult or unreliable instances. To tackle this problem, we propose a novel self-paced consensus clustering method to gradually involve instances from more reliable to less reliable ones into the ensemble learning. We first construct an initial bipartite graph from the multiple base clustering results, where the nodes represent the instances and clusters and the edges indicate that an instance belongs to a cluster. Then, we learn a structured bipartite graph from the initial one by self-paced learning, i.e., we automatically decide the reliability of each edge and involves the edges into graph learning in order of their reliability. At last, we obtain the final consensus clustering result from the learned bipartite graph. The extensive experimental results demonstrate the effectiveness and superiority of the proposed method.

Clustering is one of the major tasks in data mining. However, selecting an algorithm to cluster a dataset is a difficult task, especially if there is no prior knowledge on the structure of the data. Consensus clustering methods can be used to combine multiple base clusterings into a new solution that provides better partitioning. In this work, we present a new consensus clustering method based on detecting clustering patterns by mining frequent closed itemset. Instead of generating one consensus, this method both generates multiple consensuses based on varying the number of base clusterings, and links these solutions in a hierarchical representation that eases the selection of the best clustering. This hierarchical view also provides an analysis tool, for example to discover strong clusters or outlier instances.

Hepatocellular carcinoma (HCC) is a highly aggressive tumor characterized by significant heterogeneity and invasiveness, leading to a lack of precise individualized treatment strategies and poor patient outcomes. This necessitates the urgent development of accurate patient stratification methods and targeted therapies based on distinct tumor characteristics.By integrating gene expression data from The Cancer Genome Atlas (TCGA), International Cancer Genome Consortium (ICGC), and Gene Expression Omnibus (GEO), we identified subtypes through a multi-omics consensus clustering approach amalgamated from 10 clustering techniques. Subsequently, we developed a prognostic model, employing machine learning algorithms, based on subtype classification features. Finally, by analyzing single cell sequencing data, we investigated the mechanisms driving prognostic variations among distinct subtypes.First, we developed a novel consensus clustering method that categorizes liver cancer patients into two subtypes, CS1 and CS2. Second, we constructed a prognostic prediction model, which demonstrated superior predictive accuracy compared to several models published in the past five years. Finally, we observed differences between CS1 and CS2 in various metabolic pathways, biological processes, and signaling pathways, such as fatty acid metabolism, hypoxia levels, and the PI3K-AKT signaling pathway.

BackgroundHuman cancers are complex ecosystems composed of cells with distinct molecular signatures. Such intratumoral heterogeneity poses a major challenge to cancer diagnosis and treatment. Recent advancements of single-cell techniques such as scRNA-seq have brought unprecedented insights into cellular heterogeneity. Subsequently, a challenging computational problem is to cluster high dimensional noisy datasets with substantially fewer cells than the number of genes.MethodsIn this paper, we introduced a consensus clustering framework conCluster, for cancer subtype identification from single-cell RNA-seq data. Using an ensemble strategy, conCluster fuses multiple basic partitions to consensus clusters.ResultsApplied to real cancer scRNA-seq datasets, conCluster can more accurately detect cancer subtypes than the widely used scRNA-seq clustering methods. Further, we conducted co-expression network analysis for the identified melanoma subtypes.ConclusionsOur analysis demonstrates that these subtypes exhibit distinct gene co-expression networks and significant gene sets with different functional enrichment.

Background Glioblastoma multiforme (GBM) is a highly aggressive type of brain cancer with short survival time, poor prognosis and high mortality. Although research based on molecular data plays a crucial role in predicting cancer, few studies have been developed to combine high-throughput molecular data with clinical variables. Our goal is to establish machine learning based models to predict GBM survival. Methods The molecular data and clinical information of GBM were downloaded from The Cancer Genome Atlas (TCGA) database. The molecular data include somatic copy-number alteration (SCNA) and microRNA. Logistic regression and support vector machine (SVM) were used to establish the predictive models using molecular data, clinical variables and their combinations. Receiver operating characteristic (ROC) was conducted to estimate the sensitivity and specificity. Results The AUC scores are 0.72, 0.69 and 0.82, respectively using clinical variables, SCNA data and their combinations. The AUC scores are 0.97 and 0.98, respectively using microRNA data and the combinations of microRNA data and clinical variables. Conclusion Our study developed machine learning based models to predict GBM survival. The combination of molecular data and clinical variables can improve prediction accuracy. Moreover, it also brings us new insights into the molecular mechanisms underlying GBM cancer. Acknowledgements Supported by the National Natural Science Foundation of China (61471078, 31270903), the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry, the Program for Liaoning Excellent Talents in University (LJQ2015011) and the Fundamental Research Funds for the Central Universities (3132016330, 3132014306, 3132015213, and 3132017075).

BackgroundPolyamine modification patterns in lung adenocarcinoma (LUAD) and their impact on prognosis, immune infiltration, and anti-tumor efficacy have not been systematically explored.MethodsPatients from The Cancer Genome Atlas (TCGA) were classified into subtypes according to polyamine metabolism-related genes using the consensus clustering method, and the survival outcomes and immune profile were compared. Meanwhile, the geneCluster was constructed according to the differentially expressed genes (DEGs) of the subtypes. Subsequently, the polyamine metabolism-related score (PMRS) system was established using the least absolute shrinkage and selection operator (LASSO) multivariate regression analysis in the TCGA training cohort (n = 245), which can be applied to characterize the prognosis. To verify the predictive performance of the PMRS, the internal cohort (n = 245) and the external cohort (n = 244) were recruited. The relationship between the PMRS and immune infiltration and antitumor responses was investigated.ResultsTwo distinct patterns (C1 and C2) were identified, in which the C1 subtype presented an adverse prognosis, high CD8+ T cell infiltration, tumor mutational burden (TMB), immune checkpoint, and low tumor immune dysfunction and exclusion (TIDE). Furthermore, two geneClusters were established, and similar findings were observed. The PMRS, including three genes (SMS, SMOX, and PSMC6), was then constructed to characterize the polyamine metabolic patterns, and the patients were divided into high- and low-PMRS groups. As confirmed by the validation cohort, the high-PMRS group possessed a poor prognosis. Moreover, external samples and immunohistochemistry confirmed that the three genes were highly expressed in tumor samples. Finally, immunotherapy and chemotherapy may be beneficial to the high-PMRS group based on the immunotherapy cohorts and low half-maximal inhibitory concentration (IC50) values.ConclusionWe identified distinct polyamine modification patterns and established a PMRS to provide new insights into the mechanism of polyamine action and improve the current anti-tumor strategy of LUAD.

Clustering is the process of partitioning a dataset into groups based on the similarity between the instances. Many clustering algorithms were proposed, but none of them proved to provide good quality partition in all situations. Consensus clustering aims to enhance the clustering process by combining different partitions obtained from different algorithms to yield a better quality consensus solution. In this work, we propose a new consensus clustering method that uses a pattern mining technique in order to reduce the search space from instance-based into pattern-based space. Instead of finding one solution, our method generates multiple consensus candidates based on varying the number of base clusterings considered. The different solutions are then linked and presented as a tree that gives more insight about the similarities between the instances and the different partitions in the ensemble.

Atherosclerosis (AS) is the leading cause of coronary heart disease, which is the primary cause of death worldwide. Recent studies have identified disulfidptosis as a new type of cell death that may be involved in onset and development of many diseases. However, the role of disulfidptosis in AS is not clear. In this study, bioinformatics analysis and experiments in vivo and in vitro were performed to evaluate the potential relationship between disulfidptosis and AS. AS-related sequencing data were obtained from the Gene Expression Omnibus (GEO). Bioinformatics techniques were used to evaluate differentially expressed genes (DEGs) associated with disulfidptosis-related AS. Hub genes were screened using least absolute shrinkage and selection operator (LASSO) and random forests (RF) methods. In addition, we established a foam cell model in vitro and an AS mouse model in vivo to verify the expressions of hub genes. In addition, we constructed a diagnostic nomogram with hub genes to predict progression of AS. Finally, the consensus clustering method was used to establish two different subtypes, and associations between subtypes and immunity were explored. As the results, 9 disulfidptosis-related AS DEGs were identified from GSE28829 and GSE43292 datasets. Evaluation of DEGs using LASSO and RF methods resulted in identification of 4 hub genes (CAPZB, DSTN, MYL6, PDLIM1), which were analyzed for diagnostic value using ROC curve analysis and verified in vitro and in vivo. Furthermore, a nomogram including hub genes was established that accurately predicted the occurrence of AS. The consensus clustering algorithm was used to separate patients with early atherosclerotic plaques and patients with advanced atherosclerotic plaques into two disulfidptosis subtypes. Cluster B displayed higher levels of infiltrating immune cells, which indicated that patients in cluster B may have a positive immune response for progression of AS. In summary, disulfidptosis-related genes including CAPZB, DSTN, MYL6, and PDLIM1 may be diagnostic markers and therapeutic targets for AS. In addition, these genes are closely related to immune cells, which may inform immunotherapy for AS.

A central question for the cancer research community is how to use genomic and proteomic data to guide cancer care. We investigated two potential clinical applications of The Cancer Genome Atlas (TCGA) data: predicting patient prognosis and identifying clinically actionable alterations. For four TCGA cohorts, we systematically evaluated the power of molecular data with or without clinical variables in predicting patient survival; and to facilitate further community efforts, we established an open-access model evaluation platform. We demonstrate that incorporating molecular data with clinical variables can improve predictive power. Interestingly, one model derived from ovarian cancer can predict outcomes in kidney cancer, suggesting the potential utility of cross-tumor analysis. Across 12 cancer types, we identified 10,281 somatic alterations in clinically relevant genes in 2,928 out of 3,277 patients (89.4%), revealing recurrent and potentially targetable alterations not revealed in single-tumor datasets. Our study represents the first systematic and comprehensive analysis evaluating the prognostic power of different types of molecular data across multiple cancer types. In addition, the clinically actionable alterations we identified may inform novel clinical trial design and treatment choice. Citation Format: Han Liang, TCGA Pan-Cancer Clinical/Predictor Group. Dissect the clinical utility of TCGA genomic data across tumor types. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4168. doi:10.1158/1538-7445.AM2014-4168

Predicting functions of genes is an important issue in biology. Clustering gene expression profiles has been widely used for gene function prediction, but most clustering methods are unstable and sensitive to input parameters such as starting values and number of clusters. In this article, we develop a novel consensus clustering method to address the instability issue and thus improve the performance of clustering methods. The biological function of an unannotated gene is predicted based on the most enriched functional category in its consensus cluster. The MIPS gene annotations are used to evaluate the predictive performance. It is shown that the consensus clustering-based classification method has a significantly better predictive performance than a previously used clustering-based classification method while performing as well as support vector machines (SVMs). In addition to the obvious applicability of consensus clustering to unsupervised learning, the method's advantages in supervised learning include its being a multiclass classifier that can be trained much faster than SVMs, its generality to include any of the many existing clustering algorithms, and its flexibility to be integrated with other predictive models built with other types of data, suggesting its potential for further improved performance. As a concrete example, we consider its combined use with protein-protein interaction data for gene function prediction. It is shown that the combined analysis has a significantly higher predictive accuracy and a much broader functional coverage than using either data source alone.

The use of consensus clustering methods in chemoinformatics is motivated because of the success of consensus scoring (data fusion) in virtual screening and also because of the ability of consensus clustering to improve the robustness, novelty, consistency and stability of individual clusterings in other areas. In this paper, Cumulative Voting-based Aggregation Algorithm (CVAA) was examined for combining multiple clusterings of chemical structures. The effectiveness of clusterings was evaluated based on the extent to which they clustered compounds, which belong to the same activity class, together. Then, the results were compared to other consensus clustering and Ward's methods. The MDL Drug Data Report (MDDR) database was used for experiments and the results were obtained by combining multiple clusterings that were applied using different distance measures. The experiments show that the voting-based consensus method can efficiently improve the effectiveness of chemical structures clusterings.

ObjectiveCluster analysis of spatio-temporal event-related potential (ERP) data is a promising tool for exploring the measurement time window of ERPs. However, even after preprocessing, the remaining noise can result in uncertain cluster maps followed by unreliable time windows while clustering via conventional clustering methods. MethodsWe designed an ensemble deep clustering pipeline to determine a reliable time window for the ERP of interest from temporal concatenated grand average ERP data. The proposed pipeline includes semi-supervised deep clustering methods initialized by consensus clustering and unsupervised deep clustering methods with end-to-end architectures. Ensemble clustering from those deep clusterings was used by the designed adaptive time window determination to estimate the time window. ResultsAfter applying simulated and real ERP data, our method successfully obtained the time window for identifying the P3 components (as the interest of both ERP studies) while additional noise (e.g., adding 20 dB to −5 dB white Gaussian noise) was added to the prepared data. ConclusionCompared to the state-of-the-art clustering methods, a superior clustering performance was yielded from both ERP data. Furthermore, more stable and precise time windows were elicited as the noise increased. SignificanceOur study provides a complementary understanding of identifying the cognitive process using deep clustering analysis to the existing studies. Our finding suggests that deep clustering can be used to identify the ERP of interest when the data is imperfect after preprocessing.

Endometrial cancer is one of the most common gynecological malignancies, and DNA methylation plays a vital role in its occurrence and development. In this study, we collected the relevant data on endometrial cancer from the Cancer Genome Atlas database and UCSC website. By screening and processing the data, we obtained 410 samples and 16,381 methylation sites. Endometrial carcinoma can be divided into seven molecular subtypes using consensus clustering method. Based on the analysis of the differences among subtypes, the methylation degree of different sites was obtained, and the prognosis model of methylation sites was established. Based on the median value of the train group, the train and test groups were divided into high and low-risk groups. The survival between the high and low-risk groups was different. It also showed that this model can predict the survival of patients, with better accuracy. In conclusion, the tumor subtypes based on methylation sites can provide a better guidance for treatment, relapse, and prognosis of endometrial cancer. In this study, magnetic nanoparticles can be used to extract genomic DNA and total RNA due to their paramagnetism and biocompatibility, then transcriptome high-throughput sequencing was performed. It may serve as potential cancer immune biomarker targets for developing future oncological treatments.

<div>Abstract<p>A third of patients with epithelial ovarian cancer (OVCA) will not respond to standard treatment. The determination of a robust signature that predicts chemoresponse could lead to the identification of molecular markers for response as well as possible clinical implementation in the future to identify patients at risk of failing therapy. This pilot study was designed to identify biologic processes affecting candidate pathways associated with chemoresponse and to create a robust gene signature for follow-up studies. After identifying common pathways associated with chemoresponse in serous OVCA in three independent gene-expression experiments, we assessed the biologic processes associated with them using The Cancer Genome Atlas (TCGA) dataset for serous OVCA. We identified differential copy-number alterations (CNA), mutations, DNA methylation, and miRNA expression between patients that responded to standard treatment and those who did not or recurred prematurely. We correlated these significant parameters with gene expression to create a signature of 422 genes associated with chemoresponse. A consensus clustering of this signature identified two differentiated clusters with unique molecular patterns: cluster 1 was significant for cellular signaling and immune response (mainly cell-mediated); and cluster 2 was significant for pathways involving DNA-damage repair and replication, cell cycle, and apoptosis. Validation through consensus clustering was performed in five independent OVCA gene-expression experiments. Genes were located in the same cluster with consistent agreement in all five studies (κ coefficient ≥ 0.6 in 4). Integrating high-throughput biologic data have created a robust molecular signature that predicts chemoresponse in OVCA. <i>Cancer Res; 74(14); 3902–12. ©2014 AACR</i>.</p></div>