Rigid geometry solves "curse of dimensionality" effects in clustering methods: An application to omics data

Gene Expression Analysis through Parallel Non-Negative Matrix Factorization

Genetic expression analysis is a principal tool to explain the behavior of genes in an organism when exposed to different experimental conditions. In the state of art, many clustering algorithms have been proposed. It is overwhelming the amount of biological data whose high-dimensional structure exceeds mostly current computational architectures. The computational time and memory consumption optimization actually become decisive factors in choosing clustering algorithms. We propose a clustering algorithm based on Non-negative Matrix Factorization and K-means to reduce data dimensionality but whilst preserving the biological context and prioritizing gene selection, and it is implemented within parallel GPU-based environments through the CUDA library. A well-known dataset is used in our tests and the quality of the results is measured through the Rand and Accuracy Index. The results show an increase in the acceleration of 6.22× compared to the sequential version. The algorithm is competitive in the biological datasets analysis and it is invariant with respect to the classes number and the size of the gene expression matrix.

Correction: Rigid geometry solves "curse of dimensionality" effects in clustering methods: An application to omics data

[This corrects the article DOI: 10.1371/journal.pone.0179180.].

Visual Analytics for Dimension Reduction and Cluster Analysis of High Dimensional Electronic Health Records

Recent advancement in EHR-based (Electronic Health Record) systems has resulted in producing data at an unprecedented rate. The complex, growing, and high-dimensional data available in EHRs creates great opportunities for machine learning techniques such as clustering. Cluster analysis often requires dimension reduction to achieve efficient processing time and mitigate the curse of dimensionality. Given a wide range of techniques for dimension reduction and cluster analysis, it is not straightforward to identify which combination of techniques from both families leads to the desired result. The ability to derive useful and precise insights from EHRs requires a deeper understanding of the data, intermediary results, configuration parameters, and analysis processes. Although these tasks are often tackled separately in existing studies, we present a visual analytics (VA) system, called Visual Analytics for Cluster Analysis and Dimension Reduction of High Dimensional Electronic Health Records (VALENCIA), to address the challenges of high-dimensional EHRs in a single system. VALENCIA brings a wide range of cluster analysis and dimension reduction techniques, integrate them seamlessly, and make them accessible to users through interactive visualizations. It offers a balanced distribution of processing load between users and the system to facilitate the performance of high-level cognitive tasks in such a way that would be difficult without the aid of a VA system. Through a real case study, we have demonstrated how VALENCIA can be used to analyze the healthcare administrative dataset stored at ICES. This research also highlights what needs to be considered in the future when developing VA systems that are designed to derive deep and novel insights into EHRs.

Development of a Novel Platform of Proteome Profiling Based on an Easy-to-Handle and Informative 2D-DIGE System

Proteome profiling based on two-dimensional (2D)-DIGE might be a useful tool for investigating drug-like compounds and the mode of action of drugs. However, obtaining data for profiling requires high labor costs, and it is difficult to control the reproducibility of spot positions because 2D-DIGE usually requires large-size glass plates and spot alignments are greatly affected by the quality of DryStrips and polyacrylamide gels (PAGs). Therefore, we have developed a novel platform by employing small size DryStrips and PAGs, and an image analysis strategy based on dual correction of spot alignment and volume. Our system can automatically detect a large number of consistent spots through all images. Cytosol fractions of HeLa cells treated with dimethyl sulfoxide (DMSO) or bortezomib were analyzed, 1697 consistent spots were detected, and 775 of them were significantly changed with the treatment. Deviations between different days and lot sets of DryStrips and PAGs were investigated by calculating the correlation coefficients. The mean values of the correlation between days and lot sets were 0.96 and 0.94, respectively. Clustering analysis of all the treatment data clearly separated the DMSO or bortezomib treated groups beyond day deviations. Thus, we have succeeded in developing an easy-to-handle 2D-DIGE system that can be a novel proteome profiling platform.

Development of Orthogonal Linear Separation Analysis (OLSA) to Decompose Drug Effects into Basic Components

Drugs have multiple, not single, effects. Decomposition of drug effects into basic components helps us to understand the pharmacological properties of a drug and contributes to drug discovery. We have extended factor analysis and developed a novel profile data analysis method: orthogonal linear separation analysis (OLSA). OLSA contracted 11,911 genes to 118 factors from transcriptome data of MCF7 cells treated with 318 compounds in a Connectivity Map. Ontology of the main genes constituting the factors detected significant enrichment of the ontology in 65 of 118 factors and similar results were obtained in two other data sets. In further analysis of the Connectivity Map data set, one factor discriminated two Hsp90 inhibitors, geldanamycin and radicicol, while clustering analysis could not. Doxorubicin and other topoisomerase inhibitors were estimated to inhibit Na+/K+ ATPase, one of the suggested mechanisms of doxorubicin-induced cardiotoxicity. Based on the factor including PI3K/AKT/mTORC1 inhibition activity, 5 compounds were predicted to be novel inducers of autophagy, and other analyses including western blotting revealed that 4 of the 5 actually induced autophagy. These findings indicate the potential of OLSA to decompose the effects of a drug and identify its basic components.