Unsupervised multiple kernel learning for heterogeneous data integration

Link-HD: a versatile framework to explore and integrate heterogeneous microbial communities

MOTIVATION

We present Link-HD, an approach to integrate multiple datasets. Link-HD is a generalization of 'Structuration des Tableaux A Trois Indices de la Statistique-Analyse Conjointe de Tableaux', a family of methods designed to integrate information from heterogeneous data. Here, we extend the classical approach to deal with broader datasets (e.g. compositional data), methods for variable selection and taxon-set enrichment analysis.

RESULTS

The methodology is demonstrated by integrating rumen microbial communities from cows for which methane yield (CH4y) was individually measured. Our approach reproduces the significant link between rumen microbiota structure and CH4 emission. When analyzing the TARA's ocean data, Link-HD replicates published results, highlighting the relevance of temperature with members of phyla Proteobacteria on the structure and functionality of this ecosystem.

AVAILABILITY AND IMPLEMENTATION

The source code, examples and a complete manual are freely available in GitHub https://github.com/lauzingaretti/LinkHD and in Bioconductor https://bioconductor.org/packages/release/bioc/html/LinkHD.html.

Kernel Fusion Method for Detecting Cancer Subtypes via Selecting Relevant Expression Data

Recently, cancer has been characterized as a heterogeneous disease composed of many different subtypes. Early diagnosis of cancer subtypes is an important study of cancer research, which can be of tremendous help to patients after treatment. In this paper, we first extract a novel dataset, which contains gene expression, miRNA expression, and isoform expression of five cancers from The Cancer Genome Atlas (TCGA). Next, to avoid the effect of noise existing in 60, 483 genes, we select a small number of genes by using LASSO that employs gene expression and survival time of patients. Then, we construct one similarity kernel for each expression data by using Chebyshev distance. And also, We used SKF to fused the three similarity matrix composed of gene, Iso, and miRNA, and finally clustered the fused similarity matrix with spectral clustering. In the experimental results, our method has better P-value in the Cox model than other methods on 10 cancer data from Jiang Dataset and Novel Dataset. We have drawn different survival curves for different cancers and found that some genes play a key role in cancer. For breast cancer, we find out that HSPA2A, RNASE1, CLIC6, and IFITM1 are highly expressed in some specific groups. For lung cancer, we ensure that C4BPA, SESN3, and IRS1 are highly expressed in some specific groups. The code and all supporting data files are available from https://github.com/guofei-tju/Uncovering-Cancer-Subtypes-via-LASSO.

Integrated Multi-Omics Analyses in Oncology: A Review of Machine Learning Methods and Tools

In recent years, high-throughput sequencing technologies provide unprecedented opportunity to depict cancer samples at multiple molecular levels. The integration and analysis of these multi-omics datasets is a crucial and critical step to gain actionable knowledge in a precision medicine framework. This paper explores recent data-driven methodologies that have been developed and applied to respond major challenges of stratified medicine in oncology, including patients' phenotyping, biomarker discovery, and drug repurposing. We systematically retrieved peer-reviewed journals published from 2014 to 2019, select and thoroughly describe the tools presenting the most promising innovations regarding the integration of heterogeneous data, the machine learning methodologies that successfully tackled the complexity of multi-omics data, and the frameworks to deliver actionable results for clinical practice. The review is organized according to the applied methods: Deep learning, Network-based methods, Clustering, Features Extraction, and Transformation, Factorization. We provide an overview of the tools available in each methodological group and underline the relationship among the different categories. Our analysis revealed how multi-omics datasets could be exploited to drive precision oncology, but also current limitations in the development of multi-omics data integration.

Heterogeneous multiple kernel learning for breast cancer outcome evaluation

Background

Breast cancer is one of the common kinds of cancer among women, and it ranks second among all cancers in terms of incidence, after lung cancer. Therefore, it is of great necessity to study the detection methods of breast cancer. Recent research has focused on using gene expression data to predict outcomes, and kernel methods have received a lot of attention regarding the cancer outcome evaluation. However, selecting the appropriate kernels and their parameters still needs further investigation.

Results

We utilized heterogeneous kernels from a specific kernel set including the Hadamard, RBF and linear kernels. The mixed coefficients of the heterogeneous kernel were computed by solving the standard convex quadratic programming problem of the quadratic constraints. The algorithm is named the heterogeneous multiple kernel learning (HMKL). Using the particle swarm optimization (PSO) in HMKL, we selected the kernel parameters, then we employed HMKL to perform the breast cancer outcome evaluation. By testing real-world microarray datasets, the HMKL method outperforms the methods of the random forest, decision tree, GA with Rotation Forest, BFA + RF, SVM and MKL.

Conclusions

On one hand, HMKL is effective for the breast cancer evaluation and can be utilized by physicians to better understand the patient’s condition. On the other hand, HMKL can choose the function and parameters of the kernel. At the same time, this study proves that the Hadamard kernel is effective in HMKL. We hope that HMKL could be applied as a new method to more actual problems.

A U-statistics for integrative analysis of multilayer omics data

MOTIVATION

The emerging multilayer omics data provide unprecedented opportunities for detecting biomarkers that are associated with complex diseases at various molecular levels. However, the high-dimensionality of multiomics data and the complex disease etiologies have brought tremendous analytical challenges.

RESULTS

We developed a U-statistics-based non-parametric framework for the association analysis of multilayer omics data, where consensus and permutation-based weighting schemes are developed to account for various types of disease models. Our proposed method is flexible for analyzing different types of outcomes as it makes no assumptions about their distributions. Moreover, it explicitly accounts for various types of underlying disease models through weighting schemes and thus provides robust performance against them. Through extensive simulations and the application to dataset obtained from the Alzheimer's Disease Neuroimaging Initiatives, we demonstrated that our method outperformed the commonly used kernel regression-based methods.

AVAILABILITY AND IMPLEMENTATION

The R-package is available at https://github.com/YaluWen/Uomic.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Clustering and variable selection evaluation of 13 unsupervised methods for multi-omics data integration

Recent advances in NGS sequencing, microarrays and mass spectrometry for omics data production have enabled the generation and collection of different modalities of high-dimensional molecular data. The integration of multiple omics datasets is a statistical challenge, due to the limited number of individuals, the high number of variables and the heterogeneity of the datasets to integrate. Recently, a lot of tools have been developed to solve the problem of integrating omics data including canonical correlation analysis, matrix factorization and SM. These commonly used techniques aim to analyze simultaneously two or more types of omics. In this article, we compare a panel of 13 unsupervised methods based on these different approaches to integrate various types of multi-omics datasets: iClusterPlus, regularized generalized canonical correlation analysis, sparse generalized canonical correlation analysis, multiple co-inertia analysis (MCIA), integrative-NMF (intNMF), SNF, MoCluster, mixKernel, CIMLR, LRAcluster, ConsensusClustering, PINSPlus and multi-omics factor analysis (MOFA). We evaluate the ability of the methods to recover the subgroups and the variables that drive the clustering on eight benchmarks of simulation. MOFA does not provide any results on these benchmarks. For clustering, SNF, MoCluster, CIMLR, LRAcluster, ConsensusClustering and intNMF provide the best results. For variable selection, MoCluster outperforms the others. However, the performance of the methods seems to depend on the heterogeneity of the datasets (especially for MCIA, intNMF and iClusterPlus). Finally, we apply the methods on three real studies with heterogeneous data and various phenotypes. We conclude that MoCluster is the best method to analyze these omics data. Availability: An R package named CrIMMix is available on GitHub at https://github.com/CNRGH/crimmix to reproduce all the results of this article.

DIABLO: an integrative approach for identifying key molecular drivers from multi-omics assays

MOTIVATION

In the continuously expanding omics era, novel computational and statistical strategies are needed for data integration and identification of biomarkers and molecular signatures. We present Data Integration Analysis for Biomarker discovery using Latent cOmponents (DIABLO), a multi-omics integrative method that seeks for common information across different data types through the selection of a subset of molecular features, while discriminating between multiple phenotypic groups.

RESULTS

Using simulations and benchmark multi-omics studies, we show that DIABLO identifies features with superior biological relevance compared with existing unsupervised integrative methods, while achieving predictive performance comparable to state-of-the-art supervised approaches. DIABLO is versatile, allowing for modular-based analyses and cross-over study designs. In two case studies, DIABLO identified both known and novel multi-omics biomarkers consisting of mRNAs, miRNAs, CpGs, proteins and metabolites.

AVAILABILITY AND IMPLEMENTATION

DIABLO is implemented in the mixOmics R Bioconductor package with functions for parameters' choice and visualization to assist in the interpretation of the integrative analyses, along with tutorials on http://mixomics.org and in our Bioconductor vignette.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

MOGSA: Integrative Single Sample Gene-set Analysis of Multiple Omics Data

Gene-set analysis (GSA) summarizes individual molecular measurements to more interpretable pathways or gene-sets and has become an indispensable step in the interpretation of large-scale omics data. However, GSA methods are limited to the analysis of single omics data. Here, we introduce a new computation method termed multi-omics gene-set analysis (MOGSA), a multivariate single sample gene-set analysis method that integrates multiple experimental and molecular data types measured over the same set of samples. The method learns a low dimensional representation of most variant correlated features (genes, proteins, etc.) across multiple omics data sets, transforms the features onto the same scale and calculates an integrated gene-set score from the most informative features in each data type. MOGSA does not require filtering data to the intersection of features (gene IDs), therefore, all molecular features, including those that lack annotation may be included in the analysis. Using simulated data, we demonstrate that integrating multiple diverse sources of molecular data increases the power to discover subtle changes in gene-sets and may reduce the impact of unreliable information in any single data type. Using real experimental data, we demonstrate three use-cases of MOGSA. First, we show how to remove a source of noise (technical or biological) in integrative MOGSA of NCI60 transcriptome and proteome data. Second, we apply MOGSA to discover similarities and differences in mRNA, protein and phosphorylation profiles of a small study of stem cell lines and assess the influence of each data type or feature on the total gene-set score. Finally, we apply MOGSA to cluster analysis and show that three molecular subtypes are robustly discovered when copy number variation and mRNA data of 308 bladder cancers from The Cancer Genome Atlas are integrated using MOGSA. MOGSA is available in the Bioconductor R package "mogsa."

PIMKL: Pathway-Induced Multiple Kernel Learning

Reliable identification of molecular biomarkers is essential for accurate patient stratification. While state-of-the-art machine learning approaches for sample classification continue to push boundaries in terms of performance, most of these methods are not able to integrate different data types and lack generalization power, limiting their application in a clinical setting. Furthermore, many methods behave as black boxes, and we have very little understanding about the mechanisms that lead to the prediction. While opaqueness concerning machine behavior might not be a problem in deterministic domains, in health care, providing explanations about the molecular factors and phenotypes that are driving the classification is crucial to build trust in the performance of the predictive system. We propose Pathway-Induced Multiple Kernel Learning (PIMKL), a methodology to reliably classify samples that can also help gain insights into the molecular mechanisms that underlie the classification. PIMKL exploits prior knowledge in the form of a molecular interaction network and annotated gene sets, by optimizing a mixture of pathway-induced kernels using a Multiple Kernel Learning (MKL) algorithm, an approach that has demonstrated excellent performance in different machine learning applications. After optimizing the combination of kernels to predict a specific phenotype, the model provides a stable molecular signature that can be interpreted in the light of the ingested prior knowledge and that can be used in transfer learning tasks.

The reliable classification of biomedical samples to predict phenotypic differences requires not only robust methods, but also interpretable approaches that can explain the reasons behind a prediction. A team led by María Rodríguez Martínez at IBM Research - Zürich has developed PIMKL, a methodology that exploits prior knowledge and enables the integration of multiple types of data with varying predictive power. Even when noisy datasets are simultaneously analyzed, PIMKL is able to discard uninformative data and achieve strong prediction power. Importantly, PIMKL produces a molecular signature that enables the interpretation of the results in terms of known biological functions. This signature can be transferred to other cohorts without loss of performance, demonstrating surprising robustness across cohorts. Interpretable algorithms can effectively help gain insights about disease mechanisms and build trust in a model.

Discovering Cancer Subtypes via an Accurate Fusion Strategy on Multiple Profile Data

Discovering cancer subtypes is useful for guiding clinical treatment of multiple cancers. Progressive profile technologies for tissue have accumulated diverse types of data. Based on these types of expression data, various computational methods have been proposed to predict cancer subtypes. It is crucial to study how to better integrate these multiple profiles of data. In this paper, we collect multiple profiles of data for five cancers on The Cancer Genome Atlas (TCGA). Then, we construct three similarity kernels for all patients of the same cancer by gene expression, miRNA expression and isoform expression data. We also propose a novel unsupervised multiple kernel fusion method, Similarity Kernel Fusion (SKF), in order to integrate three similarity kernels into one combined kernel. Finally, we make use of spectral clustering on the integrated kernel to predict cancer subtypes. In the experimental results, the P-values from the Cox regression model and survival curve analysis can be used to evaluate the performance of predicted subtypes on three datasets. Our kernel fusion method, SKF, has outstanding performance compared with single kernel and other multiple kernel fusion strategies. It demonstrates that our method can accurately identify more accurate subtypes on various kinds of cancers. Our cancer subtype prediction method can identify essential genes and biomarkers for disease diagnosis and prognosis, and we also discuss the possible side effects of therapies and treatment.

A multiple kernel density clustering algorithm for incomplete datasets in bioinformatics

Background

While there are a large number of bioinformatics datasets for clustering, many of them are incomplete, i.e., missing attribute values in some data samples needed by clustering algorithms. A variety of clustering algorithms have been proposed in the past years, but they usually are limited to cluster on the complete dataset. Besides, conventional clustering algorithms cannot obtain a trade-off between accuracy and efficiency of the clustering process since many essential parameters are determined by the human user’s experience.

Results

The paper proposes a Multiple Kernel Density Clustering algorithm for Incomplete datasets called MKDCI. The MKDCI algorithm consists of recovering missing attribute values of input data samples, learning an optimally combined kernel for clustering the input dataset, reducing dimensionality with the optimal kernel based on multiple basis kernels, detecting cluster centroids with the Isolation Forests method, assigning clusters with arbitrary shape and visualizing the results.

Conclusions

Extensive experiments on several well-known clustering datasets in bioinformatics field demonstrate the effectiveness of the proposed MKDCI algorithm. Compared with existing density clustering algorithms and parameter-free clustering algorithms, the proposed MKDCI algorithm tends to automatically produce clusters of better quality on the incomplete dataset in bioinformatics.