Meta-analytic support vector machine for integrating multiple omics data

Kim, Sunghwan; Jhong, Jae-Hwan; Lee, JungJun; Koo, Ja−Yong

doi:10.1186/s13040-017-0126-8

Cited by 112 publications

(93 citation statements)

References 50 publications

(55 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Several efforts have been made for multiple omics data integration in the context of SVM learning. Kim et al (28) proposed a meta-analytic support vector machine (Meta-SVM) that can accommodate multiple omics data, making it possible to detect consensus genes associated with diseases across studies. The Meta-SVM method was applied to breast cancer expression profiles provided by TCGA including mRNA, copy number variation (CNV) and epigenetic DNA methylation.…”

Section: Cancer Classification and Subtypingmentioning

confidence: 99%

Applications of Support Vector Machine (SVM) Learning in Cancer Genomics

2018

CGP

685

347

View full text Add to dashboard Cite

Abstract. Machine learning with maximization (support) of separating margin (vector), called support vector machine (SVM) learning, is a powerful classification tool that hasMachine learning (ML) "learns" a model from past data in order to predict future data (1). The key process is the learning which is one of the artificial intelligences. Many different statistical, probabilistic, and optimization techniques can be implemented as the learning methods such as the logistic regression, artificial neural networks (ANN), K-nearest neighbor (KNN), decision trees (DT) and Naive Bayes. There are two main types of ML learning -supervised learning and unsupervised learning. The supervised learning builds a model by learning from known classes (labeled training data). In contrast, unsupervised learning methods learn the common features from unknown class data (unlabeled training data).ML algorithms have been used for key feature training and recognition and for group classification. The strength of ML methods is it could detect hard-to-discern patterns from large, noisy or complex data sets. This capability is particularly well-suited to complex genomic data, especially in cancer studies. For example, ANN and DT have been used in cancer detection and diagnosis for nearly 20 years (2-3). The clinical implication of cancer heterogeneity and various cancer genomic data available motivate the applications of ML for cancer classification using genomic data.SVM learning is one of many ML methods. Compared to the other ML methods SVM is very powerful at recognizing subtle patterns in complex datasets (4). SVM can be used to recognize handwriting, recognize fraudulent credit cards, identify a speaker, as well as detect face (5). Cancer is a genetic disease where the genomic feature patterns or feature function patterns may represent the cancer subtypes, the outcome prognosis, drug benefit prediction, tumorigenesis drivers, or a tumor-specific biological process. Therefore, the Artificial Intelligence of SVM can help us in recognizing these patterns in a variety of applications. SVM ModelSVM is a powerful method for building a classifier. It aims to create a decision boundary between two classes that enables the prediction of labels from one or more feature vectors (6). This decision boundary, known as the hyperplane, is orientated in such a way that it is as far as 41

show abstract

Section: Cancer Classification and Subtypingmentioning

confidence: 99%

Applications of Support Vector Machine (SVM) Learning in Cancer Genomics

2018

CGP

685

347

View full text Add to dashboard Cite

show abstract

“…Rare and low-frequency disruptive coding variants within LPA have been previously associated with Lp(a) 24,25 . Here, we performed two coding rare variant analyses studies (RVAS) aggregating rare (MAF <1%) variants which were 1) LOF or missense deleterious by in-silico prediction tools 35 , or 2) non-synonymous, within their respective genes, and performed association with Lp(a)-C, adjusting for KIV2-CN. All analyses were done separately for JHS and EST and metaanalyzed.…”

Section: Rare Variant Analysis: Coding and Non-coding Burden Testsmentioning

confidence: 99%

Deep coverage whole genome sequences and plasma lipoprotein(a) in individuals of European and African ancestries

Zekavat

Ruotsalainen

Handsaker

et al. 2017

Preprint

View full text Add to dashboard Cite

Lipoprotein(a), Lp(a), is a modified low-density lipoprotein particle where apolipoprotein(a) (protein product of the LPA gene) is covalently attached to apolipoprotein B. Lp(a) is a highly heritable, causal risk factor for cardiovascular diseases and varies in concentrations across ancestries. To comprehensively delineate the inherited basis for plasma Lp(a), we performed deep-coverage whole genome sequencing in 8,392 individuals of European and African American ancestries. Through whole genome variant discovery and direct genotyping of all structural variants overlapping LPA, we quantified the 5.5kb kringle IV-2 copy number (KIV2-CN), a known LPA structural polymorphism, and developed a model for its imputation. Through common variant analysis, we discovered a novel locus (SORT1) associated with Lp(a)-cholesterol, and also genetic modifiers of KIV2-CN. Furthermore, in contrast to previous GWAS studies, we explain most of the heritability of Lp(a), observing Lp(a) to be 85% heritable among African Americans and 75% among Europeans, yet with notable inter-ethnic heterogeneity. Through analyses of aggregates of rare coding and non-coding variants with Lp(a)-cholesterol, we found the only genome-wide significant signal to be at a non-coding SLC22A3 intronic window also previously described to be associated with Lp(a); however, this association was mitigated by adjustment with KIV2-CN. Finally, using an additional imputation dataset (N=27,344), we performed Mendelian randomization of LPA variant classes, finding that genetically regulated Lp(a) is more strongly associated with incident cardiovascular diseases than directly measured Lp(a), and is significantly associated with measures of subclinical atherosclerosis in African Americans.

show abstract

“…An important limitation of support vector machines shared by the majority of machine learning approaches is that, the larger the number of features relative to observations, the higher the probability that the model overfits the data [14]. Taking this limitation into consideration, Madhavan et al combined multiple machine learning approaches to integrate gene expression, micro-RNA expression, copy number variant, and serum and urine metabolomics data.…”

Section: Methodsmentioning

confidence: 99%

Data mining and machine learning approaches for the integration of genome-wide association and methylation data: methodology and main conclusions from GAW20

et al. 2018

View full text Add to dashboard Cite

BackgroundMultiple layers of genetic and epigenetic variability are being simultaneously explored in an increasing number of health studies. We summarize here different approaches applied in the Data Mining and Machine Learning group at the GAW20 to integrate genome-wide genotype and methylation array data.ResultsWe provide a non-intimidating introduction to some frequently used methods to investigate high-dimensional molecular data and compare the different approaches tried by group members: random forest, deep learning, cluster analysis, mixed models, and gene-set enrichment analysis. Group contributions were quite heterogeneous regarding investigated data sets (real vs simulated), conducted data quality control and assessed phenotypes (eg, metabolic syndrome vs relative differences of log-transformed triglyceride concentrations before and after fenofibrate treatment). However, some common technical issues were detected, leading to practical recommendations.ConclusionsDifferent sources of correlation were identified by group members, including population stratification, family structure, batch effects, linkage disequilibrium and correlation of methylation values at neighboring cytosine-phosphate-guanine (CpG) sites, and the majority of applied approaches were able to take into account identified correlation structures. The ability to efficiently deal with high-dimensional omics data, and the model free nature of the approaches that did not require detailed model specifications were clearly recognized as the main strengths of applied methods. A limitation of random forest is its sensitivity to highly correlated variables. The parameter setup and the interpretation of results from deep learning methods, in particular deep neural networks, can be extremely challenging. Cluster analysis and mixed models may need some predimension reduction based on existing literature, data filtering, and supplementary statistical methods, and gene-set enrichment analysis requires biological insight.

show abstract

Meta-analytic support vector machine for integrating multiple omics data

Cited by 112 publications

References 50 publications

Applications of Support Vector Machine (SVM) Learning in Cancer Genomics

Applications of Support Vector Machine (SVM) Learning in Cancer Genomics

Deep coverage whole genome sequences and plasma lipoprotein(a) in individuals of European and African ancestries

Data mining and machine learning approaches for the integration of genome-wide association and methylation data: methodology and main conclusions from GAW20

Contact Info

Product

Resources

About