Computational Systems Biology Methods and Protocols.7z

through parameters that describe the agreement among the

datasets [10]. Meanwhile, a nonparametric Bayesian model

has been introduced to discover prognostic cancer subtypes

by constructing a hierarchy of Dirichlet processes and has

shown a good ability to distinguish concordant and discordant

signals within each patient sample [26].

2.Machine-learning-based integration model. The main idea

under such methods is to extract significant data pattern

along with integrative analysis. An extended multiple kernel

learning has been applied for dimensionality reduction

approaches, and several kernels per data type are applicable to

avoid the unnecessary choice of the best kernel functions and

Table 4
The representative approaches of top-down integration

Categories Methods Computational instructions

Statistic (factor-

centered)

“Residuals” [115] A two-stage approach based on regularized singular value

decomposition, and regularized estimation of prediction

model

iCluster [24] A joint latent variable model incorporating the variance-

covariance structures

iCluster+ [11] Joint modeling is proposed to ensemble discrete and

continuous variables

moCluster [101] Multiblock multivariate analysis and an ordinary clustering

algorithm

iBAG [25] Hierarchical modeling within Bayesian analysis

MDI [10] Dirichlet-multinomial allocation (DMA) mixture model

within Bayesian analysis

“Nonparametric

Bayesian model”

[26]

A hierarchy of Dirichlet processes within a nonparametric

Bayesian model

“Factor analysis”

[116]

Factor analysis

Optimization

(matrix-centered)

“Joint matrix

factorization” [21]

Joint nonnegative matrix factorization

“Multi-view

bi-clustering” [30]

Rank matrix factorization

GSVD [104][105] Higher-order generalized singular value decomposition

“Ping-pong” [29] Ping-pong algorithm

Machine learning

(pattern-

centered)

“Linear discriminant

analysis” [22]

Factor analysis, combined with linear discriminant analysis

“Kernel-based” [27] Multiple kernel learning

JointCluster [28] Simultaneous clustering of multiple networks

SNF [102] Similarity network fusion based on theoretical multi-view

learning framework

PFA [103] Pattern fusion analysis based on local tangent space

alignment (LTSA) theory

122 Xiang-Tian Yu and Tao Zeng