Sparse group penalized integrative analysis of multiple cancer prognosis datasets

Abstract

SUMMARY In cancer research, high-throughput profiling studies have been extensively conducted, searching for markers associated with prognosis. Because of the “large d, small n” characteristic, results generated from the analysis of a single dataset can be unsatisfactory. Recent studies have shown that integrative analysis, which simultaneously analyzes multiple datasets, can be more effective than single-dataset analysis and classic meta-analysis. In most of existing integrative analysis, the homogeneity model has been assumed, which postulates that different datasets share the same set of markers. Several approaches have been designed to reinforce this assumption. In practice, different datasets may differ in terms of patient selection criteria, profiling techniques, and many other aspects. Such differences may make the homogeneity model too restricted. In this study, we assume the heterogeneity model, under which different datasets are allowed to have different sets of markers. With multiple cancer prognosis datasets, we adopt the AFT (accelerated failure time) model to describe survival. This model may have the lowest computational cost among popular semiparametric survival models. For marker selection, we adopt a sparse group MCP (minimax concave penalty) approach. This approach has an intuitive formulation and can be computed using an effective group coordinate descent algorithm. Simulation study shows that it outperforms the existing approaches under both the homogeneity and heterogeneity models. Data analysis further demonstrates the merit of heterogeneity model and proposed approach.

Sparse Regularization for Bi-Level Variable Selection

“…Note that, however, only the group bridge, the composite MCP and the group exponential lasso appropriately share the roles of f I and f O described above. Bi-level selection is applied for the analysis of real data in several fields (Peng et al, 2010;Zhou et al, 2010;Chatterjee et al, 2012;Liu et al, 2013). Furthermore, penalties or methods for bi-level selection other than those described in this paper have been proposed in Abramovich and Grinshtein (2013), Xiang et al (2014) and Jiang and Huang (2015).…”

Promoting similarity of model sparsity structures in integrative analysis of cancer genetic data

“…However, the stopping rule is selected by jointly considering the M datasets. Loosely speaking, this amounts to applying a comparable amount of regularization to all datasets, which has been suggested in integrative analysis using the penalization technique [].…”

“…Here, it is possible that I( m,j = 0) ≠ I( k,j = 0) for some (j, m, k)'s. The heterogeneity structure includes the homogeneity structure as a special case and is more flexible [5,6].In this study, we conduct integrative analysis under the heterogeneity structure. Although multiple datasets are allowed to have different sets of markers, as the basis of integrating multiple datasets, it is reasonable to expect that they share some common markers.…”

Penalized integrative analysis under the accelerated failure time model

“…A lung cancer prognosis study with gene expression measurements is presented in this article, and more are available in the literature. With such "large p, small n" data, results generated in the analysis of a single dataset are often unsatisfactory because of the small sample size (Guerra and Goldstein, 2009;Liu et al, 2013;Ma et al, 2011b). For outcomes of common interest, there are often multiple independent studies with comparable designs.…”

“…This makes integrative analysis even more complicated. Penalization methods for integrative analysis have been developed (Liu et al, 2013;Ma et al, 2011b), however, in an unsystematic manner.…”