Integrative Analysis of Cancer Prognosis Data With Multiple Subtypes Using Regularized Gradient Descent

Abstract

In cancer research, high-throughput profiling studies have been extensively conducted, searching for genes/SNPs associated with prognosis. Despite seemingly significant differences, different subtypes of the same cancer (or different types of cancers) may share common susceptibility genes. In this study, we analyze prognosis data on multiple subtypes of the same cancer, but note that the proposed approach is directly applicable to the analysis of data on multiple types of cancers. We describe the genetic basis of multiple subtypes using the heterogeneity model, which allows overlapping but different sets of susceptibility genes/SNPs for different subtypes. An accelerated failure time (AFT) model is adopted to describe prognosis. We develop a regularized gradient descent approach, which conducts gene-level analysis and identifies genes that contain important SNPs associated with prognosis. The proposed approach belongs to the family of gradient descent approaches, is intuitively reasonable, and has affordable computational cost. Simulation study shows that when prognosis-associated SNPs are clustered in a small number of genes, the proposed approach outperforms alternatives with significantly more true positives and fewer false positives. We analyze an NHL (non-Hodgkin lymphoma) prognosis study with SNP measurements, and identify genes associated with the three major subtypes of NHL, namely DLBCL, FL and CLL/SLL. The proposed approach identifies genes different from using alternative approaches and has the best prediction performance.

Integrative analysis of prognosis data on multiple cancer subtypes

“…Beyond the proposed method, simulated data are also analyzed using a thresholding method (Ma et al, , referred to as “Threshold” below and in the tables) and a meta‐analysis method. With Threshold, Ma et al () proposes selecting the tuning parameters using fivefold cross‐validation. This method is referred to as “Threshold‐CV.” In addition, to make a fair comparison, we also consider “Threshold‐BIC,” which uses the BIC criterion to select tunings.…”

“…With high-dimensional measurements such as SNPs, data on individual subtypes have the "large d, small n" characteristic, with the sample size n much smaller than the number of SNPs d. Because of low sample size, susceptibility genes/SNPs identified from the analysis of individual subtypes may have unsatisfactory properties. Recent studies have shown that, when multiple datasets (multiple subtypes in this study) have overlapped susceptibility SNPs/genes, integrative analysis methods simultaneously analyze the raw data of multiple datasets and outperform single-dataset and classic meta-analysis methods (Ma, Huang, and Moran, 2009;Ma et al, 2012;Liu, Huang, and Ma, 2014).…”

Integrating multi-platform genomic data using hierarchical Bayesian relevance vector machines

“…Second, instead of using MKL methods, H-RVM learns parameters α and β from t, X, and Y using the variational learning algorithm of hierarchical kernel learning (HKL) [14,16]. H-RVM posits the following generative model for the (noisy) log survival time measurements t. Similar to MKL, K T β α represents the mean of t. The error distribution is Gaussian with mean 0 and precision parameter γ (8). Further, we impose a Gamma prior on γ such that Motivated by the "automatic relevance determination" idea of RVM, we impose independent Gaussian priors on the α j 's with the same mean 0 and different precision parameters φ j 's (10), where φ j controls (a priori) predictive power of the j-th feature vector from the three data sources for the log survival time.…”

Integrative analysis of ‘‐omics’ data using penalty functions

“…Meta‐analysis, which in a narrow sense refers to data integration solely based on summary statistics such as effect size, has played an important role in summarizing evidences from, for example, multiple genome‐wide association studies . Accumulating evidence has (e.g., Refs ) shown that jointly analyzing multiple datasets can improve analysis results. The scientific findings of these analyses encompass a wide range of topics in the field of biology and biomedicine, including inferring potential biological mechanisms, discovering novel disease subtypes, predicting novel complex biological networks as well as identifying biologically and clinically meaningful cancer prognostic markers, which have opened up new and exciting areas of investigation.…”