A Multi-Omics Framework for Survival Mediation Analysis of High-Dimensional Proteogenomic Data
Seungjun Ahn, Weijia Fu, Maaike van Gerwen, Lei Liu, Zhigang Li · 2025
Read A Multi-Omics Framework for Survival Mediation Analysis of High-Dimensional Proteogenomic Data by Seungjun Ahn, Weijia Fu, Maaike van Gerwen, Lei Liu, Zhigang Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-08606.txt.
Introduction
Seungjun Ahn and co-authors built a numeric method in 2025 for tracing how exposures affect survival through chains of molecular mediators. Patients' survival often depends on proteins and genes acting as middlemen between a risk factor and death. Standard mediation tools for survival data handle one exposure and one omics layer at a time, missing how layers interact. Their method, SMAHP, handles high-dimensional exposures and mediators jointly across multi-omics data.
The method is among the first to bring the accelerated failure time model into multi-omics causal mediation for survival. Models across multiple cases were used to test its actions. Identifying true causal pathways, rather than mere links, could show drug builders exactly which molecule to push on. In proteogenomic cohorts where thousands of traits compete for focus, a principled filter like this is needed infrastructure.
Catalog record
| CREATOR | Seungjun Ahn, Weijia Fu, Maaike van Gerwen, Lei Liu, Zhigang Li |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-08606
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Multi-Omics Framework for Survival Mediation Analysis of High-Dimensional Proteogenomic Data Seungjun Ahn, Weijia Fu, Maaike van Gerwen, Lei Liu, Zhigang Li Abstract Motivation: Survival analysis plays a crucial role in understanding time-to-event (survival) outcomes such as disease progression. Despite recent advancements in causal mediation frameworks for survival analysis, existing methods are typically based on Cox regression and primarily focus on a single exposure or individual omics layers, often overlooking multi-omics interplay. This limitation hinders the full potential of integrated biological insights. Results: In this paper, we propose SMAHP, a novel method for survival mediation analysis that simultaneously handles high-dimensional exposures and mediators, integrates multi-omics data, and offers a robust statistical framework for identifying causal pathways on survival outcomes. This is one of the first attempts to introduce the accelerated failure time (AFT) model within a multi-omics causal mediation framework for survival outcomes. Through simulations across multiple scenarios, we demonstrate that SMAHP achieves high statistical power, while effectively controlling false discovery rate (FDR), compared with two other approaches. We further apply SMAHP to the largest head-and-neck carcinoma proteogenomic data, detecting a gene mediated by a protein that influences survival time. Availability and implementation: R package is freely available on CRAN repository (https://CRAN.R-project.org/package=SMAHP) and published under General Public License version 3. 2024 keywords Mediation analysis, Multi-omics, Time-to-event, Causality, Proteogenomic data, CPTAC 1 Introduction Survival analysis is a powerful tool for understanding time-to-event outcomes, such as patient survival or disease progression. Recent advancements in high-throughput technologies have enabled the profiling of biological data on a scale that was once unimaginable, including proteomics data and RNA-Seq data. These advances open new avenues for uncovering complex relationships between biological entities and survival outcomes. For example, recent studies have identified protein or gene biomarkers associated with survival outcomes in patients with idiopathic pulmonary fibrosis ( 1 ) , glioblastoma ( 2 ) , hepatocellular carcinoma ( 3 ) , small-cell lung carcinoma ( 4 ) , cardiovascular diseases ( 5 , 6 ) , Alzheimer’s disease and related dementias ( 7 ) , and oropharyngeal carcinoma ( 8 ) , often using Cox proportional hazards (PH) regression models with or without penalization methods. While traditional survival analysis methods, such as Cox PH regression, have been widely used to assess the direct effects of individual predictor variables on survival outcomes, they often fall short in capturing indirect effects or mediating pathways. Specifically, predictor models are fitted separately to assess associations between genes and survival, and between proteins and survival. The mediation analysis offers a solution by exploring how a hypothetical intermediate variable (i.e., mediator) bridges the effect of an exposure variable on an outcome. In the context of survival analysis, this approach provides valuable insights into how molecular biomarkers, such as protein expression levels, mediate the relationship between RNA-Seq gene expression and survival. Over the past few years, pioneering methodological approaches have emerged for causal mediation frameworks with survival outcomes in the analysis of omics data. Notably, High-dimensional Mediation Analysis in Survival Models (HIMAsurvival) ( 9 ) first introduced a Cox PH regression framework, incorporating variable selection techniques via sure independence screening (SIS) ( 10 ) and minimax concave penalty (MCP)-penalization ( 11 ) , to estimate and test the effects of high-dimensional mediators on survival outcome. Following this, another study presented a Cox regression-based framework ( 12 ) , featuring adaptations of de-biased Lasso inference and a joint significance test for better false discovery control compared to HIMAsurvival ( 9 ) . More recently, the Mediation Analysis of Survival Outcomes and High-Dimensional Omics Mediators (MASH) ( 13 ) proposed a three-step high-dimensional mediator selection procedure (i.e., pre-screening with SIS, MCP for variable selection, and FDR control using the Benjamini-Hochberg (BH) procedure). While the mediator selection procedure is similar to the two methods described earlier ( 12 , 9 ) , MASH distinguishes itself by using R^{2} -based measures to estimate mediation effects, with its primary framework again centered on the Cox regression model. Despite these advances in survival mediation analysis, two interconnected limitations persist. First, existing survival mediation approaches have primarily focused on high-dimensional mediators, with much less attention given to the setting of exposures. These studies typically consider a single binary or continuous exposure, overlooking the opportunity to capture the rich information present in high-dimensional exposure variables. A previous study ( 14 ) explored a frailty model (or Cox model with random effects) to identify mediation effects in the presence of high-dimensional exposures, but it only considers one mediator at a time. Second, these methods have showcased applications to individual omics layer (e.g., DNA methylation, metabolomics, and copy number variation data). However, biological systems are complex and driven by interplay across multiple omics layers. As demonstrated in numerous clinical and bioinformatics studies ( 15 , 16 , 17 , 18 ) , the integration of multilayer (or multi-omics) analysis has been used to characterize key multi-omics pathways, offering a more comprehensive view of the molecular mechanisms underlying complex diseases such as cancer and Alzheimer’s disease. Significant gaps still remain, as only one study has attempted to address this issue in the context of survival causal mediation in the analysis of omics data ( 19 ) . This unified mediation analysis framework accounts for both multivariable exposures and mediators in relation to survival outcomes, applying it to proteogenomic data to identify genes mediated by proteins associated with survival. However, the primary focus was not specifically on survival outcomes. This study considered a Cox model, which may not be valid if the PH assumption is violated. Additionally, the simulations in the study did not explore a range of censoring rates (set to 50 \% ). The motivation of this paper is to develop a statistical methodology capable of (1) handling both high-dimensional mediators and exposures, (2) testing the mediation effect between exposure and survival outcomes with proper FDR control, and (3) analyzing multiple omics platforms simultaneously. In addition, from a clinical perspective, most existing mediation methods have been implemented based on individual omics layers, with the exception of proteomics and RNA-Seq data. However, many clinical proteogenomic studies ( 20 , 21 , 22 , 23 , 24 , 25 , 26 ) have highlighted the synergy between protein expression/proteomics and genomics, as well as their links to phenotypes across various disease types. To overcome these challenges and achieve the objectives of this study, we propose a novel approach to mediation analysis, specifically focused on survival outcomes and framed within the counterfactual paradigm. By implementing and applying this methodology to multilayered, high-dimensional proteogenomic data (high-dimensional genomic exposures from RNA-Seq data and high-dimensional protein mediators from proteomics data), we introduce this framework as S urvival M ediation A nalysis of H igh-dimensional P roteogenomic data (SMAHP). Our method procedure involves the following parts: First, we conduct a preliminary screening of high-dimensional mediators and exposures using penalized accelerated time failure (AFT) ( 27 ) and MCP-penalized regression models ( 11 ) , based on their disjoint indirect effects (exposure \rightarrow mediator and mediator \rightarrow outcome) or direct effects (exposure \rightarrow outcome). Second, using proteins and genes selected from the preliminary screening, we adopt the SIS ( 10 ) to identify “important” protein mediators that relate the gene exposures to