Rank-Based Identification of High-dimensional Surrogate Markers: Application to Vaccinology

2025-02-05

AI Summary

Addressing the challenges of predicting long-term immunogenicity in vaccine clinical trials and modeling high-dimensional, low-sample-size RNA-seq data, this study proposes RISE—a two-stage, rank-based nonparametric framework for interpretable and robust identification of early surrogate biomarkers. RISE circumvents strong distributional assumptions and large-sample requirements inherent in conventional models by employing univariate rank tests for initial screening, independent surrogate dataset validation, and pathway enrichment analysis—ensuring both biological plausibility and statistical rigor. Applied to an inactivated influenza vaccine cohort, RISE identified an interferon-related gene signature measured one day post-vaccination that accurately predicted neutralizing antibody titers at day 28 (AUC > 0.85). This significantly shortens the timeline for vaccine efficacy assessment and provides a reliable, biologically grounded surrogate endpoint for early immunogenicity evaluation.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Metareasoning and Metaheuristics
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web search models and ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

In vaccine trials with long-term participant follow-up, it is of great importance to identify surrogate markers that accurately infer long-term immune responses. These markers offer practical advantages such as providing early, indirect evidence of vaccine efficacy, and can accelerate vaccine development while identifying potential biomarkers. High-throughput technologies like RNA-sequencing have emerged as promising tools for understanding complex biological systems and informing new treatment strategies. However, these data are high-dimensional, presenting unique statistical challenges for existing surrogate marker identification methods. We introduce Rank-based Identification of high-dimensional SurrogatE Markers (RISE), a novel approach designed for small sample, high-dimensional settings typical in modern vaccine experiments. RISE employs a non-parametric univariate test to screen variables for promising candidates, followed by surrogate evaluation on independent data. Our simulation studies demonstrate RISE's desirable properties, including type one error rate control and empirical power under various conditions. Applying RISE to a clinical trial for inactivated influenza vaccination, we sought to identify genes whose post-vaccination expression could serve as a surrogate for the induced immune response. This analysis revealed a signature of genes whose combined expression at 1 day post-injection appears to be a reasonable surrogate for the neutralising antibody titres at 28 days after vaccination. Pathways related to innate antiviral signalling and interferon stimulation were strongly represented in this derived surrogate, providing a clear immunological interpretation.

Problem

• Identify high-dimensional surrogate markers for long-term immune responses
• Address statistical challenges in small sample, high-dimensional vaccine data
• Develop a non-parametric method to evaluate surrogate marker candidates

Innovation

• Non-parametric univariate test screens variables
• Two-stage rank-based surrogate marker identification
• Handles high-dimensional data in small samples