A Likelihood Ratio Approach for Precise Discovery of Truly Relevant Protein Markers

TL;DR

This paper introduces a likelihood ratio method for accurately identifying truly relevant protein markers in high-dimensional, small-sample discovery studies, improving biomarker discovery efficiency and reducing false positives.

Contribution

The paper presents a novel likelihood ratio approach tailored for discovery studies, effectively controlling false discovery rate and distinguishing relevant markers from irrelevant ones.

Findings

High sensitivity in identifying relevant markers

Low empirical false discovery rate achieved

Effective in both simulated and real data

Abstract

The process of biomarker discovery is typically lengthy and costly, involving the phases of discovery, qualification, verification, and validation before clinical evaluation. Being able to efficiently identify the truly relevant markers in discovery studies can significantly simplify the process. However, in discovery studies the sample size is typically small while the number of markers being explored is much larger. Hence discovery studies suffer from sparsity and high dimensionality issues. Currently the state-of-the-art methods either find too many false positives or fail to identify many truly relevant markers. In this paper we develop a likelihood ratio-based approach and aim for accurately finding the truly relevant protein markers in discovery studies. Our method fits especially well with discovery studies because they are mostly balanced design due to the fact that experiments are limited and controlled. Our approach is based on the observation that the underlying distributions of expression profiles are unimodal for those irrelevant plain markers. Our method has asymptotic chi-square null distribution which facilitates the efficient control of false discovery rate. We then evaluate our method using both simulated and real experimental data. In all the experiments our method is highly effective to discover the set of truly relevant markers, leading to accurate biomarker identifications with high sensitivity and low empirical false discovery rate.