Multiset multicover methods for discriminative marker selection.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Markers are crucial for high-throughput data analysis, including cell type assignment in single-cell RNA sequencing (scRNA-seq) and deconvolving bulk gene expression.
• Current marker selection methods, primarily based on differential expression (DE) analysis, are inadequate for large-scale datasets with numerous cell types and tissues.

Purpose of the Study:

• To address the limitations of existing marker selection methods for large, complex biological datasets.
• To introduce a novel framework, the phenotype cover (PC) problem, for marker selection.
• To develop and present algorithms that enhance the discriminative power of marker sets.

Main Methods:

• Definition of the phenotype cover (PC) problem tailored for marker selection in large datasets.
• Development of new algorithms designed to solve the PC problem.
• Evaluation of the proposed algorithms on various marker-selection tasks using high-throughput biological data.

Main Results:

• The developed algorithms effectively address the phenotype cover problem.
• The proposed marker selection methods demonstrate improved discriminative power compared to traditional approaches.
• Analysis confirms the ability of these methods to accurately distinguish different phenotypes within complex datasets.

Conclusions:

• The novel phenotype cover (PC) problem and associated algorithms offer a significant advancement for marker selection in large-scale biological studies.
• These methods provide accurate solutions for distinguishing phenotypes, particularly in scRNA-seq and spatial proteomics.
• The findings pave the way for more robust and precise analysis of complex biological systems using marker-based approaches.