A robust rerank approach for feature selection and its application to pooling-based GWA studies

Area of Science:

• Biostatistics
• Genomics
• Bioinformatics

Background:

• Large-p-small-n datasets are prevalent in modern biomedical research, posing challenges for traditional statistical methods.
• Conventional methods like t-tests and Area Under the Curve (AUC) estimates struggle with variance instability and tied values.
• Significance Analysis of Microarrays (SAM) is sensitive to tuning parameters, complicating its application.

Purpose of the Study:

• To propose a robust rerank approach for feature selection in large-p-small-n biomedical datasets.
• To overcome the limitations of conventional statistical methods and SAM in high-dimensional, low-sample-size scenarios.
• To develop a method that is insensitive to tuning parameter selection for practical implementation.

Main Methods:

• A novel rank-based statistic, 'rank-over-variable', is computed for each feature.
• Iterative application of 'random subset' and 'rerank' techniques to rank features.
• Selection of leading features based on the reranking process for downstream analysis.

Main Results:

• The proposed rerank approach demonstrates effectiveness in identifying significant features from large-p-small-n data.
• The method shows robustness and applicability in simulation studies and real-world pooling-based genome-wide association (GWA) studies.
• The approach is particularly suitable for datasets with a large number of features and a small number of samples.

Conclusions:

• The robust rerank approach provides a valuable tool for feature selection in challenging large-p-small-n biomedical datasets.
• Its insensitivity to tuning parameters makes it a practical and reliable method for researchers.
• The method enhances the analysis of complex genomic data, such as in GWA studies.