EPS: automated feature selection in case-control studies using extreme pseudo-sampling.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• High-dimensional biological case-control datasets present challenges for identifying informative predictive features.
• Effective feature selection is crucial for accurate biological data analysis and interpretation.

Purpose of the Study:

• To introduce and evaluate the Extreme Pseudo-Sampling (EPS) algorithm for feature selection in high-dimensional biological data.
• To present an enhanced, open-source Python implementation of the EPS algorithm with improved customizability.

Main Methods:

• The EPS algorithm combines deep learning (variational autoencoder) with logistic regression to generate latent sample representations.
• It creates pseudo-samples around extreme cases and controls to augment the dataset.
• Feature significance is determined by training a regression model on the upsampled data.

Main Results:

• The EPS algorithm effectively identifies significant predictive features in complex biological datasets.
• The open-source implementation offers enhanced customizability for data preparation, model training, and classification.

Conclusions:

• The Extreme Pseudo-Sampling algorithm provides a robust solution for feature selection in challenging high-dimensional biological data.
• The enhanced open-source package facilitates wider adoption and application across diverse biological datasets.