Feature selection with vector-symbolic architectures: a case study on microbial profiles of shotgun metagenomic samples of colorectal cancer.

Abstract

The continuously decreasing cost of next-generation sequencing has recently led to a significant increase in the number of microbiome-related studies, providing invaluable information for understanding host-microbiome interactions and their relation to diseases. A common approach in metagenomics consists of determining the composition of samples in terms of the amount and types of microbial species that populate them, with the goal of identifying microbes whose profiles are able to differentiate samples under different conditions with advanced feature selection techniques. Here, we propose a novel backward variable selection method based on the hyperdimensional computing (HDC) paradigm, which takes inspiration from how the human brain works in the classification of concepts by encoding features into vectors in a high-dimensional space. We validated our method on public metagenomic samples collected from patients affected by colorectal cancer in a case/control scenario, by performing a comparative analysis with other state-of-the-art feature selection methods, obtaining promising results. Characterizing the microbial composition of metagenomic samples is crucial for identifying potential biomarkers that can distinguish between healthy and diseased states. However, the high dimensionality and complexity of metagenomic data present significant challenges in the context of accurately selecting features. Our backward variable selection method, based on the HDC paradigm, offers a promising approach to overcoming these challenges. By effectively reducing the feature space while preserving essential information, this method enhances the ability to detect critical microbial signatures associated with diseases like colorectal cancer, leading to more precise diagnostic tools.