Disease Classification in Metagenomics with 2D Embeddings and Deep Learning
This addresses the problem of overfitting in deep learning for high-dimensional, small-sample metagenomic data, offering a domain-specific solution for bioinformatics researchers.
The authors tackled disease classification from metagenomic data, where features outnumber samples, by proposing Met2Img to transform abundance data into synthetic images using taxonomic and t-SNE embeddings, achieving significant improvements over state-of-the-art methods like Random Forest and SVM on twelve datasets with over 1400 samples.
Deep learning (DL) techniques have shown unprecedented success when applied to images, waveforms, and text. Generally, when the sample size ($N$) is much bigger than the number of features ($d$), DL often outperforms other machine learning (ML) techniques, often through the use of Convolutional Neural Networks (CNNs). However, in many bioinformatics fields (including metagenomics), we encounter the opposite situation where $d$ is significantly greater than $N$. In these situations, applying DL techniques would lead to severe overfitting. Here we aim to improve classification of various diseases with metagenomic data through the use of CNNs. For this we proposed to represent metagenomic data as images. The proposed Met2Img approach relies on taxonomic and t-SNE embeddings to transform abundance data into "synthetic images". We applied our approach to twelve benchmark data sets including more than 1400 metagenomic samples. Our results show significant improvements over the state-of-the-art algorithms (Random Forest (RF), Support Vector Machine (SVM)). We observe that the integration of phylogenetic information alongside abundance data improves classification. The proposed approach is not only important in classification setting but also allows to visualize complex metagenomic data. The Met2Img is implemented in Python.