A Deep learning approach for host depletion in metagenomic samples

Abstract

Metagenomic studies often struggle with excessive host DNA, which reduces the sensitivity and accuracy of microorganism detection. Traditional lab-based host depletion is costly and time-consuming, while computational methods using reference databases are resource-intensive and often less accurate. To overcome these limitations, there is a growing need for efficient, accurate, and resource-friendly host depletion techniques. Machine learning (ML) offers a promising alternative by enabling read classification without relying on large reference databases, reducing computational load and improving speed and reliability. Such approaches can greatly enhance the effectiveness of metagenomic analyses across diverse host species.