Hostile: accurate decontamination of microbial host sequences

Area of Science:

• Genomics
• Bioinformatics
• Microbiology

Background:

• Clinical sequencing samples are frequently contaminated with human DNA, necessitating removal for ethical, legal, and analytical reasons.
• Inaccurate host sequence removal can compromise downstream analyses like variant calling and de novo assembly of microbial genomes.

Purpose of the Study:

• To develop an accurate and efficient computational tool for removing human host sequences from both short and long sequencing reads.
• To ensure the preservation of target microbial sequences during the decontamination process.

Main Methods:

• Development of Hostile, a Python-based tool for host sequence decontamination.
• Evaluation of Hostile's performance on simulated and real sequencing data, assessing human read removal and microbial read retention.
• Comparison of Hostile's efficiency and accuracy against existing host decontamination tools.

Main Results:

• Hostile effectively removes at least 99.6% of human reads while retaining at least 99.989% of bacterial reads.
• Utilizing a masked reference genome with Hostile further improved bacterial read retention to ≥99.997% with minimal impact on human read removal.
• Hostile demonstrated superior performance compared to an existing tool, removing more human reads and fewer bacterial reads, often in less time.

Conclusions:

• Hostile provides accurate and efficient host decontamination for diverse sequencing read types.
• The tool's performance, especially with masked reference genomes, makes it valuable for microbial genomics research and clinical applications.
• Hostile is available as an open-source Python package, facilitating its adoption in the scientific community.