Contamination source modeling with SCRuB improves cancer phenotype prediction from microbiome data.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbiome sequencing is prone to contamination, which can obscure true biological signals.
• Existing in silico decontamination methods have limitations in utilizing shared information and handling partial contamination.

Purpose of the Study:

• To introduce Source tracking for Contamination Removal in microBiomes (SCRuB), a novel probabilistic method for in silico decontamination.
• To enhance the accuracy and robustness of microbiome data analysis by effectively removing contamination.

Main Methods:

• Developed SCRuB, a probabilistic method integrating information across multiple samples and controls.
• Validated SCRuB using data-driven simulations, induced contamination experiments, and diverse microbiome datasets.
• Compared SCRuB's performance against state-of-the-art decontamination techniques.

Main Results:

• SCRuB precisely identifies and removes contamination, outperforming existing methods by 15-20 times on average.
• Demonstrated SCRuB's robustness across various ecosystems, data types, and sequencing depths.
• Showcased improved host phenotype predictions, including melanoma patient treatment response, using decontaminated data.

Conclusions:

• SCRuB offers a significant advancement in microbiome data decontamination.
• The method enhances the reliability of microbiome analyses and downstream applications.
• SCRuB has broad applicability in microbiome research for more accurate biological insights.