AMAISE: a machine learning approach to index-free sequence enrichment

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Metagenomics is crucial for diagnosing infectious diseases.
• Clinical samples are often contaminated with host DNA, hindering analysis.
• Current host-depletion methods are time-consuming or computationally intensive.

Purpose of the Study:

• To develop an efficient and accurate host-depletion method for metagenomic analysis.
• To introduce AMAISE, a novel index-free tool for sequence enrichment.

Main Methods:

• Developed AMAISE, a machine learning approach for index-free sequence enrichment.
• Applied AMAISE to separate host from microbial DNA reads.
• Evaluated AMAISE's performance in reducing memory usage for metagenomic classification.

Main Results:

• AMAISE achieved over 98% accuracy in separating host from microbial reads.
• Incorporating AMAISE reduced memory usage by 14-18% compared to metagenomic classification alone.
• Demonstrated the efficacy of a reference-independent machine learning approach.

Conclusions:

• AMAISE offers an accurate and efficient solution for host DNA depletion in clinical metagenomics.
• The index-free, machine learning strategy overcomes limitations of existing methods.
• This approach enhances the utility of metagenomics for infectious disease diagnostics.