PlasClass improves plasmid sequence classification.

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• Distinguishing bacterial plasmid DNA from genomic DNA is challenging, particularly in metagenomic assemblies with numerous unknown sequence origins.
• Current plasmid classification tools struggle with short sequences, are trained on limited plasmid data, and can be complex to implement.

Purpose of the Study:

• To introduce PlasClass, a novel and improved computational tool for classifying DNA sequences of plasmid origin.
• To enhance the accuracy and efficiency of identifying plasmid contigs in bacterial and metagenomic datasets.

Main Methods:

• PlasClass employs standard classifiers trained on an extensive and up-to-date collection of known plasmid sequences.
• The tool is optimized for various sequence lengths, with a focus on improving classification of shorter sequences.
• Performance was evaluated using held-out data, simulations, and real-world bacterial isolate, plasmidome, and metagenomic samples.

Main Results:

• PlasClass demonstrates superior performance compared to state-of-the-art methods, especially for shorter sequences, achieving higher F1 scores across diverse datasets.
• The tool significantly reduces computational time and memory usage.
• Accurate classification of plasmid and bacterial genome sequences was achieved in both metagenomic and isolate assembly contexts.

Conclusions:

• PlasClass offers a more reliable, efficient, and user-friendly solution for plasmid sequence classification.
• Its improved performance on short sequences addresses a critical limitation in current bioinformatics tools for analyzing complex microbial communities.
• The tool facilitates easier and more accurate analysis of plasmid DNA in various genomic contexts.