Plasmidity: A Novel Pipeline for Plasmid Sequence Prediction

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Plasmid DNA is crucial for bacterial adaptation and antibiotic resistance.
• Accurate plasmid detection from whole genome sequencing (WGS) data remains challenging.
• Existing tools often lack comprehensive feature analysis for robust plasmid identification.

Purpose of the Study:

• To develop and validate PLASMIDITY, an automated pipeline for plasmid sequence detection.
• To leverage machine learning for enhanced characterization of plasmid contigs.
• To improve the accuracy and efficiency of identifying plasmid DNA in genomic datasets.

Main Methods:

• PLASMIDITY integrates contig assembly with machine learning classifiers (Gradient Boosting).
• Features analyzed include contig length, multiplicity, circularity, and genetic markers (plasmid and chromosome).
• The pipeline utilizes k-mer profiles and similarity searches against plasmid reference databases.

Main Results:

• Trained Gradient Boosting classifiers achieved a maximum F1 score of 0.901 on simulated WGS data.
• The pipeline was trained and tested on 200 taxonomically diverse samples.
• PLASMIDITY demonstrated improved performance compared to other plasmid prediction tools.

Conclusions:

• PLASMIDITY offers a robust and automated solution for plasmid sequence detection.
• The machine learning approach effectively utilizes diverse sequence and marker features.
• This tool enhances the analysis of plasmid-borne traits in genomic studies.