PlasTrans

PlasTrans identifies transmissible plasmid-derived sequences in metagenomic plasmidomes to distinguish conjugative and mobilizable plasmid fragments and assess their potential for horizontal gene transfer.

Key Features:

• Transferability Annotation: Produces per-sequence annotations indicating potential for horizontal gene transfer (transferability) for plasmid-derived fragments.
• Input and Output Formats: Accepts plasmid sequences in FASTA format and outputs a tabular file with transferability annotations.
• Information Entropy Analysis: Applies information entropy from information theory to assess randomness of synonymous codon usage across 4424 plasmid genomes, revealing distinct signatures between transmissible and non-transmissible plasmids.
• Deep Learning (CNN): Employs a convolutional neural network to extract complex sequence signatures from triplet code sequences and base sequences of plasmid DNA fragments.
• Performance: Achieves area under the curve (AUC) values of 84%–91% for classifying transmissible fragments, including short DNA fragments of only hundreds of base pairs.

Scientific Applications:

• Horizontal Gene Transfer Studies: Identifies transmissible plasmid fragments to study mechanisms and prevalence of horizontal gene transfer in microbial communities.
• Antibiotic Resistance and Virulence Surveillance: Enables detection of conjugative or mobilizable plasmid fragments that may mediate dissemination of antibiotic resistance genes and virulence factors.
• Plasmidome Diversity and Regulation: Facilitates quantitative analysis of sequence signatures to elucidate plasmid diversity and potential regulatory roles within microbial ecosystems.

Methodology:

Calculates information entropy of synonymous codon usage across 4424 plasmid genomes and trains/applies a convolutional neural network on triplet code sequences and base sequences of plasmid DNA fragments to classify transferability, taking FASTA input and producing tabular transferability annotations.