Transduction is a form of horizontal gene transfer, the phenomenon that allows bacteria to acquire genes from organisms other than their own offspring. Unlike transformation, in which cells take up naked DNA from their environment, or conjugation, which relies on direct cell-to-cell contact, transduction depends on virus-like particles, or VLPs. These particles, produced by bacteria themselves or by the viruses that infect them, can accidentally encapsulate fragments of bacterial DNA instead of viral genomes. When the particles deliver that cargo to new cells, they can spread traits such as antibiotic resistance, virulence factors, and metabolic capabilities across microbial communities with striking efficiency.

Detecting this hidden traffic of DNA has been the province of transductomics, a sequencing-based method introduced to catalog the DNA carried by VLPs in a sample. The approach is conceptually elegant. Researchers first purify virus-like particles from a biological sample, such as a fecal pellet, and sequence the DNA inside them. They then assemble metagenomic contigs, longer stretches of DNA reconstructed from the whole microbial community in the same sample. By mapping the VLP sequencing reads back onto those community contigs, researchers can see which pieces of bacterial DNA were packaged into particles. The telltale signature lies in the read coverage patterns: different shapes of coverage along a contig reveal whether the DNA was packaged randomly, from a provirus integrated in a bacterial genome, or through other mechanisms.

The problem, until now, has been that interpreting those coverage patterns required a trained human eye. Every contig had to be inspected and classified manually, a process that could consume hours of expert attention per dataset. For studies involving many samples, or time courses with dozens of conditions, the manual bottleneck rendered transductomics effectively unfeasible at scale. It also introduced a reproducibility concern, since different analysts might classify borderline coverage patterns differently, and even the same analyst might vary from day to day.

TrIdent addresses this bottleneck with a pattern-matching algorithm that automates the classification step entirely. The software systematically compares the read coverage profile of each contig against a library of expected patterns corresponding to distinct packaging mechanisms, assigning each contig to a class that reflects how its DNA was likely mobilized. Because the classification follows explicit, deterministic rules, every run of the tool on the same data produces identical results, a property the manual workflow could never guarantee. The package is platform independent, requires no specialized dependencies beyond the R environment, and is distributed under an open-source GPL-2 license through Bioconductor, making it straightforward for any laboratory to adopt.

Because no equivalent software existed, the development team faced an unusual validation challenge: how do you benchmark an algorithm against a standard that is itself manual? Their solution was to compare TrIdent’s classifications against those of experienced human classifiers. On a previously generated transductomics dataset that had already been carefully classified by hand, TrIdent’s assignments were generally comparable to the manual ones. More rigorously, when the team applied the tool to newly generated transductomics data from the mouse gut microbiota, TrIdent agreed with two independent human classifiers about as much as those two human classifiers agreed with each other. In other words, the algorithm performed at the level of inter-human consistency, the practical ceiling for a task that inherently involves judgment calls.

Speed and reproducibility were equally decisive advantages. TrIdent classified complete transductomics datasets in a fraction of the time required by human classifiers, turning what had been a per-sample slog into a routine computational step. The authors emphasize that this efficiency does not merely save time; it changes the kind of science that becomes possible. Studies that were previously constrained to a handful of samples can now encompass dozens of conditions, time points, and biological replicates, allowing researchers to ask how transduction responds to antibiotics, infection, diet, or other perturbations across entire experiments rather than isolated snapshots.

To demonstrate the tool’s power, the researchers turned TrIdent loose on transductomics datasets generated from murine fecal pellets, including samples collected before and after antibiotic treatment and in a model of Clostridioides difficile infection. The analysis yielded a genuinely surprising biological insight: bacterial DNA associated with two specific bacterial families, the Oscillospiraceae and the Turicibacteraceae, was highly enriched in the DNA packaged by virus-like particles compared with the same families’ abundance in the whole-community metagenomes. Both families are relatively low-abundance members of the murine gut community, yet their DNA appeared disproportionately often inside VLPs. This finding suggests that certain scarce bacteria may be heavily involved in transduction, punching far above their numerical weight in the horizontal movement of genes through the gut ecosystem.

The implications extend well beyond the mouse intestine. Horizontal gene transfer is a central force shaping microbial evolution, and transduction in particular is implicated in the dissemination of antibiotic resistance genes, one of the most pressing public health challenges of the era. A tool that makes transductomics fast, scalable, and reproducible opens the door to systematic surveys of DNA mobilization in soil, ocean, clinical, and human-associated microbiomes. Researchers studying phage therapy, microbiome engineering, or the spread of virulence factors can now incorporate transductomics into standard pipelines without needing a specialist to hand-annotate every contig.

The work was carried out by Jessie L. Maier, Craig Gin, Jorden Rabasco, Yixuan Yang, Avery Bass, Wynter Spencer, Breck A. Duerkop, Benjamin Callahan, and Manuel Kleiner, with teams spanning North Carolina State University and the University of Colorado Anschutz Medical Campus. The research was supported by a seed grant from the North Carolina State University Data Science Academy and by the National Institutes of Health. The authors have released the software with full documentation, and the underlying study is published open access, reflecting a broader movement in microbiology toward transparent, reusable computational methods. As transductomics datasets accumulate across laboratories worldwide, tools like TrIdent are poised to become as indispensable to the field as the sequencing machines that generate the data, turning a once-manual art into an automated science and revealing, contig by contig, the hidden gene highways that connect the microbial world.

Subject of Research: Automated transductomics analysis of virus-like particle mediated DNA mobilization in microbiomes

Article Title: TrIdent—an R package to automate transductomics analysis of virus-like particle mediated DNA mobilization

Article References: Maier, J. L., Gin, C., Rabasco, J., Yang, Y., Bass, A., Spencer, W., Duerkop, B. A., Callahan, B., & Kleiner, M. (2026). TrIdent—an R package to automate transductomics analysis of virus-like particle mediated DNA mobilization. Microbiome. https://doi.org/10.1186/s40168-026-02546-y

Image Credits: AI Generated

DOI: 10.1186/s40168-026-02546-y

Keywords: transduction, horizontal gene transfer, transductomics, virus-like particles, R package, microbiome, metagenomics, gut microbiota, DNA mobilization, pattern-matching, bioinformatics, antibiotic resistance