Julian Dicken

I am a PhD student at CompMicroLab supervised by Laura Carroll and Johan Henriksson.

Microbial communities are found almost everywhere, in the human gut, soil and lakes, around deep-sea hydrothermal vents, and even in toxic mine drainage. Metagenomic sequencing studies these communities by extracting and sequencing all of the DNA in a sample at once. From this single experiment, researchers can work out which organisms are present, how abundant they are, and what genes they carry, giving a snapshot of an entire community directly from its natural environment. This approach has transformed microbiology, but it has an important limitation. Because DNA from all the cells is pooled together, it is not possible to tell which fragments of DNA came from which cell. Genomes must instead be pieced back together from the mixture, a process that can blend closely related strains into a single composite and miss rare organisms altogether. This loss of cell identity also matters for mobile genetic elements, pieces of DNA that bacteria may pass between one another and that often carry genes for antibiotic resistance. Bulk sequencing can detect these elements but usually cannot determine which organisms carry them, whether they sit on a separate piece of DNA or have been integrated into the host's own genome, or how many copies each cell contains. As a result, fundamental research questions remain open. We still know little about how antibiotic resistance spreads through a community, how much of its diversity, in rare strains and in organisms that have never been characterised, is still hidden from us, or how the two are connected, for instance whether these hidden organisms act as reservoirs of resistance.

Single-cell metagenomic (scMetaG) sequencing solves these problems by keeping every piece of sequenced DNA associated with the cell it came from. Genomes can be traced back to real cells, allowing individual strains and rare or previously unknown organisms to be resolved, and mobile genetic elements can be linked directly to the cells that carry them.

However, the technology is rapidly scaling towards millions of cells per sample, far beyond what existing single-cell genomics tools can handle, and bulk metagenomic tools were never designed for cell-resolved data at all. Single-cell data also has its own technical artefacts, such as uneven amplification of DNA, chimeric reads, and contamination between cells. Co-assembly addresses several of these challenges together. By combining the data from related cells while keeping track of which cell each read came from, it can recover genomes even from cells whose DNA was poorly amplified, and build pangenomes that capture the genetic variation between cells of the same species.

My PhD project is about developing the software needed to analyse this data. This includes, for example, a preprocessing pipeline, a de novo co-assembler for recovering genomes and pangenomes, and tools for genome annotation. I write these tools in Rust using succinct data structures, sketching methods, and other techniques that compress and summarise data without losing the information needed for analysis, so that tasks that would overwhelm conventional tools become fast and practical at this scale. I will evaluate each tool on simulated datasets and mock communities with known composition.