Bacterial Whole-Genome Sequencing Analysis: From Raw Reads to Comparative Genomics

Whole-genome sequencing has become the reference method for characterising bacterial isolates — identifying species, typing strains, screening for resistance and virulence, and reconstructing how isolates relate to one another in an outbreak. Because a bacterial genome is small (typically a few megabases) and usually sequenced without a reference guiding the process, the workflow centres on de novo assembly rather than alignment. This guide walks the full path from raw reads to comparative genomics.

What bacterial WGS enables

From a single sequencing run you can determine an isolate’s species, assign a standardised strain type, screen for antimicrobial-resistance and virulence genes, and — across a collection — reconstruct phylogeny and identify outbreak clusters. This breadth is why WGS has largely replaced older, single-purpose typing methods in both research and public-health microbiology. The analytical foundation for almost all of it is a good genome assembly.

Step 1 — Read quality control

Start by assessing raw reads with FastQC/MultiQC and trimming adapters and low-quality bases with fastp or Trimmomatic. For most bacteria, sequencing to around 100× depth gives ample coverage for a high-quality assembly. It is also good practice to confirm there is no unexpected contamination — for example by screening reads with Kraken2 — because a contaminated read set produces a chimeric assembly that quietly derails everything downstream.

Step 2 — De novo assembly

Assembly reconstructs the genome from overlapping reads without a reference. For short-read (Illumina) data, SPAdes is the long-standing standard, and Shovill is a popular pipeline that wraps SPAdes for speed. Short-read assemblies are typically good drafts — accurate at the base level but fragmented into contigs, because repeats longer than the read length cannot be resolved. To obtain a complete, closed genome (single circular chromosome plus plasmids), a hybrid approach that combines Illumina accuracy with Oxford Nanopore or PacBio long reads — assembled with a tool such as Unicycler — is the reliable route. The choice depends on whether you need a draft for typing and gene screening (short reads suffice) or a finished genome for plasmid and structural analysis (hybrid is worth it).

Step 3 — Assembly QC

Never analyse an assembly you haven’t checked. QUAST reports contiguity metrics — total length, number of contigs, and N50 (the length such that half the assembly sits in contigs at least that long) — which should match expectations for the species. CheckM (or BUSCO) estimates completeness and contamination using lineage-specific marker genes; a good isolate assembly is highly complete with minimal contamination. An assembly that is unexpectedly large, highly fragmented, or contaminated is a warning that reads, sample purity, or the assembly itself need revisiting before you go further.

Step 4 — Annotation

Annotation predicts and labels genomic features — protein-coding sequences, rRNAs, tRNAs, and more. Prokka has been the community workhorse for years; Bakta is a modern, standardised alternative that provides richer, database-backed annotation of genomes, MAGs, and plasmids. Consistent annotation is a prerequisite for the comparative steps that follow, because pan-genome tools compare genes across isolates using these annotations as input.

Step 5 — Typing and species identification

Typing places an isolate into a standardised, shareable classification. Multilocus sequence typing (MLST) assigns a sequence type from the alleles of about seven housekeeping genes — coarse but universally comparable, with schemes hosted at resources like PubMLST. Core-genome MLST (cgMLST) extends this idea to hundreds or thousands of core loci, giving the high resolution needed to distinguish closely related isolates in outbreak investigations; tools such as chewBBACA implement assembly-based cgMLST against published schemes. Species identification itself can be confirmed from the assembly (for example via average nucleotide identity against reference genomes). Together these provide a common language for comparing isolates across labs and studies, which is central to our infectious-disease work.

Step 6 — Pan-genome analysis

Bacteria of the same species can differ substantially in gene content, so comparative genomics starts with the pan-genome — the full complement of genes across a set of isolates. Tools such as Roary and its successor Panaroo partition genes into the core genome (present in essentially all isolates), and the accessory genome, often subdivided into soft-core, shell, and cloud categories by how widely they are shared. The core genome provides a stable backbone for phylogenetics, while the accessory genome captures the mobile, variable content — resistance cassettes, virulence factors, phage — that often explains phenotypic differences between strains.

Step 7 — Phylogenomics

Phylogenomics reconstructs how isolates are related. Two complementary approaches dominate: building a core-genome alignment (from the pan-genome analysis) and inferring a tree with a maximum-likelihood method such as IQ-TREE or RAxML; or reference-based SNP calling, mapping reads to a reference with a tool like Snippy to produce a SNP alignment for tree-building. In recombining species, recombinant regions can distort the tree, so tools such as Gubbins are used to detect and remove them first. For outbreak analysis, pairwise SNP or allele distances between isolates are compared — closely related isolates cluster tightly — though the exact threshold that defines “linked” is species- and context-dependent and should be interpreted alongside epidemiological data, not in isolation. This interpretive care is where our biostatistics support matters.

Common pitfalls

The recurring bacterial-WGS mistakes are avoidable. Analysing an unchecked assembly propagates contamination or fragmentation into every result. Undetected sample contamination produces chimeric genomes and false gene calls. Expecting complete genomes from short reads alone misjudges what a draft can deliver. Ignoring recombination in phylogenetics distorts trees. And reading outbreak links off SNP distances without epidemiology over-interprets the genomics. A documented, versioned pipeline — the kind we build for every project — keeps these in check and makes results reproducible across labs.

Virulence, plasmids, and mobile elements

Beyond typing and phylogenetics, a bacterial genome carries the determinants that explain why a strain behaves as it does. Virulence factor screening — against curated databases such as VFDB — identifies genes associated with pathogenicity (adhesins, toxins, secretion systems), using the same identity-and-coverage matching logic as resistance screening. Plasmid analysis matters because plasmids are the vehicles that spread resistance and virulence between strains and species: tools like PlasmidFinder identify plasmid replicon types, and reconstruction tools such as MOB-suite attempt to separate plasmid from chromosomal sequence and predict mobility.

These mobile genetic elements — plasmids, transposons, integrons, and prophages — are precisely the accessory-genome content that a pan-genome analysis highlights, and they are often the most epidemiologically important part of a genome. Because short-read assemblies fragment repetitive and plasmid sequences, resolving complete plasmids usually benefits from long-read or hybrid sequencing. Screening for resistance determinants specifically is a large enough topic that we cover it separately in our guide to AMR gene detection from WGS.

Choosing a sequencing platform

The sequencing technology shapes what your bacterial analysis can deliver. Illumina short reads are highly accurate at the base level and inexpensive, which makes them ideal for typing, gene screening, and SNP-based phylogenetics — but repeats longer than the read length leave the assembly fragmented into contigs, so plasmids and structural features often stay unresolved. Long-read platforms — Oxford Nanopore and PacBio — span those repeats and can close a genome into complete circular replicons, revealing genome structure and full plasmid sequences that short reads cannot.

Historically long reads carried a higher per-base error rate, which is why hybrid assembly — using long reads for structure and accurate short reads to polish base-level errors — became the reliable route to a finished, accurate genome. Long-read accuracy has improved considerably (notably PacBio HiFi and newer Nanopore chemistries), narrowing that gap. The practical decision comes back to the question: for population typing and resistance screening across many isolates, short reads are efficient and sufficient; for reference-quality genomes, plasmid biology, or structural questions, long-read or hybrid sequencing is worth the added cost.

Conclusion

Bacterial WGS turns a set of isolates into species calls, standardised types, gene content, and an evolutionary and epidemiological picture — but only through a disciplined, assembly-first workflow with quality control at every stage. Get the assembly, annotation, typing, and phylogenetics right, and the results are reproducible and shareable across the wider microbiology community. If you’d like this handled end to end, from reads to comparative genomics, our team supports the full bacterial-genomics pipeline — tell us about your isolates and we’ll scope it.

Frequently asked questions