Why Bacterial Genomes Fail to Circularize: Diagnosis, Troubleshooting, and Rescue

A complete, circularized bacterial genome—one contig per replicon, no gaps, no ambiguities—is the gold standard for strain characterization, plasmid identification, and regulatory submissions. But many projects stall at the final step: the assembly graph refuses to resolve into a single circle. This troubleshooting guide walks through why bacterial genomes fail to circularize, how to diagnose the specific cause, and what to do before ordering another sequencing run.

Key takeaways

• Repeats longer than your reads break assemblies. rRNA operons, insertion sequences, and transposons are the most common culprits. If a repeat exceeds the read length, the assembler cannot determine which copy goes where.
• Mixed cultures—even at 5%—wreck circularization. Strain-level mixtures are worse than species-level contamination because the assembler cannot separate closely related subpopulations.
• DNA damage mimics sequencing gaps. Freeze-thaw cycles, residual ethanol, and shearing during extraction create false breakpoints that no assembler can close.
• A k-mer spectrum is the fastest diagnostic. A shoulder peak means repeats. A low-frequency secondary peak means contamination. A single clean peak means the problem is elsewhere.
• Most non-closing genomes can be rescued. Hybrid assembly with long reads, selective subsampling, or manual gap closure fixes the majority of cases without re-culturing.

When Genomes Refuse to Close

A bacterial genome assembly that produces 50 contigs instead of one closed chromosome is not necessarily a failed project. It is a signal—and the signal points to one of a small number of root causes. Understanding which one applies to your dataset is the difference between a wasted sequencing run and a completed genome. The cost difference is substantial: re-sequencing a genome that could have been rescued with the right diagnostic step wastes both budget and months of project time.

Wick and colleagues demonstrated in 2017 that hybrid assembly—combining Illumina short reads for accuracy with long reads for contiguity—can reliably close most bacterial genomes. The approach has since become standard practice: short reads provide base-level accuracy while long reads resolve the structural complexity that short reads alone cannot address. Yet even with both data types, a subset of genomes resist circularization. The three most common reasons, in order of frequency, are repetitive elements that exceed read length, mixed or contaminated input cultures, and DNA that was damaged before library preparation began.

What Repeats Do to Assembly Graphs

The most common reason a bacterial genome fails to circularize is that it contains one or more repetitive elements longer than the sequencing read length. The assembler encounters identical sequence at two or more genomic locations and cannot determine unique flanking context. In the assembly graph, this appears as a node with multiple incoming and outgoing edges—a branch point that the algorithm cannot resolve.

In bacteria, the major repeat classes that cause this problem are well characterized. rRNA operons—typically 5–7 kb and present in one to 15 nearly identical copies—are the leading cause. Insertion sequence (IS) elements, ranging from 0.7 to 2.5 kb with up to hundreds of copies per genome in some species, produce highly fragmented assemblies. Transposons, prophage remnants, and large segmental duplications round out the list. A 2024 study of Porphyromonas gingivalis reference strains found that fragmentation in draft genomes linearly correlated with the number of genomic repeats larger than the read length, and that PCR-based contiguity verification of rrn operons identified wrongly assembled regions even in genomes previously reported as complete.

If you are working with a genome that has already been sequenced with short reads only, the practical fix is adding long reads—PacBio HiFi or Oxford Nanopore—that span the repeat length. A single multiplexed MinION flow cell can close most bacterial genomes when combined with existing Illumina data through a hybrid assembler. For researchers planning a new project, starting with long reads—through Nanopore-based microbial genome sequencing or PacBio SMRT platforms—avoids the repeat problem entirely. Full-length plasmid sequencing services that use long-read platforms can simultaneously resolve chromosomal repeats and recover complete plasmid sequences in a single run.

Mixed Cultures and the Coverage Problem

When a contaminated culture is sequenced, the assembler cannot cleanly separate sequences from closely related strains, producing low-coverage contigs that inflate the total assembly size to twice the expected genome size or more. Strain-level mixtures inflict more damage than species-level contaminants because the assembler attempts to reconcile sequences that are 99.5% identical but come from different genomes, producing a fragmented assembly that reflects neither subpopulation accurately. Orakov and colleagues quantified the scale of the problem in 2021 when their GUNC tool detected chimerism or contamination in 5.7% of genomes in GenBank and 5.2% in RefSeq, with rates reaching 15–30% among metagenome-assembled genomes that had already passed standard quality filters. The diagnostic signature is unmistakable in a k-mer frequency spectrum: a secondary low-frequency peak sits to the left of the main coverage peak, representing contaminant-derived k-mers at lower abundance.

Before ordering re-sequencing, run a k-mer spectrum analysis on the existing reads. If a contaminant signal is present, k-mer abundance filtering—removing reads that contain only low-frequency k-mers—can dramatically improve the assembly without additional sequencing. In some cases, subsampling reads to 20–50% of the original dataset paradoxically improves assembly by diluting the contaminant signal relative to the dominant genome. For projects that need the cleanest possible starting material, plasmid and BAC clone whole-genome sequencing from a single-colony pick eliminates mixed-culture ambiguity at the source.

DNA Quality Before the Sequencer

DNA damage and degradation create assembly problems that no algorithm can fix, because the information is physically gone. The most common pre-analytical issues include freeze-thaw cycles that shear high-molecular-weight DNA into fragments too short for long-read library preparation, residual ethanol from precipitation steps that inhibits enzymatic reactions, and mechanical shearing during vortexing or pipetting.

A genome that fails to circularize due to DNA quality typically produces contigs that end abruptly at non-repetitive sequence—the assembler simply runs out of reads that span a particular region. This is distinct from a repeat-driven break, which occurs at identifiable repeat boundaries. Checking the fragment size distribution before library preparation is the simplest preventive measure: for long-read sequencing, the DNA should show a predominant band above 30 kb with minimal smear below 10 kb. If the input DNA is degraded, no amount of sequencing depth will close the gaps.

For difficult organisms—Gram-positive bacteria with thick cell walls, biofilm-forming species, or organisms that produce copious nucleases—gentle enzymatic lysis protocols that avoid mechanical shear are critical. Lysozyme-based lysis for Gram-positives, proteinase K digestion for nuclease-rich samples, and plug-based extraction for ultra-long-read library preparation can each recover DNA integrity where standard column-based kits fail. When the DNA extraction step is the suspected bottleneck, optimized microbiome sample preparation protocols tailored to the organism type can recover higher-integrity DNA for downstream assembly.

Plasmid Ambiguity and Copy-Number Confusion

Even when the chromosome circularizes cleanly, a small contig may remain unplaced. This contig is often a plasmid—but not always in the straightforward way it appears. Plasmids complicate assembly because they exist at different copy numbers than the chromosome, and because identical plasmid sequences can integrate into the chromosome or exist in both integrated and episomal forms.

Coverage depth is the most informative diagnostic. Plasmid contigs typically have either higher or lower coverage than the chromosome—higher for multi-copy plasmids, lower for large low-copy plasmids. A contig with coverage that matches the chromosome depth but refuses to circularize may represent a plasmid that shares terminal repeats with another replicon, creating an unresolvable branch in the assembly graph. Hybrid assembly with Unicycler explicitly models plasmid copy number during the bridging step, which is why it resolves many plasmid ambiguities that short-read-only assemblers leave fragmented. Schäfer and colleagues published a practical Galaxy workflow in 2024 that guides researchers through manual curation of Unicycler assembly graphs to correct inter-plasmidic repeat collapse and restore single-copy sequences that were erroneously dropped during automated assembly.

For labs that need plasmid sequences verified independently of the chromosome, copy number analysis provides quantitative confirmation of plasmid abundance, helping to distinguish between genuinely independent replicons and assembly artifacts.

What to Do Before Re-Sequencing

FAQ

My assembly has 120 contigs. Is it worth trying to close it, or should I start over?

Start by checking whether the total assembly size matches the expected genome size. If the total is roughly 1× the expected size, the genome is likely complete but fragmented—adding long reads will close most of the gaps. If the total is 1.5× or larger, contamination or a mixed culture is likely present, and k-mer filtering may rescue the assembly without re-sequencing. If the total is well below the expected size, DNA was likely degraded before library preparation, and re-extraction is the correct next step.

How do I know if the problem is repeats or contamination?

A k-mer frequency spectrum distinguishes them cleanly. A shoulder peak to the right of the main coverage peak indicates repeats (k-mers present at 2× or higher copy number). A secondary low-frequency peak to the left of the main peak indicates contamination (k-mers from a different genome at lower abundance). No secondary peaks but persistent breaks at non-repeat boundaries points to DNA damage.

Can I close a genome with only short-read data?

Occasionally, if the genome is small, has few repeats, and was sequenced deeply from a pure culture. But for most bacteria, at least one long-read sequencing run is required to span rRNA operons and IS elements—see our guide to choosing microbial sequencing methods for platform comparisons. The cost of a single MinION flow cell multiplexed across several genomes is typically far lower than the labor cost of manual gap closure attempts on short-read assemblies.

What if my assembly looks complete but I suspect it is misassembled?

Check for uniform read coverage across the entire assembly—systematic coverage drops can indicate a missed region or collapsed repeat. Run a self-dot-plot of the assembled sequence to look for off-diagonal matches that indicate duplicated or rearranged regions. And verify that all expected single-copy marker genes are present at exactly one copy each; duplicated or missing markers are strong evidence of assembly errors.

When should I hand this off to a sequencing service instead of troubleshooting further?

If you have completed the six-step diagnostic workflow described above and the genome still resists closure, or if the organism is known to be challenging (high-GC, extensively methylated, massive genome, extreme repeat density), a service provider with experience in difficult bacterial genomes can design a tailored strategy—often combining ultra-long reads, Hi-C scaffolding, or optical mapping—that resolves issues beyond the reach of standard assembly pipelines.

References

1. Wick RR, Judd LM, Gorrie CL, Holt KE. Unicycler: Resolving bacterial genome assemblies from short and long sequencing reads. PLOS Computational Biology. 2017;13(6):e1005595. doi:10.1371/journal.pcbi.1005595
2. Orakov A, Fullam A, Coelho LP, Khedkar S, Szklarczyk D, Mende DR, Schmidt TSB, Bork P. GUNC: detection of chimerism and contamination in prokaryotic genomes. Genome Biology. 2021;22:178. doi:10.1186/s13059-021-02393-0
3. Acuña-Amador L, Primot A, Cadieu E, Roulet A, Barloy-Hubler F. Genomic repeats, misassembly and reannotation: a case study with long-read resequencing of Porphyromonas gingivalis reference strains. BMC Genomics. 2018;19:54. doi:10.1186/s12864-017-4429-4
4. Schäfer L, Jehle JA, Kleespies RG, Wennmann JT. A practical guide and Galaxy workflow to avoid inter-plasmidic repeat collapse and false gene loss in Unicycler's hybrid assemblies. Microbial Genomics. 2024;10(1):001173. doi:10.1099/mgen.0.001173
5. Wick RR, Howden BP, Stinear TP. Autocycler: long-read consensus assembly for bacterial genomes. Bioinformatics. 2025;41(9):btaf474. doi:10.1093/bioinformatics/btaf474
6. Liang H, Zou Y, Wang M, et al. Efficiently constructing complete genomes with CycloneSEQ to fill gaps in bacterial draft assemblies. GigaByte. 2025;2025:gigabyte154. doi:10.46471/gigabyte.154
7. Feng X, Li H. Evaluating and improving the representation of bacterial contents in long-read metagenome assemblies. Genome Biology. 2024;25(1):92. doi:10.1186/s13059-024-03234-6
8. Rashid U, Wu C, Shiller J, Smith K, Crowhurst R, Davy M, Chen T-H, Carvajal I, Bailey S, Thomson S, Deng CH. AssemblyQC: a Nextflow pipeline for reproducible reporting of assembly quality. Bioinformatics. 2024;40(8):btae477. doi:10.1093/bioinformatics/btae477