Contamination Screening Genome Assembly
If you assemble a genome from mixed sequencing data you risk producing a chimeric assembly that misrepresents your target organism. Contamination screening is the process of identifying and removing sequences from foreign sources before or after assembly. This guide is for researchers who assemble genomes from metagenomic samples, host associated samples, or any dataset where cross species or reagent contamination is possible. The CSI-SSU method for phylogenetic contamination screening provides a rigorous example of how to detect foreign sequences in genomic datasets.
Accurate genome assembly depends on clean input data. Even low level contamination can introduce spurious contigs, inflate genome size estimates, and mislead downstream annotation or comparative genomics. Systematic detection of Oxford Nanopore adapter and barcode contamination in GenBank shows that contamination can persist in public databases and affect many analyses. Every genome assembler should integrate contamination screening into their workflow.
At a Glance
| Aspect | Key Information |
|---|---|
| What is contamination screening? | Identifying and removing sequences that originate from organisms other than the intended target |
| Why is it important? | Prevents chimeric assemblies, inaccurate gene predictions, and false biological conclusions |
| When to screen? | On raw reads before assembly and on assembled contigs after assembly |
| Common tools | Kraken2, Centrifuge, BLAST, BlobToolKit, CheckM, taxon annotator scripts |
| Main databases | NCBI nt, RefSeq, Silva, UniVec, custom contaminant sets |
| Output | Filtered reads, decontaminated assembly, taxon assignment reports |
Core Concepts of Contamination in Genome Assembly
Contamination in genome sequencing can arise from multiple sources. Cross species contamination occurs when DNA from another organism is present in the sample (e.g. host DNA in a parasite genome or environmental bacteria in a target culture). Reagent contamination comes from laboratory reagents, such as enzymes or water, that contain microbial DNA. Barcode hopping or index swapping is a known issue in multiplexed sequencing, where reads are misassigned to a sample. The cautionary case for host assignment based on broad environmental bla(OXA) carriers illustrates how misidentification of contamination can lead to incorrect host microbe associations.
The consequences depend on the level of contamination. High level contamination may cause assembly failure or produce many contigs from the wrong organism. Low level contamination can introduce a few foreign contigs that are difficult to spot without systematic screening. In metagenomic assemblies, contamination blurs the boundary between species and complicates binning. For single isolate genomes, contamination inflates genome size and adds spurious genes.
Decision Points Before Screening
You need to decide when to screen and what reference databases to use. Screening raw reads removes contamination before assembly which can improve assembly quality and reduce computational cost. Screening the assembly afterward can detect contamination that was not caught earlier. Many practitioners perform both steps.
The choice of reference database is critical. A general database like NCBI nt covers many taxa but is large and slow. A custom database built from likely contaminants (e.g. human, common lab bacteria, vector sequences) runs faster and can be more sensitive for known threats. The EMBL EBI Training resources on quality control recommend testing multiple databases to understand false positive rates. You should also consider the taxonomic level needed: strain level resolution is rarely necessary for contamination screening, genus or family level is usually sufficient.
Stringency thresholds affect sensitivity. A low identity cutoff (e.g. 80%) will capture divergent contaminants but may assign many reads to the wrong taxon. A higher threshold (e.g. 95%) reduces false positives but may miss contamination from distantly related species. The Galaxy Training Network contamination screening tutorial suggests starting with a kmer based classifier (Kraken2) and then confirming hits with a slower alignment method.
Practical Workflow for Contamination Screening
I recommend a two pass workflow: screen raw reads, then screen the assembly.
Step 1: Initial Read Quality Assessment
Before screening, evaluate your raw reads. Compute read quality metrics (see our guide on Fastq Read Quality Metrics). Check for adapter contamination and trim adapters (see Adapter Trimming Validation). Remove low quality bases and short reads. This step reduces false positives in taxonomic classification.
Step 2: Screen Raw Reads with a Taxonomic Classifier
Run a kmer based tool such as Kraken2 or Centrifuge against a curated database that includes known contaminants. For example, the NCBI RefSeq database for bacteria, archaea, viruses, and human is a common choice. The NCBI Bookshelf provides guidance on using RefSeq for sequence classification. Examine the classification report: if more than a few percent of reads map to non target organisms, consider filtering them. Many pipelines use a cutoff of 1% to 5% contamination before flagging.
If you are working with host associated samples, first map reads to the host genome and remove those that map. This reduces host contamination and computational load. Tools like Bowtie2 or BWA can perform this mapping.
Step 3: Assembly
After filtering, assemble the remaining clean reads. Use an assembler appropriate for your data type (e.g. SPAdes for bacteria, Flye for long reads). The assembly will be less prone to chimeras if contamination was removed.
Step 4: Screen the Assembly for Contaminant Contigs
You cannot assume the assembly is clean. Use tools designed to detect contamination in assembled sequences. BlobToolKit plots contigs by GC content, coverage, and taxonomic assignment, making visually obvious outliers easy to spot. CheckM estimates genome completeness and contamination using lineage specific marker genes. A typical acceptable contamination level for a high quality bacterial genome is below 5%, though some standards require below 2%.
For metagenomic bins, the CSI SSU method screens for phylogenetic contamination using small subunit rRNA gene sequences. This approach is especially useful for protist and eukaryotic genomes where marker gene sets are less standardized.
Step 5: Verify and Remove Contaminants
Based on the screening results, remove contigs that are confidently assigned to a foreign taxon. Keep contigs that are unclassified if they cannot be reliably assigned, but flag them for further investigation. Re evaluate the assembly after removal. The Bioconductor package contamination offers functions for statistical detection of contaminant contigs based on coverage and composition.
Quality Checks and Validation
After screening, validate the quality of your final assembly. Calculate read mapping rates to the assembly: high rates (above 90%) indicate the assembly represents most of the input reads. Check that the genome size and GC content match expected values for your target organism. For bacterial genomes, the presence of a complete set of single copy marker genes is a good sign of purity.
Compare your screened assembly against an independent method. For example, resequence the sample and repeat the screening. The study on convergent GenX biodegradation by synthetic bacterial consortia used both read based and assembly based screening to confirm the absence of contamination. Such cross validation increases confidence.
Common Mistakes
Over filtering. Removing reads that are genuinely from your target organism because they resemble a reference database entry. Use loose thresholds and confirm ambiguous hits with alignment.
Using an outdated or too narrow database. If your contaminant is not in the database, you will miss it. Update databases regularly and include possible contaminants from your laboratory environment.
Ignoring adapter and barcode contamination. The systematic detection of adapter and barcode contamination in GenBank shows that these artifacts can mimic foreign sequences. Always trim adapters and check barcode assignment.
Screening only reads or only assembly. Contamination can be introduced at any stage. Screen both to be thorough.
Assuming that unclassified sequences are safe. Some novel or highly divergent organisms will not match any database. They could be real or contaminant. Use additional evidence like GC content and coverage to decide.
Limits and Uncertainty
No contamination screening method is perfect. Uncharacterized contaminants that are absent from reference databases will remain undetected. Novel organisms that are phylogenetically similar to your target may be mistakenly retained. Conversely, your target organism might be removed if it resembles a known contaminant (e.g. a non pathogenic relative of a pathogen). The resolving of plasmid encoded carbapenem resistance through nanopore sequencing highlights how plasmids can be misidentified as contamination because they share sequences with other species.
Screening results depend heavily on database completeness and algorithm settings. Reporting the database version, tool parameters, and thresholds used is essential for reproducibility. When contamination is suspected but not confirmed, consider orthogonal methods such as PCR or fluorescence in situ hybridization to check for the presence of foreign DNA in the original sample.
Frequently Asked Questions
Q: Can I skip contamination screening if I am working with a pure culture? A: No. Even pure cultures can be contaminated by reagent DNA or index hopping. Screening is a low cost safeguard that can save months of incorrect analysis.
Q: What is the best tool for screening raw reads? A: Kraken2 is fast and widely used. For higher sensitivity, use a BLAST based approach but expect longer run times. The NCBI SRA provides example datasets to test different tools.
Q: How do I decide a threshold for removing contigs? A: There is no universal threshold. Examine the taxonomic assignment scores, coverage values, and GC content. Remove contigs with strong assignment to a clearly foreign taxon. Keep contigs with weak assignments and investigate further.
Q: What if my target organism is not well represented in databases? A: Use a broader database like NCBI nt and accept a higher false positive rate. Consider aligning contigs to close relatives and checking synteny. The EMBL EBI Training offers materials on dealing with novel sequences.
References and Further Reading
- NCBI Bookshelf: free biomedical books and technical references
- EMBL EBI Training: official training resources for biological data
- Galaxy Training Network: open bioinformatics workflow training
- Bioconductor: open software and documentation for genomic analysis
- NCBI Sequence Read Archive: public repository for sequencing data
- Convergent GenX biodegradation by genomically designed synthetic bacterial consortia , example of cross validation in contamination screening
- Long read whole genome sequencing dataset of microbial communities from freshwater wetlands , case study with contamination assessment
- CSI SSU: phylogenetic contamination screening demonstrated on the Protist 10,000 Genomes Project , method for eukaryotic genomes
- A Cautionary Case for Host Assignment Based on Broad Environmental bla(OXA) Carriers , pitfalls of host microbe assignments
- Systematic detection of Oxford Nanopore adapter and barcode contamination in GenBank , evidence of contamination in public data
- Resolving plasmid encoded carbapenem resistance dynamics through nanopore sequencing , illustrating plasmid related misclassification