Assembly Completeness and Continuity

  • Assembly completeness and continuity are two important characteristics used to evaluate the quality of genome assemblies in bioinformatics. Genome assembly reconstructs long DNA sequences from shorter sequencing reads, producing contigs, scaffolds, or chromosome-scale sequences. However, an assembly is not considered reliable simply because it contains long sequences or a large amount of DNA. Researchers must also determine whether the assembly represents the expected genomic content and whether its sequences are sufficiently continuous and correctly organized. Completeness describes how much of the expected biological sequence has been recovered, whereas continuity describes how extensively the recovered sequence has been assembled into uninterrupted regions.
  • Assembly completeness refers to the extent to which an assembled genome contains the expected genomic regions, genes, and other biological sequences. A complete assembly should represent most or all of the genomic content that was intended to be reconstructed. Missing regions may include genes, repetitive elements, regulatory sequences, structural regions, or other parts of the genome. In practice, completeness is assessed using different types of evidence, including conserved gene sets, reference genomes, k-mer analysis, sequencing coverage, and known genome characteristics. No single method can establish that every nucleotide has been recovered correctly, so several complementary approaches are often required.
  • Assembly continuity refers to the extent to which genomic sequences are assembled into long, uninterrupted contigs or scaffolds. A highly continuous assembly contains relatively long sequences and fewer breaks. A fragmented assembly contains many short contigs or scaffolds, indicating that the reconstruction has been interrupted at numerous positions. Continuity is particularly important when studying long-range genomic organization, gene structures, repetitive regions, structural variations, and chromosome-scale relationships. However, a highly continuous assembly is not necessarily complete or accurate. An assembly program can produce long sequences while incorrectly joining regions, collapsing repeats, or introducing structural errors.
  • Completeness and continuity therefore measure different aspects of assembly quality. An assembly may be highly complete but fragmented into many short contigs. Conversely, it may contain a small number of very long contigs but omit substantial portions of the genome. A reliable genome assembly should aim for high completeness, good continuity, accurate nucleotide sequences, and correct structural organization. These properties should be evaluated separately rather than treating one metric as a substitute for all the others.
  • Several statistical measures are used to describe assembly continuity. Contig count represents the number of continuous assembled sequences, while scaffold count represents the number of larger structures formed by ordering and orienting contigs, sometimes with gaps between them. A lower number of contigs or scaffolds can indicate improved continuity, although it does not automatically demonstrate better accuracy. The total assembly length is another important measure because it can be compared with an expected genome size or a suitable reference assembly. An assembly that is substantially shorter than expected may indicate missing sequence, whereas an unexpectedly large assembly may contain duplicated haplotypes, uncollapsed allelic regions, contamination, or other redundant sequences.
  • The N50 statistic is one of the most widely used measures of assembly continuity. To calculate N50, assembled sequences are arranged in descending order of length, and their lengths are added until at least half of the total assembly length has been reached. The length of the sequence that crosses this halfway point is the N50 value. A higher N50 generally indicates that a larger proportion of the assembly is contained in longer sequences. N50 can be calculated separately for contigs and scaffolds, and these values should be reported with their corresponding sequence types.
  • For example, consider an assembly containing five contigs with lengths of 500,000, 300,000, 150,000, 100,000, and 50,000 base pairs. The total assembly length is 1,100,000 base pairs, and half of this total is 550,000 base pairs. The largest contig contributes 500,000 base pairs, which is not sufficient to reach half of the assembly length. Adding the second contig brings the cumulative length to 800,000 base pairs. Therefore, the N50 is 300,000 base pairs, the length of the contig that crosses the halfway threshold.
  • The L50 statistic complements N50 by reporting the minimum number of longest sequences required to cover at least half of the total assembly length. In the example above, the two longest contigs are sufficient to exceed half of the assembly length, so L50 is two. A lower L50 generally indicates that fewer sequences are needed to account for half of the assembly. N50 and L50 should be interpreted together because they describe related aspects of continuity but do not measure sequence accuracy or completeness.
  • Other useful continuity statistics include the largest contig, largest scaffold, mean sequence length, median sequence length, total number of contigs, total number of scaffolds, and the number of sequences above specified length thresholds. Researchers may also report NG50 and LG50, which use an estimated or independently established genome size rather than the total assembled length as the reference for calculating the halfway point. This distinction is important when an assembly is incomplete. If the assembly is substantially shorter than the expected genome, an N50 calculated from the assembled sequence alone may give an overly favourable impression of continuity. NG50 provides a way to evaluate sequence continuity against the expected genome size, although its interpretation depends on the accuracy of that genome-size estimate.
  • Assembly completeness can be evaluated using conserved single-copy orthologs. These are genes that are expected to be present in most members of a particular taxonomic group and are commonly used as indicators of genome content. BUSCO (Benchmarking Universal Single-Copy Orthologs) is a widely used tool that searches an assembly or gene set for expected conserved orthologs from a selected lineage dataset. BUSCO results classify detected genes into categories such as complete single-copy, complete duplicated, fragmented, and missing. A high proportion of complete BUSCO genes generally indicates that much of the expected conserved gene content has been recovered.
  • BUSCO is useful for evaluating completeness in many animal, plant, fungal, and microbial genomes, but its results must be interpreted carefully. The selected lineage dataset should be appropriate for the organism being studied. A low BUSCO completeness score may indicate missing sequence, but it can also reflect genuine biological differences, unusual gene content, highly divergent genes, or limitations in gene prediction. Similarly, duplicated BUSCO genes may indicate uncollapsed haplotypes, gene duplication, polyploidy, or genuine biological duplication rather than an assembly error. BUSCO completeness is therefore an indicator of conserved gene recovery, not a direct measurement of the percentage of every nucleotide in the genome that has been assembled.
  • Other tools provide complementary completeness assessments. CheckM and CheckM2 are commonly used to estimate completeness and contamination in microbial genomes, particularly metagenome-assembled genomes. They use marker-gene information to assess whether expected genomic components are present and whether markers occur in patterns suggesting contamination or duplication. CheckM2 uses machine-learning-based approaches for genome quality assessment. These tools are particularly relevant to microbial genome reconstruction, although their estimates depend on the organism, available reference information, and the suitability of the marker sets.
  • K-mer-based approaches provide another way to investigate assembly completeness. A k-mer is a short sequence of k consecutive nucleotides. K-mers are counted in sequencing reads and compared with those found in the assembled genome. If many reliable read k-mers are absent from the assembly, this may indicate missing sequence or assembly errors. K-mer analysis can also reveal duplicated sequence, collapsed repeats, and differences between expected and observed genomic content. Merqury is a commonly used tool for evaluating genome assemblies through k-mer comparisons, particularly when reliable high-accuracy sequencing reads are available. K-mer-based methods can provide information about consensus accuracy and completeness, but their interpretation depends on sequencing quality, coverage, genome complexity, and the treatment of repetitive sequences.
  • Reference-based assessment is useful when a closely related, high-quality reference genome is available. The assembled sequences can be aligned against the reference to examine how much of the reference is represented, whether large genomic regions are missing, and whether the assembly contains structural differences. Alignment tools such as minimap2 and MUMmer can help identify sequence correspondence, rearrangements, duplications, and gaps. Reference-based evaluation is not always appropriate for highly divergent species or organisms without a suitable reference. A reference genome may also contain its own errors, omissions, or population-specific differences. Consequently, reference alignment should not be treated as an absolute measure of completeness.
  • Genome size estimation provides an additional way to evaluate assembly completeness. The total length of an assembly can be compared with genome-size estimates obtained from flow cytometry, k-mer analysis, published studies, or other independent measurements. A large difference between expected and assembled genome sizes may indicate missing sequence, redundant haplotypes, contamination, or differences in genome representation. However, genome-size estimates themselves can be uncertain, particularly in species with large repetitive genomes, high heterozygosity, polyploidy, or complex genome organization. Genome-size comparisons should therefore be interpreted alongside other quality assessments.
  • Sequencing coverage is also relevant to completeness and continuity. Coverage describes how many sequencing reads, on average, represent each position in the genome. Insufficient or uneven coverage can prevent assembly algorithms from reconstructing certain regions. Low-coverage regions may remain fragmented or absent, while repetitive regions may receive ambiguous coverage because reads cannot be assigned to individual repeat copies. High coverage can improve assembly in some circumstances, but excessive duplication, uneven sequencing, contamination, or high sequence heterogeneity can complicate reconstruction. Coverage should therefore be evaluated across the assembly rather than summarized only by a single average value.
  • Different sequencing technologies influence the completeness and continuity of genome assemblies. Short-read technologies, such as Illumina sequencing, generally provide highly accurate reads but may struggle with long repetitive regions, complex structural variants, and regions that require long-range information. Long-read technologies, including PacBio HiFi and Oxford Nanopore sequencing, can span many repetitive or structurally complex regions and often improve contig continuity. PacBio HiFi reads combine long-read sequencing with high consensus accuracy, while Oxford Nanopore technologies can generate very long reads that are useful for resolving large repetitive regions and structural arrangements. Hybrid approaches combine short- and long-read data to benefit from complementary strengths, although the results depend on data quality, coverage, and the assembly strategy.
  • Genome assembly software also influences continuity and completeness. Tools such as SPAdes and MEGAHIT are widely used for short-read and metagenomic assembly, while Flye, Canu, Hifiasm, Shasta, and Verkko support different long-read or haplotype-aware assembly strategies. Unicycler is commonly used for bacterial genome assembly, particularly in workflows combining short- and long-read data. These programs use different algorithms and assumptions, so the best choice depends on genome size, sequencing technology, ploidy, heterozygosity, repeat content, and the biological objective. No assembler can guarantee a complete or error-free genome under all conditions.
  • Several tools are available for evaluating genome assembly quality. QUAST (Quality Assessment Tool for Genome Assemblies) reports statistics such as total assembly length, contig count, largest contig, N50, and other continuity measures. When a suitable reference genome is supplied, it can also provide reference-based comparisons. BUSCO evaluates conserved gene completeness, while Merqury uses k-mer-based methods to assess assembly quality and consensus accuracy. CheckM2 is particularly useful for assessing microbial genome completeness and contamination. BlobToolKit and related approaches can help identify possible contamination by combining sequence composition, coverage, and taxonomic information. Bandage can be used to visualize assembly graphs and investigate alternative paths, repeats, and unresolved connections.
  • A basic assembly statistics report can be generated using QUAST. The following command illustrates how to evaluate an assembled genome in FASTA format:

quast.py assembly.fasta -o quast_results

  • The output directory contains reports describing assembly length, contig statistics, N50, and other metrics. If a suitable reference genome is available, it can be included in the assessment:
  • quast.py assembly.fasta -r reference.fasta -o quast_reference_results
  • The reference-based report can help identify regions that are missing, rearranged, duplicated, or inconsistent with the selected reference. However, differences may reflect genuine biological variation rather than assembly errors.
  • BUSCO can be used to assess conserved gene completeness. The following example illustrates a genome-mode analysis using an appropriate lineage dataset:

busco \

  -i assembly.fasta \

  -m genome \

  -l mammalia_odb \

  -o busco_results \

  -c 8

  • The lineage dataset must be selected according to the organism being analysed. The example uses a mammalian lineage and is not suitable for organisms outside that group. BUSCO results should be interpreted together with taxonomic context, gene prediction quality, and other assembly metrics.
  • For k-mer-based assessment, Merqury can compare an assembly with k-mer information derived from high-quality sequencing reads. A simplified workflow may begin by generating a read k-mer database and then evaluating the assembly:

meryl count k=21 output reads.meryl reads.fastq

merqury.sh reads.meryl assembly.fasta merqury_results

  • The exact commands, input requirements, and recommended parameters depend on the Merqury and Meryl versions and the sequencing data. Paired-end reads, multiple read files, compressed inputs, and different data types may require adjustments to the workflow. The k-mer size should also be selected with consideration of genome complexity and sequencing depth.
  • A simple Python script can calculate basic continuity statistics from a FASTA assembly using Biopython. The following example calculates total length, sequence count, largest sequence, N50, and L50:

from Bio import SeqIO

assembly_file = “assembly.fasta”

lengths = [

    len(record.seq)

    for record in SeqIO.parse(assembly_file, “fasta”)

]

if not lengths:

    raise ValueError(“No sequences found in the assembly file.”)

total_length = sum(lengths)

sorted_lengths = sorted(lengths, reverse=True)

half_length = total_length / 2

cumulative_length = 0

n50 = 0

l50 = 0

for length in sorted_lengths:

    cumulative_length += length

    l50 += 1

    if cumulative_length >= half_length:

        n50 = length

        break

print(“Number of sequences:”, len(lengths))

print(“Total assembly length:”, total_length)

print(“Largest sequence:”, max(lengths))

print(“N50:”, n50)

print(“L50:”, l50)

  • This script calculates N50 and L50 using the total assembled sequence length. It does not calculate genome-size-based NG50 or LG50, and it does not assess biological completeness, contamination, nucleotide accuracy, or structural correctness. It is intended as a simple continuity-statistics example rather than a complete genome quality assessment pipeline.
  • Completeness and continuity are particularly important in different biological applications. In microbial genomics, a highly complete and accurate assembly can support gene identification, metabolic reconstruction, comparative genomics, and the study of antimicrobial resistance. In metagenomics, completeness is important for evaluating metagenome-assembled genomes, while contamination must also be assessed because assembled sequences may originate from multiple organisms. In plant and animal genomics, continuity is valuable for studying gene organization, repetitive regions, structural variation, chromosome evolution, and trait-associated genomic regions. In human genomics, high-quality assemblies can improve the analysis of complex genomic regions, structural variants, and sequences that are difficult to represent using older reference assemblies.
  • The distinction between contig continuity, scaffold continuity, and chromosome-level organization is also important. Contig N50 measures continuity without relying on gaps between contigs. Scaffold N50 can be much higher because scaffolding connects contigs using long-range evidence, even when the intervening sequence is unknown and represented by runs of ambiguous nucleotides such as N. Chromosome-scale scaffolding indicates that sequences have been organized at chromosome scale, but it does not necessarily mean that every chromosome is complete, gapless, or correctly phased. Researchers should report the type of sequence used for each continuity statistic and describe the evidence supporting chromosome-scale organization.
  • Haplotypes and ploidy can further complicate completeness assessment. A diploid genome contains two homologous chromosome sets, which may differ at many positions. Some assembly strategies produce separate haplotype-resolved sequences, while others collapse similar regions into a single representation. A haplotype-resolved assembly may have a total sequence length greater than that of a collapsed haploid representation because it includes both chromosome copies. This does not necessarily indicate duplication or overassembly. Similarly, duplicated BUSCO genes may reflect genuine gene duplication, polyploidy, or separate haplotypes rather than technical errors. The expected genome representation must therefore be defined before comparing assembly size and completeness.
  • Contamination is another important issue because foreign DNA can increase the total assembly length and may produce apparently continuous sequences. Contamination may originate from laboratory organisms, host DNA, environmental organisms, sequencing adapters, or other sources. Taxonomic classification, GC-content analysis, read coverage, and sequence similarity searches can help identify suspicious regions. Tools such as Kraken2, BlobToolKit, and CheckM2 can support contamination screening in appropriate workflows. However, unusual sequences should not automatically be removed because they may represent genuine genomic features, horizontally transferred genes, organelles, symbionts, or sequences that are poorly represented in reference databases.
  • Assembly continuity should also be interpreted in relation to repeat resolution. Long repetitive regions can prevent assemblers from determining the correct order and copy number of sequences. Repeats may be collapsed into a single sequence, incorrectly duplicated, or left unresolved in the assembly graph. Long reads, Hi-C data, optical mapping, and other long-range technologies can help resolve these regions, but the results still require validation. A high N50 cannot demonstrate that repetitive regions have been reconstructed correctly.
  • For reliable genome assembly evaluation, researchers should combine multiple metrics and evidence sources. Continuity statistics such as N50, L50, contig count, and scaffold count describe sequence organization. Completeness assessments such as BUSCO, CheckM2, and k-mer analysis provide information about recovered genomic content. Reference alignment can reveal missing or rearranged regions when a suitable reference exists. Read mapping, consensus quality, contamination screening, and structural validation provide additional evidence about the accuracy of the reconstruction. The final evaluation should distinguish sequence completeness, nucleotide-level accuracy, structural correctness, and continuity because these are related but independent properties.
  • Reproducibility is essential when reporting assembly quality. Researchers should document the sequencing platforms, read types, sequencing depth, assembler and software versions, parameters, polishing procedures, scaffolding methods, and evaluation tools. The expected genome size, reference genome, BUSCO lineage dataset, and k-mer parameters should also be reported where relevant. Assembly statistics should be calculated from clearly identified files, distinguishing raw contigs, polished contigs, scaffolds, and final chromosome-level sequences. Reporting only a single N50 value or completeness percentage is insufficient for a comprehensive evaluation.
  • Assembly completeness and continuity are fundamental components of genome assembly quality assessment in bioinformatics. Completeness indicates how much of the expected genomic content has been recovered, while continuity describes how effectively that content has been organized into long sequences. Metrics such as N50 and L50, together with tools such as QUAST, BUSCO, CheckM2, and Merqury, provide complementary information about assembly quality. A dependable genome assembly should be evaluated not only for continuity and completeness but also for nucleotide accuracy, structural correctness, contamination, and biological representation. Combining these assessments helps researchers select reliable genome assemblies for downstream genomic, evolutionary, medical, agricultural, and microbial research.
Author:

Leave a Reply

Your email address will not be published. Required fields are marked *