Tandem repeats lead to sequence assembly errors and impose multi-level challenges for genome and protein databases.
Tørresen OK, Star B, Mier P, Andrade-Navarro MA, Bateman A, Jarnot P, Gruca A, Grynberg M, Kajava AV, Promponas VJ, Anisimova M, Jakobsen KS, Linke D. · 2019
Read Tandem repeats lead to sequence assembly errors and impose multi-level challenges for genome and protein databases. by Tørresen OK, Star B, Mier P, Andrade-Navarro MA, Bateman A, Jarnot P, Gruca A, Grynberg M, Kajava AV, Promponas VJ, Anisimova M, Jakobsen KS, Linke D. (2019) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc6868369.txt.
Introduction
Tørresen and co-authors review a hidden source of error in genome databases, in Nucleic Acids Research. Tandem repeats are stretches where a short DNA pattern repeats head to tail. They occur across the tree of life. Sequencing machines and assembly software stumble over them, since reads from inside a repeat look alike.
Errors then slip into genome and protein databases, often without any flag. A gene largely made of repeats may be dropped from annotation altogether. Users downstream treat the records as ready to use, with no easy way to judge their trust. The authors walk through each stage where repeats cause trouble. They argue the problem deserves open acknowledgement and better handling.
Catalog record
| CREATOR | Tørresen OK, Star B, Mier P, Andrade-Navarro MA, Bateman A, Jarnot P, Gruca A, Grynberg M, Kajava AV, Promponas VJ, Anisimova M, Jakobsen KS, Linke D. |
|---|---|
| YEAR | 2019 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 78 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Tandem repeats lead to sequence assembly errors and impose multi-level challenges for genome and protein databases By Ole K Tørresen, Bastiaan Star, Pablo Mier, Miguel A Andrade-Navarro, Alex Bateman, Patryk Jarnot, Aleksandra Gruca, Marcin Grynberg, Andrey V Kajava, Vasilis J Promponas, Maria Anisimova, Kjetill S Jakobsen Published in: Nucleic Acids Research DOI: 10.1093/nar/gkz841 ABSTRACT Abstract The widespread occurrence of repetitive stretches of DNA in genomes of organisms across the tree of life imposes fundamental challenges for sequencing, genome assembly, and automated annotation of genes and proteins. This multi-level problem can lead to errors in genome and protein databases that are often not recognized or acknowledged. As a consequence, end users working with sequences with repetitive regions are faced with ‘ready-to-use’ deposited data whose trustworthiness is difficult to determine, let alone to quantify. Here, we provide a review of the problems associated with tandem repeat sequences that originate from different stages during the sequencing-assembly-annotation-deposition workflow, and that may proliferate in public database repositories affecting all downstream analyses. As a case study, we provide examples of the Atlantic cod genome, whose sequencing and assembly were hindered by a particularly high prevalence of tandem repeats. We complement this case study with examples from other species, where mis-annotations and sequencing errors have propagated into protein databases. With this review, we aim to raise the awareness level within the community of database users, and alert scientists working in the underlying workflow of database creation that the data they omit or improperly assemble may well contain important biological information valuable to others. INTRODUCTION The availability of DNA and protein sequence data has revolutionized the way we study cellular, molecular, physiological, evolutionary and developmental processes, allowing the association of phenotypes with genotypes at a single nucleotide (or single amino acid) resolution. Researchers rely on public sequence depositories and other databases for sharing their data, such as GenBank or UniProt, and the content of these databases has grown exponentially in the last decades. While such databases initially consisted predominantly of submissions of individual gene or protein sequences that were carefully curated, large proportions of the content of genome and protein databases today originate from different types of metagenome and genome sequencing and assembly projects. GenBank, for example, included more than 2635 Gbp (billion base pairs) in its 2017 release number 221, of which 2242 Gbp (85%) originated from whole-genome shotgun sequencing (1). For an informed use of such data, it is essential that end users understand the distinct contrast in quality between individual, well-curated submissions and entries generated from automated sequence annotation pipelines. The latter procedures can contain unrecognized errors. Here, we argue that awareness of potential database errors is especially relevant with regards to repetitive stretches of DNA, which can occur in both noncoding and coding regions of genomes. The specific nature of this type of DNA sequences can introduce and propagate bias during multiple levels of analyses, and resulting uncertainties and errors are automatically translated further into protein sequences where they become impossible to recognize. Such issues may arise from problems originating from DNA sequencing, from difficulties with assembling repetitive DNA regions and from inaccuracies generated during the annotation process. The multiplicity of these error sources makes it particularly difficult for researchers to understand and assess the bias that may be underlying the sequences that they retrieve from public databases. As an example, in Table 1, we have listed the total number of proteins in UniProtKB/Swiss-Prot that have changed the length of their repetitive region from the first occurrence in the database to the latest—suggesting that errors in repetitive region length have been identified and corrected. The average difference in length is 13.57 amino acids, a substantial number. The 1669 proteins with differences in repeats (Table 1) are 6% of all proteins in the database that have a repetitive region (see Table 2). These numbers do not reflect a true error rate but suggest that errors in repeat numbers and repeat length are frequent and might often go unnoticed, especially in databases that are less well curated than UniProtKB/Swiss-Prot. [Table: Summary of proteins from UniProtKB/Swiss-Prot where the length of repetitive region has changed between different versions of the database] [Table: Differences of repetitive region lengths in evolutionarily distinct groups of organisms] In this review, we discuss different types of sequencing and database errors, using prominent, published examples where such errors have been found. We first provide a description of the different types of repeats that occur on the DNA and protein level and an overview of DNA sequencing technologies with their benefits and limitations. We then describe the genome assembly, annotation, and database deposition processes, and then link these processes to the different types of errors that may occur at different points in this workflow. We aim to alert the ever-growing community of database end-users of these errors, and to raise awareness among the scientists working in the underlying workflow of database creation, that data that they omit or improperly assemble may well contain important biological information valuable to others. Repetitive elements in genomes Repetitive DNA occurs in all domains of life—Bacteria, Archaea and Eukaryota—and can be grouped into two categories: interspersed repeats, such as transposable elements occurring in multiple loci across the genome, and tandem repeats (TRs) that occur in a single locus. In eukaryotes, repetitive DNA also occurs in specific chromosomal regions, such as the (sub)telomeric regions (2,3) and the centromeres (4). Transposable elements (TEs) are typically several thousand base pairs (kbp) in size, and in eukaryotes their size can range from 100 base pairs (bp) to 20 kbp (5). Large fractions of vertebrate genomes are filled with active and inactive fragments of TEs, with more than 40% of the genome of zebrafish and more than a third of mammalian genomes consisting of TEs (6). Evolutionarily old TEs will accumulate mutations and will diverge from the original sequence, and TEs can therefore lose their repetitive nature over time. In contrast, TRs may consist of motifs as short as 1 bp, where the motif is repeated in tandem. Short tandem repeats (with a motif shorter than 10 bp) were originally called microsatellites (7), longer tandem repeats (with a motif between 10 and 100 bp) were called minisatellite DNA (8), and long tandem repeats (with a repeating motif longer than 100 bp) were called satellite DNA (9). In eukaryotes (based on studies done on metazoans, green algae, plants and yeast), the content of TRs with a unit size of 1–50 bp usually varies between 2000 bp/Mbp and 55 000 bp/Mbp (corresponding to 0.2–5.5% of the genome) (10,11). Repeats also lead to significant intra-specific variation (i.e. variation between individuals of the same species) (12,13) as shown in a wide range of eukaryotes, for instance Arabidopsis (13,14) and Drosophila (15). Within humans, repeats outnumber the number of bases affected by SNP variation by an order of magnitude (4–5 fold) (16). Intra-specific variation poses its own intrinsic challenges for instance when sequencing samples from pooled individuals (17). Short tandem repeats (STR) are less prevalent in bacteria compared to eukaryotes—presumably due to the typically compact bacterial genomes—but nonetheless regularly occur in bacterial coding regions (18). TEs can cause ‘breakage’ of a continuous assembly and lead to assembly collapse, where the