Findings

Genome databases are of great value for biomedical research, and have significantly advanced our understanding of the biology of multiple parasite species, including Plasmodium falciparum and Plasmodium vivax, the two most common malaria parasites [1, 2]. The latter genome sequence was produced by shotgun sequencing by Carlton et al. at TIGR in 2008 at five fold coverage, and is deposited at GenBank and PlasmoDB [3]. Assembly errors are inevitable when constructing genomes, and, in the case of intracellular parasites, contamination with host DNA sequence also poses a problem. Indeed, recent research has shown that many published genomes, including mammalian, contain contaminating sequence from a variety of microorganisms [4]. Considering gene prediction errors and malaria parasites specifically, Lu et al. reported that about 20 % of genes are incorrectly predicted in the P. falciparum genome database, although these are mostly due to errors arising from the gene prediction software used [5].

During a search for repetitive protein fragments in the P. vivax genome conducted on the nucleotide sequences deposited in PlasmoDB [6] we found a 47 amino acid (aa) sequence (KGQDNSADIQHSGGRSSLEGPRFEGKPIPNPLLGLDSTRTGHHHHHH) repeated a total of 17 times in several annotated contigs. A His6 tag (Fig 1A) was attached to the 3’ end, and multiple restriction enzyme sites (Fig 1B) were present within the 141 bp nucleotide sequence (AAG GGT CAA GAC AAT TCT GCA GAT ATC CAG CAC AGT GGC GGC CGC TCG AGT CTA GAG GGC CCG CGG TTC GAA GGT AAG CCT ATC CCT AAC CCT CTC CTC GGT CTC GAT TCT ACG CGT ACC GGT CAT CAT CAC CAT CAC CAT). This sequence, when run through a VecScreen search (NCBI, http://www.ncbi.nlm.nih.gov/tools/vecscreen/) shows significant similarity to the promoter probe vector pMQ354 (Fig 1C). These features suggest cloning vector sequence contamination. We performed BLASTN searches of these 17 coding sequences against whole-genome shotgun contigs (WGS) of four whole genome sequences (India VII [GenBank: AFMK01000000], North Korean [GenBank: AFBK01000000], Brazil I [GenBank: AFNI01000000], Mauritania I [GenBank: AFNJ01000000]) [7]. All hits were aligned with the reference sequence, and the results showed missing or substituted base pairs at the 3′ end of the query sequences, resulting in the absence of the correct stop codon of the parasite gene, and the incorporation of the vector sequence into the predicted parasite gene protein product, which then terminated at the vector stop codon. Considering that there may be a possibility of frame shifting, we translated the coding sequence in all three frames (Fig 1A), and frames two and three protein were used as query sequences against the PlasmoDB protein database. This resulted in five and four sequence hits respectively, and these nine sequences were subjected to alignment and correction as described before. In total, we discovered 26 sequences in PlasmoDB contaminated by the vector sequence (Table 1).

Fig. 1
figure 1

Cloning vector source sequence contamination in PlasmoDB. a: A 141 bp vector source sequence with a his6 tag repeatedly occurred in the Plasmodium vivax nucleotide database. b: Dozens of restriction enzyme sites are present in the sequence. c: VecScreen search showed the contaminating sequence strongly match to pMQ354. d: Typical errors in Sal-1 strain sequencing results due to the contaminating sequence. The missing ends are marked in yellow, and contaminating vector sequences are underlined

Full size image
Table 1 Correction of 26 genes affected by a contaminated cloning vector sequence in PlasmoDB
Full size table

Generally, cloning vector source sequences are relatively easily recognized by a variety of tools, such as VecScreen. The P. vivax database has been updated more than ten times [8], and yet this vector sequence contamination persists, suggesting that it may have special characteristics that render it difficult to identify automatically. Attempted PCR amplification of Sal-1 genomic DNA using primers specific for the potential contaminating sequence would provide definitive proof of whether these sequences really are present in the genome, a scenario we believe to be highly unlikely.

The publication of four geographical reference strain whole genome sequences now provides an opportunity for the correction of the genome sequence of the Sal-I reference genome. Given our findings, it is possible that further interrogation of the P. vivax genome deposited in PlasmoDB may reveal further contamination. It is also possible that any previous work that made use of these sequences may require reappraisal.