Policy Statement on Arabidopsis thaliana Reference Sequence

The first genome sequence of Arabidopsis thaliana was completed in 2000. Sequencing was accomplished by a consortium, using several different clone libraries with varying amounts of documentation with regard to which stock of the ecotype/accession Columbia (Col-0) was used in creating the libraries, and how these stocks were related to each other.

Responsibility for updating the reference genome annotation was passed from TIGR to TAIR after the TIGR5 genome release in January 2004. The following two releases (TAIR6 and TAIR7) contained large numbers of updates to gene structure and function, reflecting the continued accumulation of new transcript sequences and function data. However, updates to the underlying chromosome assemblies were not made for these releases, as there was little or no new genomic sequence data to be incorporated during this period.

In the past year the situation has begun to change rapidly, with new low-cost sequencing technologies becoming widely available. These technologies are sparking a new round of genome sequencing for Arabidopsis, mainly focused on identifying ecotype differences. Although these new sequencing projects are not primarily focused on the Columbia ecotype, they do include this ecotype as a reference standard. This new Columbia sequence data has begun to reveal a significant number of differences from the existing reference genome sequence. Many of these are likely to be errors in the original reference, while others probably simply reflect spontaneous mutations that have accumulated in different Columbia seed stocks.

This policy document aims to more rigorously define the Arabidopsis reference genome sequence and outline a standard of evidence to be used in making updates. These steps will help to maximize the usefulness of the reference sequence to the research community and ensure that any changes to the reference sequence meet a standard of evidence support endorsed by the community. This document does not address the separate but related need for a sequence resource containing a complete set of Arabidopsis genes from all sequenced accessions but is limited to how we can maximize the sequence quality for a single reference accession.

Defining the Arabidopsis reference sequence

The reference sequence is currently defined as a composite sequence from several different laboratory stocks of the Columbia ecotype. While this definition has served well enough up to this point, the increasing availability of sequences from various Columbia seed stocks has begun to reveal sequence differences even at this level. Currently there are 11 Columbia lines described in TAIR, each of which has one or more seed stocks associated with it, for a total of 30 different stocks. For the line Col-0 alone ABRC maintains 11 separate stocks submitted by different labs. While differences among these Col-0 stocks are not expected to be large, there is little doubt that they will exist, and specific examples have already been documented.

There are several advantages to agreeing on a single seed stock as the reference sequence stock. For those who choose to work with the reference stock for studies of natural variation, gene expression, etc. the advantage will be a greater certainty that the reference genome sequence accurately represents the genome sequence of the biological samples being used, thereby avoiding miscalling of SNPs and improving accuracy of microarray and tiling chip experiments and other technologies that rely on an accurate genome sequence. Identifying a single, highly inbred seed stock as the standard will also improve the consistency of the reference sequence and minimize repeated changes that update the genome first to one reference seed stock, and then another, depending on which sequence was contributed most recently or judged to be most reliable.

A good choice for the reference seed stock would be the stock that has been used to generate the largest number of publicly available clones, mutant lines, etc., since this approach will maximize the utility of these public resources. Using this criterion we propose that the Col-0 seed stock CS70000, the stock used to generate the Ecker T-DNA and cDNA collections, be designated as the reference seed stock.

Standard of evidence for updating the reference sequence

In addition to defining what the reference sequence represents, it's necessary to define a standard of evidence that will ensure that the reference genome sequence will continue to meet community expectations of quality.

Proposed evidence standards:

1. Corrections will be made if supported by two or more independently derived sequence libraries from the Columbia ecotype (and at least one of the two is derived from genomic DNA), and the correction is supported by the majority of available sequence data. Corrections will not be made if only one independently derived sequence supports the change, or if the majority of Columbia sequence data at that position does not support the change (except in the case outlined below).
2. At positions where CS70000 is found to differ from other Columbia seed stocks based on at least two independently derived CS70000 sequences (one of which is genomic), the CS70000 sequence will be adopted.
3. Quality standards will be required for sequences used to support a correction. These standards must necessarily be specific to the sequencing technology used, but will include such parameters as depth of coverage and the percentage of reads that support the correction.

Process for updating the reference sequence

Updates to the reference sequence will be carried out as a part of TAIR's regular genome releases, and the changes will be released at TAIR and propagated to NCBI RefSeq along with updates to gene structure and function that are already propagated in this way. No changes will be made to the original BAC sequence records stored in GenBank, DDBJ and EMBL as these are owned by the submitters and represent archival copies of the original experimental data from the sequencing project.

Procedure for adopting these recommendations

This document will be circulated to the following groups for comment and revision: TAIR advisory board, NAASC, MASC. Following incorporation of suggested changes, the document will be posted on TAIR for public comment for a period of three months. The broader community will be informed of the opportunity to comment through postings to TAIR and the BioNet international Arabidopsis newsgroup. At the end of that period a final round of revision will be carried out and the resulting document will be published in a refereed journal and posted on the TAIR website.