Natural antisense transcripts (NATs) are endogenous RNA molecules that exhibit partial or complete complementarity to other transcripts, through which they may contribute to the regulation of molecular expression at various levels. Though many natural antisense transcripts were discovered through their regulatory function on the expression of mRNAs 12, some global predictions of NATs in several species have also been published 345678910. The first of these used mRNA data to predict natural antisense transcripts 4. With the appearance of more draft genomes and full length cDNA data, the scale of NATs predictions has been extended. Several datasets, mainly based on full length cDNAs, have been published for mouse 811, rice 12 and Arabidopsis thaliana 7. Since 2006, the trend in NATs prediction has turned to multi-species comparisons 613. A number of published NATs have been validated by various experimental approaches, such as RT-PCR 10 and microarray 5, further confirming that antisense transcript is a common occurrence in eukaryote transcriptomes.

The background for the emergence of so much NAT data in recent years, is on the one hand the availability of more genomic and full length cDNA data, and on the other hand a growing realization of the important functions of natural antisense transcripts. Antisense RNAs may contribute regulatory activity at various levels, such as post-transcription 1415, splicing 1617, transport 18, and genomic imprinting1920, and have been shown to be involved in the control of developmental processes 21, adaptation to various stresses 22, and viral infection 2324 through annealing to complementary sequences.

To facilitate research, previous publications have suggested a few classification systems for NATs. The most basic of these is the cis/trans system 4 in which an antisense transcript from the same genomic loci as the sense transcript is labelled a cis-NAT, whereas a trans-NAT is an antisense transcript expressed from a genomic locus different from that of the sense transcript. A second classification system is based on the overlapping position of the complementary pair, which will be divided into 5–6 categories according to their patterns of gene structure, e.g. depending on whether the pair overlaps at their 5' ends, 3' ends, completely, or in the introns 671011. A third classification system considers the respective coding potential of the complementary pair, and includes the categories coding-coding, coding-noncoding and noncoding-noncoding 813.

Up to present, a number of large-scale NAT data have been published and several functional studies of NATs have been carried out, however, thus far no database has been set up to collect and order all these transcripts. In order to serve the need of the NAT research, we have over the past two years built the antiCODE database. The purpose of the database is to collect the existing NAT data, and to provide a useful browsing and search platform for these data. This release of antiCODE contains more than 30,000 natural sense-antisense transcript pairs from the 12 model organisms Homo sapiens (human), Mus musculus (mouse), Rattus norvegicus (rat), Xenopus tropicalis (western clawed frog), Drosophila melanogaster (fruit fly), Caenorhabditis elegans (nematode), Ciona intestinalis (seasquirt), Gallus gallus (chicken), Danio rerio (zebrafish), Bos taurus (cow), Oryza sativa (rice) and Arabidopsis thaliana (thale cress).

All NATs in the database have been collected from recent articles 4567810111213. The original datasets used for construction of the database are listed in Table 1, which include 11,287 human NAT pairs 4510, 14,199 mouse NAT pairs 811, 1,339 A. thaliana NAT pairs 7, 687 rice NAT pairs 12 and more than 5,000 NAT pairs from other species 613.

Recently, new technologies, such as microarray, SAGE, and MPSS have played prominent roles in the identification of NAT pairs. Before 2005 only EST (UniGene) and mRNAs had been used for NAT prediction. Later large scale full-length cDNA data emerged, based on which more than 1,000 rice NATs12 were first reported, closely followed by mouse 811 and Arabidopsis 7 NATs. For NAT prediction in Arabidopsis 7 also MPSS data has been used, and in 2005, a new NAT dataset based on SAGE was reported in mouse 26. In 2007, data 27 from whole-genome arrays was employed for NAT prediction in Arabidopsis. It is expected that along with the improvement in array technology, more transcripts from tilling microarrays will be used for future NAT predictions, hopefully resulting in an accurate and exhaustive set of NAT data.

The most recently released NAT datasets 9262728 have yet not been included in antiCODE, but will be included in the next release of the database. However, compared with other existing databases 29, antiCODE is presently the most comprehensive and integrated database for NAT pairs. The most distinctive features of antiCODE are as follows; (i) antiCODE includes almost all known natural antisense transcript (NAT) pairs from 12 eukaryotic model organisms, (ii) antiCODE provides substantial and compact information relating to NATs (e.g. accession number, clone ID, species, classification etc.), (iii) we have introduced a classification system based on the previous notions which should give users an immediate impression of the basic features of each NAT pair, (iv) a Blast service is provided, and (v) antiCODE provides a user-friendly interface and a convenient search option, allowing efficient investigation and verification of natural antisense pairs from different species.

The antiCODE database and related resources can be freely accessed at its websites http://bioinfo.ibp.ac.cn/ANTICODE or http://www.anticode.org

Yifei Yin and Yi Zhao carried out the design and the collection of data. Jie Wang carried for building the database. Changning Liu participated in the design of the study. Shuguang Chen helped to draft the manuscript. Runsheng Chen and Haitao Zhao participated in the design and coordination. All authors read and approved the final manuscript.