First, RACE (5’, 3’ or both) is performed with primers (black arrows) from one or more annotated exons of known loci. Second, the RACE products are hybridized onto a tiling array, possibly across different cell types and conditions. Third, the detected sites of transcription (RACEfrags, in yellow in the figure) are used to design RT-PCR primers (black arrows). Primers are designed only on RACEfrags corresponding to previously undetected exons. Fourth, one RT-PCR reaction is performed for each primer in a novel RACEfrag, using the original RACE primer as the second primer. Fith, each RT-PCR reaction is cloned separately into a mini-pool. Finally, clones are randomly selected from the RT-PCR mini-pools and sequenced, leading to the identification of novel transcripts.

(a) RACEarray interrogation of the MECP2 locus. Probe intensity values of RACE products originating from annotated exons of the MECP2 locus hybridized into the ENCODE tiling-array including the region in which the MECP2 locus resides. Two isoforms are known for this gene. Fifteen new transcript sequences have been discovered through RACEarray normalization (cloned RT-PCR products).
(b) RACEarray interrogation of the CHAF1B locus (UCSC Genome Browser screenshot). RACEfrags are depicted in orange, RT-PCR primer pairs in cyan, sequenced RT-PCR products in purple, and index genes (“Q3 target genes” track) in blue. A 5’ RACE reaction originating from an exon of gene CHAF1B (a.k.a. NM_005441) produced the RACEfrags showed in orange. The most distal RACEfrag—overlapping an exon from upstream gene ZCWCC3 (a.k.a. NM_015358, on the same strand as CHAF1B)—was chosen for RT-PCR verification. RT-PCR products were cloned. Sixteen clones were selected at random and sequenced. Eight different sequences were obtained (under “RT-PCR products”; two end sequences from one clone could not be assembled and are not shown here). All the novel sequences connect the two loci using a variety of novel exon combinations. No previous evidence existed supporting these transcripts. For reference, various UCSC annotation tracks are also represented at the bottom of each screenshot. Some tracks (“Q3 RxFRAGS”, “Human mRNAs + ESTs”) were collapsed (“Dense” mode) for clarity purposes (a more detailed figure as well as other examples are available at http://genome.imim.es/datasets/racearrays2007/).

(a)total number of nucleotides in RACEfrags as a function of the tissue in which the RACE reaction has been performed.
(b)cumulative number of transcribed nucleotides detected by RACEfrags per tissue. The cumulative coverage is obtained iteratively. At each step, a new tissue is included in the carrying combination of tissues. The tissue included at each step is the one for which RACEfrags include the maximum number of nucleotides in the genome, not previously included in the carrying combination of tissues. Correspondingly, tissues are ordered on the X-axis from left to right, so that the tissue at a given position is the one producing more novel RACEfrags with respect to the RACEfrags produced by tissues to its left on the axis. While this is a heuristic approach that does not guarantee optimality, we believe that for this particular problem, it will certainly produce a nearly optimal ranking.

A projected RACEfrag is a maximal set of RACEfrags that transitively overlap (see Supplementary Methods and Supplementary Figure 3). In the X-axis, index exons are ordered from 3' to 5' of the gene. For instance, the most 3’ exon generates 22% of all projected RACEfrags. The two most 3’ exons generate 40 % of them, and so on. Adding the 5’ most index exon generates 10% of all projected RACEfrags.

The blue bars reflect the frequency of extension lengths among different length classes. The solid yellow line shows the cumulative frequency of extensions of that length or greater. Most of the RACEfrags map within 1MB of the index primer. The pick at 10MB could reflect RACEfrags 3’ to the 5’ index primer (originated by circularization of the cDNA); note that the average distance between pooled primers is about 10MB.