Targeted next-generation sequencing of a cancer transcriptome enhances detection of sequence variants and novel fusion transcripts
- 33k Downloads
Targeted RNA-Seq combines next-generation sequencing with capture of sequences from a relevant subset of a transcriptome. When testing by capturing sequences from a tumor cDNA library by hybridization to oligonucleotide probes specific for 467 cancer-related genes, this method showed high selectivity, improved mutation detection enabling discovery of novel chimeric transcripts, and provided RNA expression data. Thus, targeted RNA-Seq produces an enhanced view of the molecular state of a set of "high interest" genes.
KeywordscDNA Library Chronic Myeloid Leukemia Additional Data File Sequence Coverage Fusion Transcript
chronic myeloid leukemia
molecular inversion probe
sequence read format.
In recent years, a technologic revolution has shifted DNA sequencing from traditional Sanger methods to "next-generation" sequencing (see review ). Applying these new sequencing methods to cDNA libraries, termed RNA-Seq, generates a wealth of information beyond that obtained from sequencing genomic DNA (see review ). RNA-Seq provides insights at multiple levels into the transcription of the genome as it yields sequence, splicing, and expression-level information leading to the identification of novel transcripts [3, 4] and sequence alterations. For research into somatic mutations in cancer (for example, The Cancer Genome Atlas [5, 6, 7]), this method has the advantage of enriching for changes in coding sequences, which are more likely to affect function, compared with sequencing genomic DNA. Chromosomal rearrangements, including translocations, are an important class of mutations in cancer . Although chromosomal rearrangements can be detected by next-generation sequencing of genomic DNA [9, 10], RNA-Seq is a powerful tool to identify those rearrangements that lead to chimeric transcripts and are more likely to have functional consequences in cancer [3, 11].
Despite these advantages of RNA-Seq, the complexity of the transcriptome and the wide dynamic range of expression levels render whole-transcriptome sequencing an expensive proposition, particularly at the depth required to call mutations and identify structural rearrangements or aberrant splice forms in low-abundance mRNAs. Mortazavi and colleagues  reported that 40 million reads were required to provide onefold coverage of a transcriptome, whereas the calling genotypes with high confidence may require coverage levels of at least fivefold to 20-fold . This magnitude of coverage invariably results in vast oversampling of abundant transcripts, which adversely affects the efficiency and overall power of the approach.
Cost and efficiency considerations have prompted the emergence of methods that allow "targeted" next-generation sequencing. Two suitably high-throughput approaches to enrich specific sequences from genomic DNA have been developed: multiplexed molecular inversion probes (MIPs) [14, 15, 16] and capture by hybridization to oligonucleotide probes on microarrays [17, 18, 19] or in solution . MIPs are similar to PCR primers in that they enrich loci defined by two flanking specific sequences. Thus, they are not appropriate for the discovery of novel chromosomal rearrangements such as translocations. By contrast, capture by hybridization can enrich DNA fragments that extend beyond the probe sequence, including sequences that are not contiguous in the reference sequence. Solution hybrid selection is a capture method that uses a complex mixture of RNA baits derived from PCR-amplified oligodeoxynucleotides to select hybridizing sequences in a library of DNA fragments . To date, however, hybridization-based capture approaches have been applied primarily to genomic DNA, typically for the purpose of enriching exonic DNA of interest. Although targeted sequencing of genomic DNA facilitates mutation-discovery/profiling, it is unable to interrogate the myriad additional genomic alterations affecting DNA and mRNA that are critical to tumor biology and therapeutic development.
In this study, we explore the feasibility and power of "targeted RNA-Seq," the application of hybridization capture methods to transcriptome analysis. When applied to 467 cancer-related genes, this novel approach increased the coverage of low-abundance transcripts to levels that enabled reliable identification of sequence changes. In addition, this method provided information about relative expression levels, facilitated the discovery of novel splice variants, and enabled detection of novel fusion transcripts and isoforms thereof that would otherwise have escaped detection. As such, this method fills an important niche in cancer research, as well as other areas of genomics, by generating all the multifaceted genomic and gene-expression information in a single, straightforward experiment.
cDNA hybrid selection
To develop a targeted RNA-Seq method, we created a complex pool of biotinylated oligonucleotide probes (baits) for cancer-related transcripts and used them to capture cDNAs from a library prepared for Illumina sequencing. We targeted 467 genes in total (887 distinct transcripts; Table S1 in Additional data file 1), representing the majority of all protein tyrosine kinase genes, nuclear hormone-receptor genes, and genes catalogued in the Cancer Gene Census . Baits were designed in a tiling fashion with minimal overlap to span the entire protein-coding region of each transcript. To test the method, a cDNA library for Illumina sequencing was constructed from the K-562 chronic myeloid leukemia (CML) cell line. From an aliquot of this library, we selected cDNAs hybridizing with these cancer cDNA baits. We used PCR to regenerate a double-stranded DNA library that was sequenced in a single lane on the Illumina Genome Analyzer platform. To obtain a baseline for comparison, we also sequenced the original unenriched cDNA library.
Analysis of Illumina sequence in cDNA hybrid selection
Sequence filter criteria
Total purity-filtered reads
Aligned to all transcriptome
Unique in transcriptome
Mapping to 1 of 887 target transcripts
This increase in sequence coverage also increased the sensitivity for detecting sequence variants in these target genes. At positions with sufficient sequence coverage, we identified nonreference bases, including SNPs and candidate mutations. Hybrid selection enabled us to identify 257 known SNPs at high confidence (LOD > 5) in the coding sequences of target genes, compared with only 76 before hybrid selection. Similarly, we identified four novel variants before hybrid selection and an additional 12 for a total of 16 after hybrid selection (Table S2 in Additional data file 2). Thirteen (81%) of the 16 were successfully validated by traditional Sanger sequencing of PCR products amplified from K-562 genomic DNA. By comparison, three (75%) of the four novel variants detected before hybrid selection were validated.
The overall enrichment in sequence coverage for the target transcripts also enabled the identification of a greater number of alternatively spliced isoforms of these genes. Considering all possible logical intragenic combinations of exons annotated in RefSeq, 70,344 hypothetical splice junctions exist for the 467 target genes, and 6,593 of these have been annotated in RefSeq. The number of confirmed exon junctions involving target genes increased from 2,958 before hybrid selection to 4,720 after hybrid selection (Table S3 in Additional data file 3). Of these confirmed junctions, 294 are previously unannotated in RefSeq, involving 130 target genes. Genes exhibiting alternative splicing in K-562 were identified as described in Materials and methods. Hybrid selection revealed at least 177 target genes to be alternatively spliced, compared with 52 target genes before hybrid selection. Taken together, these results demonstrate the power of targeted RNA-Seq to illuminate both SNPs and splicing variants in an efficient manner.
Hybrid selection-enhanced detection of fusion transcripts
XKR3 b, c
SNHG3-RCC1 b, c
PRIM1 c, d
NACA b, d
CELSR3 c, d
SLC29A1 c, d
In addition to BCR-ABL1 and NUP214-XKR3, which were both detected before hybrid selection, we identified four gene fusions after hybrid selection that may have otherwise gone undetected and were not found previously . In each case, only one of the two genes was specifically targeted by baits (Table 2 and Table S4 in Additional data file 4). Three of the four gene fusions involve the production of "read-through" transcripts in which exons from separate, adjacent genes are joined together in a single mRNA molecule. Read-through transcripts have previously been discovered in cancer and have been shown to contribute to tumorigenicity . The fourth novel gene fusion involves the previously annotated SNHG3-RCC1 read-through transcript on chromosome 1 and PICALM on chromosome 11. As with NUP214-XKR3, multiple splice isoforms were detected for SNHG3-RCC1-PICALM, and four of five of them were confirmed by sequencing RT-PCR products (data not shown). Although these observations are consistent with a single genomic translocation followed by alternative splicing of the resulting RNA in both cases, it also is possible that further amplifications and rearrangements at this locus contributed to the multiple fusion transcripts observed.
In this study, we demonstrated that combining hybridization capture of a cDNA library with Illumina sequencing provides a robust and sensitive method to detect a wide range of DNA and RNA sequence alterations present in cancer cells. First, this method has a very high specificity, as 98% of the sequence mapping to RefSeq aligns to targeted transcripts after hybrid selection (Table 1). Second, this selectivity leads to improved detection of SNPs (Figure 1), splice isoforms, and fusion transcripts (Table 2) in the targeted transcripts. Importantly, this property reduces the amount of sequencing, and consequently costs, required to identify mutations and other cancer-associated variants. Third, differences in transcript abundance are generally preserved after hybrid selection (Figure 2), likely because the baits are in molar excess during the hybridization . Similarly, preservation of genomic copy-number alterations (that is, amplifications and deletions) has been observed after hybrid selection of genomic DNA in cell lines with well-characterized chromosomal aberrations (M.F.B. and L.A.G., unpublished data) and in filter-based hybridization experiments . Fourth, information that reflects function, such as expression levels, alternative splicing, and RNA editing, can be obtained by RNA-Seq directly from RNA input material rather than from genomic DNA. Beyond RNA expression levels (Figure 2), it is also possible to demonstrate that a particular fusion transcript is expressed, to identify fusion transcripts with partner genes that are not in targeted baits, and even to show the relative abundance of different spliced fusion transcripts (Table 2 and Figure 3). Fifth, fusion transcripts due to trans-splicing  would also be detected by this method, but not by genomic sequencing, and could be distinguished from translocations by validation experiments with genomic DNA. In summary, by sequencing cDNA rather than genomic DNA, we generated a richer view of the biologic state of this cell line. Recent studies that used MIPs to select for sequences subject to RNA editing  or to analyze allele-specific expression  provide additional examples of how targeted RNA-Seq can enhance our understanding of the molecular state of the transcriptome. Finally, this method is easily scalable to larger numbers of samples and genes.
Our targeted RNA-Seq method provides a direct and powerful approach to discover and characterize translocations and to study their prevalence in all types of cancer , including solid tumors, which is an area of active research [3, 4]. Although the role of translocations in leukemias and sarcomas is well established, we were able to identify novel fusion transcripts in the well-studied K-562 CML cell line (Table 2). By enriching for cDNA sequences from genes of known relevance to cancer, targeted RNA-Seq makes possible the identification of translocations for any number of targeted genes in a single experiment. In addition, oncogenes often have multiple translocation partners , and this method provides an effective tool to identify new partners for genes previously identified in translocations, because only one of the two translocated genes must be present in the hybrid-selection baits. This method is able to recover fusion transcripts in which incomplete matches to baits exist, probably because baits adjacent to the bait whose sequence contains the breakpoint enable this recovery and enrichment. This may be possible because the cDNA library inserts are 290 to 390 bp, which is larger than the 170-base baits.
It is interesting to compare the efficacy of targeted RNA-Seq to enhance detection of low-abundance transcripts with that of cDNA library normalization. Normalization is better suited for discovery of sequence changes in transcripts not known to be associated with a particular biologic question. By contrast, targeted RNA-Seq is ideal for increasing coverage for a subset of "high interest" transcripts. Further, unlike normalization, targeted RNA-Seq preserves expression-level information (Figure 2). In addition, targeted RNA-Seq can achieve higher increases in coverage for a subset of targeted transcripts, depending on the number of unique baits designed. If coverage of lower abundance transcripts is a priority in a given experiment, information about transcript abundance can be used during bait design to focus on those transcripts with targeted RNA-Seq. Conventional normalization methods [30, 31] are unlikely to achieve easily the approximately 30-fold enrichment for most low-abundance transcripts observed in our targeted RNA-Seq experiments (Figure 1; JZL., XA, unpublished results).
By combining hybridization capture of cDNAs and next-generation sequencing, targeted RNA-Seq provides an efficient and cost-effective means to analyze a specific subset of a transcriptome simultaneously for mutations, structural alterations, and expression levels. This method overcomes the limitations of conventional RNA-Seq that requires significantly greater sequencing depth to generate sufficient coverage of low-abundance transcripts. It also circumvents certain limitations of targeted genomic DNA sequencing, in which detection of chromosomal rearrangements may be challenging (and analysis of mRNA effects is impossible). In a single experiment, targeted RNA-Seq provides a wealth of qualitative as well as quantitative information that cannot be obtained by any single other method. Targeted RNA-Seq is therefore a powerful and convenient new approach that is well suited for a wide range of large-scale tumor-profiling studies in many clinical or research settings.
Materials and methods
Illumina library construction and sequencing
A K-562 cDNA library (insert size of 290 to 390 bp) for Illumina sequencing was constructed from a 500 ng aliquot of double-stranded cDNA prepared from 3 μg polyA+ RNA (Ambion, Austin, TX USA) primed with 0.3 μg random hexamers (Invitrogen, Carlsbad, CA USA), as described previously , except (a) no RNase inhibitor was used, (b) low-intensity shearing was performed for 5 seconds rather than 4 seconds, and (c) PCR primers were removed with 1.8× volumes of AMPure beads (Agencourt Bioscience Corporation, Beverly, MA USA). We used 14 PCR cycles to generate the library before hybrid selection and an additional 18 cycles afterward with the same PCR conditions. Single reads of 76 bases were generated on an Illumina Genome Analyzer II. The raw unaligned Illumina sequences in SRF (sequence-read format) are available at .
Bait design and synthesis
We designed 11,566 bait sequences (Table S5 in Additional data file 5) targeting the coding sequence of 887 transcripts of 467 genes described in the NCBI RefSeq database. The RefSeq file used contained 45,376 transcript sequences from all NM and XM human transcripts (downloaded from  on June 23, 2008). Each bait was composed of 170 bases matching the transcript sequence it was intended to enrich. Baits were tiled across the coding region of each transcript, from 5' to 3', such that all coding bases were covered by at least one bait and that overlap between baits was minimized and distributed evenly among all baits. The median and mean overlaps between adjacent baits were five and seven bases, respectively. Where genes existed with multiple splice variants, baits were designed for each splice variant independently, so that 9,913 unique baits were designed. The baits were synthesized by Agilent Technologies (Santa Clara, CA USA) on a custom 55,000 spot array. To fully use the array, the 11,566 baits were replicated so that at least two copies of each oligonucleotide were ordered, plus two copies of the reverse complement of each oligonucleotide. Oligonucleotides and their reverse complements give rise to the same PCR products. Thus, although sense and antisense oligonucleotides were chemically synthesized, only sense RNA baits were present in the hybridization.
Five hundred nanograms of the K-562 cDNA Illumina library was selected as described previously , except that the MEGAshortscript T7 Kit (Ambion) was used for the bait preparation.
Sequence alignment and coverage
Purity-filtered  76-mer reads were aligned to all curated protein-coding transcripts in RefSeq (downloaded from  on March 1, 2009) allowing up to four mismatches, and mapped back to their genomic coordinates in the reference human genome (hg18), preserving splice junctions. Alignments were performed by using the ImperfectLookupTable (ILT) tool of the ARACHNE genome assembly suite . Reads were considered informative if all placements to RefSeq transcripts originated from a unique genomic locus, and the next-best placement contained at least three additional mismatches. Sequence coverage was determined separately for 5' UTRs, coding sequences (CDS), and 3' UTRs.
Sequence variant identification
To eliminate false positives in calling mutations, reads aligning to RefSeq transcripts were also aligned directly to the genome, and uniqueness was required in both the transcriptome and the genome. Each position was assigned a LOD score indicating the likely accuracy of the call, according to the observed sequence coverage, allele distribution, and reference base . Of 1,085,748 bases in the coding sequence of targeted genes, 297,693 bases exhibited LOD greater than 5 before hybrid selection, and 724,211 bases exhibited LOD greater than 5 after hybrid selection. Bases that disagreed with the reference genome were classified as known SNPs if present in dbSNP  (build 129) or as novel variants. Novel variants were discarded if they occurred within five bases of another novel variant (to compensate for alignment artifacts produced by indels), if they were observed on Illumina reads in only one orientation, or if they fell within segmental duplications . The remaining novel variants were considered high confidence and submitted for validation (see later).
Splice isoform identification
To catalog the exon junctions detected by RNA-Seq, we created a database of all hypothetical intragenic exon junctions involving RefSeq genes. Each 76-mer read was aligned to this new reference-sequence database in the same manner as described earlier. Exon junctions were "confirmed" in K-562 if they harbored at least two distinct 76-mer reads mapping to the junction with, at most, four mismatches but with at least 10 mismatches with their best placement on the genome. Those genes with at least two confirmed exon junctions that overlapped each other (for example, one upstream exon joined to two downstream exons, two upstream exons joined to one downstream exon, or alternating exons) were considered to be alternatively spliced.
Fusion transcript identification
To identify candidate gene fusions from individual 76-mer reads, the first 30 and last 30 bases were separately aligned to all curated protein-coding transcripts in RefSeq (allowing up to two mismatches). Reads for which both ends mapped to separate genes were flagged for further analysis. Gene pairs implicated by at least two distinct reads (for which the orientation was consistent with a gene fusion) were nominated as candidate fusions. The entire set of 76-mer reads was then searched for instances joining any exon of the upstream gene to any exon of the downstream gene across the full 76 bases, requiring at least 12 bases overlap with each gene. To call confidently a gene pair as a fusion event, we required at least two distinct instances that could not be placed anywhere else in the transcriptome or the genome. The criteria used are conservative to avoid false-positive fusion transcript alignments; additional fusion transcripts may be present, but not detectable with these alignment parameters and coverage levels.
Validation of sequence alterations
Novel SNPs called by RNA-Seq were validated by traditional bidirectional Sanger sequencing of PCR products that had been amplified from 20 ng of K-562 genomic DNA by 35 PCR cycles with Herculase Hotstart DNA polymerase (Stratagene, La Jolla, CA, USA).
For NUP214-XKR3 and SNHG3-RCC1-PICALM fusion transcripts, confirmation was attempted by Sanger sequencing of RT-PCR products. First-strand cDNA was synthesized from K-562 mRNA with random hexamers (Invitrogen) or gene-specific primers (Eurofins MWG Operon, Huntsville, AL, USA) as described earlier. For random priming, 1 μg mRNA and 1.5 μg random hexamers were used, and for gene-specific priming, 500 ng mRNA and 2 pmol gene-specific primers were used. The cDNA was purified by using 1.8× volume of Agencourt AMPure PCR Purification kit. Fusion transcript-containing cDNAs were then amplified by 30 to 40 PCR cycles by using 1/50 of the purified first-strand cDNA, 25 pmol forward and reverse gene-specific primers, 100 μmol Betaine (Sigma-Aldrich, St. Louis, MO USA), and Phusion Master Mix with GC Buffer (New England BioLabs, Ipswich, MA, USA) in a 50 μl volume. PCR products were gel purified from a 10% TBE Criterion Gel (BioRad, Hercules, CA, USA). Gel slices were excised, crushed, and eluted with 250 μl 0.3 M NaCl for more than 4 hours followed by ethanol precipitation. The purified PCR products were sequenced as described earlier and compared with junctions identified by Illumina sequencing. All primer sequences are available on request.
Additional data files
The following additional data are available with the online version of this article: a table listing the bait gene names and transcript accession numbers (Additional data file 1), a table listing novel SNPs (Additional data file 2), a table listing splice junctions (Additional data file 3), a table listing Illumina reads from fusion transcripts (Additional data file 4), and a table listing the bait sequences (Additional data file 5).
We thank the staff of the Broad Institute Genome Sequencing Platform for generating sequencing data. We are grateful for help from Terrance Shea and Sarah Young with SNP calling. This work was supported by National Human Genome Research Institute grant HG03067-05, the Starr Cancer Consortium (L.A.G.), and National Institutes of Health grant DP2OD002750 (LAG).
- 4.Zhao Q, Caballero OL, Levy S, Stevenson BJ, Iseli C, de Souza SJ, Galante PA, Busam D, Leversha MA, Chadalavada K, Rogers YH, Venter JC, Simpson AJ, Strausberg RL: Transcriptome-guided characterization of genomic rearrangements in a breast cancer cell line. Proc Natl Acad Sci USA. 2009, 106: 1886-1891. 10.1073/pnas.0812945106.PubMedPubMedCentralCrossRefGoogle Scholar
- 5.McLendon R, Friedman A, Bigner D, Van Meir EG, Brat D, Mastrogianakis GM, Olson JJ, Mikkelsen T, Lehman N, Aldape K, Yung WK, Bogler O, Weinstein JN, VandenBerg S, Berger M, Prados M, Muzny D, Morgan M, Scherer S, Sabo A, Nazareth L, Lewis L, Hall O, Zhu Y, Ren Y, Alvi O, Yao J, Hawes A, Jhangiani S, Fowler G, et al: Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature. 2008, 455: 1061-1068. 10.1038/nature07385.CrossRefGoogle Scholar
- 7.The Cancer Genome Atlas. [http://cancergenome.nih.gov/]
- 9.Campbell PJ, Stephens PJ, Pleasance ED, O'Meara S, Li H, Santarius T, Stebbings LA, Leroy C, Edkins S, Hardy C, Teague JW, Menzies A, Goodhead I, Turner DJ, Clee CM, Quail MA, Cox A, Brown C, Durbin R, Hurles ME, Edwards PA, Bignell GR, Stratton MR, Futreal PA: Identification of somatically acquired rearrangements in cancer using genome-wide massively parallel paired-end sequencing. Nat Genet. 2008, 40: 722-729. 10.1038/ng.128.PubMedPubMedCentralCrossRefGoogle Scholar
- 10.Hampton OA, Den Hollander P, Miller CA, Delgado DA, Li J, Coarfa C, Harris RA, Richards S, Scherer SE, Muzny DM, Gibbs RA, Lee AV, Milosavljevic A: A sequence-level map of chromosomal breakpoints in the MCF-7 breast cancer cell line yields insights into the evolution of a cancer genome. Genome Res. 2009, 19: 167-177. 10.1101/gr.080259.108.PubMedPubMedCentralCrossRefGoogle Scholar
- 11.Maher CA, Palanisamy N, Brenner JC, Cao X, Kalyana-Sundaram S, Luo S, Khrebtukova I, Barrette TR, Grasso C, Yu J, Lonigro RJ, Schroth G, Kumar-Sinha C, Chinnaiyan AM: Chimeric transcript discovery by paired-end transcriptome sequencing. Proc Natl Acad Sci USA. 2009, 106: 12353-12358. 10.1073/pnas.0904720106.PubMedPubMedCentralCrossRefGoogle Scholar
- 13.Bentley DR, Balasubramanian S, Swerdlow HP, Smith GP, Milton J, Brown CG, Hall KP, Evers DJ, Barnes CL, Bignell HR, Boutell JM, Bryant J, Carter RJ, Keira Cheetham R, Cox AJ, Ellis DJ, Flatbush MR, Gormley NA, Humphray SJ, Irving LJ, Karbelashvili MS, Kirk SM, Li H, Liu X, Maisinger KS, Murray LJ, Obradovic B, Ost T, Parkinson ML, Pratt MR, et al: Accurate whole human genome sequencing using reversible terminator chemistry. Nature. 2008, 456: 53-59. 10.1038/nature07517.PubMedPubMedCentralCrossRefGoogle Scholar
- 20.Gnirke A, Melnikov A, Maguire J, Rogov P, LeProust EM, Brockman W, Fennell T, Giannoukos G, Fisher S, Russ C, Gabriel S, Jaffe DB, Lander ES, Nusbaum C: Solution hybrid selection with ultra-long oligonucleotides for massively parallel targeted sequencing. Nat Biotechnol. 2009, 27: 182-189. 10.1038/nbt.1523.PubMedPubMedCentralCrossRefGoogle Scholar
- 22.Yassour M, Kaplan T, Fraser HB, Levin JZ, Pfiffner J, Adiconis X, Schroth G, Luo S, Khrebtukova I, Gnirke A, Nusbaum C, Thompson DA, Friedman N, Regev A: Ab initio construction of a eukaryotic transcriptome by massively parallel mRNA sequencing. Proc Natl Acad Sci USA. 2009, 106: 3264-3269. 10.1073/pnas.0812841106.PubMedPubMedCentralCrossRefGoogle Scholar
- 24.Graux C, Cools J, Melotte C, Quentmeier H, Ferrando A, Levine R, Vermeesch JR, Stul M, Dutta B, Boeckx N, Bosly A, Heimann P, Uyttebroeck A, Mentens N, Somers R, MacLeod RA, Drexler HG, Look AT, Gilliland DG, Michaux L, Vandenberghe P, Wlodarska I, Marynen P, Hagemeijer A: Fusion of NUP214 to ABL1 on amplified episomes in T-cell acute lymphoblastic leukemia. Nat Genet. 2004, 36: 1084-1089. 10.1038/ng1425.PubMedCrossRefGoogle Scholar
- 31.Zhulidov PA, Bogdanova EA, Shcheglov AS, Vagner LL, Khaspekov GL, Kozhemyako VB, Matz MV, Meleshkevitch E, Moroz LL, Lukyanov SA, Shagin DA: Simple cDNA normalization using kamchatka crab duplex-specific nuclease. Nucleic Acids Res. 2004, 32: e37-10.1093/nar/gnh031.PubMedPubMedCentralCrossRefGoogle Scholar
- 32.Broad Institute Targeted RNA-Seq Sequence Data. [http://www.broad.mit.edu/annotation/cDNA/cDNA.html]
- 33.NCBI RefSeq Human mRNA FTP site. [ftp://ftp.ncbi.nih.gov/refseq/H_sapiens/mRNA_Prot/]
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium provided the original work is properly cited.