Research article

Transcriptome sequencing of the Microarray Quality Control (MAQC) RNA reference samples using next generation sequencing

Shrinivasrao P Mane¹ , Clive Evans¹ , Kristal L Cooper¹ , Oswald R Crasta¹ , Otto Folkerts¹ , Stephen K Hutchison² , Timothy T Harkins³ , Danielle Thierry-Mieg⁴ , Jean Thierry-Mieg⁴ and Roderick V Jensen⁵

¹Virginia Bioinformatics Institute, Virginia Tech, Blacksburg, VA 24061, USA

²454 Life Sciences, Inc., 20 Commercial Street, Branford, CT 06405, USA

³Roche Applied Science, Indianapolis, IN 46250, USA

⁴National Center for Biotechnology Information, National Library of Medicine, National Institutes of Health, 8600 Rockville Pike, Bethesda, MD 20894, USA

⁵Department of Biological Sciences, Virginia Tech, Blacksburg, VA 24061, USA

author email corresponding author email

BMC Genomics 2009, 10:264doi:10.1186/1471-2164-10-264

Published:	12 June 2009

Abstract

Background

Transcriptome sequencing using next-generation sequencing platforms will soon be competing with DNA microarray technologies for global gene expression analysis. As a preliminary evaluation of these promising technologies, we performed deep sequencing of cDNA synthesized from the Microarray Quality Control (MAQC) reference RNA samples using Roche's 454 Genome Sequencer FLX.

Results

We generated more that 3.6 million sequence reads of average length 250 bp for the MAQC A and B samples and introduced a data analysis pipeline for translating cDNA read counts into gene expression levels. Using BLAST, 90% of the reads mapped to the human genome and 64% of the reads mapped to the RefSeq database of well annotated genes with e-values ≤ 10^-20. We measured gene expression levels in the A and B samples by counting the numbers of reads that mapped to individual RefSeq genes in multiple sequencing runs to evaluate the MAQC quality metrics for reproducibility, sensitivity, specificity, and accuracy and compared the results with DNA microarrays and Quantitative RT-PCR (QRTPCR) from the MAQC studies. In addition, 88% of the reads were successfully aligned directly to the human genome using the AceView alignment programs with an average 90% sequence similarity to identify 137,899 unique exon junctions, including 22,193 new exon junctions not yet contained in the RefSeq database.

Conclusion

Using the MAQC metrics for evaluating the performance of gene expression platforms, the ExpressSeq results for gene expression levels showed excellent reproducibility, sensitivity, and specificity that improved systematically with increasing shotgun sequencing depth, and quantitative accuracy that was comparable to DNA microarrays and QRTPCR. In addition, a careful mapping of the reads to the genome using the AceView alignment programs shed new light on the complexity of the human transcriptome including the discovery of thousands of new splice variants.