Data Release: Alzheimer Brain Isoform Sequencing (Iso-Seq) Dataset
Tuesday, September 20, 2016
Alzheimer’s disease (AD) is a devastating neurodegenerative disease that affects ~44 million people worldwide, making it the most common form of dementia. Pathologically it is defined by severe neuronal loss, aggregation of amyloid β (Aβ) in extracellular senile plaques in the brain, and formation of intraneuronal neurofibrillary tangles consisting of hyperphosphorylated tau protein. Studies looking into disease mechanism have shown that changes in gene expression due to alternative splicing likely contribute to the initiation and progression of AD. Hence, efforts have been made to better understand the gene expression changes in the AD brain by sequencing the transcriptome of affected brain regions.
Most transcriptome studies conducted to date have used short-read sequencing technologies, which provide the abundance of transcript reads needed for evaluating expression profiles. However, the ability to accurately identify alternative splicing and the associated expression patterns for different splice isoforms is limited by the short read-lengths. Given that the average size of a human gene transcript is several kilobases long, a 150bp to 300bp read will fail to span the entire transcript and therefore assembly will be required. In most cases this process can be very difficult, if not impossible, given the high similarity between expressed gene isoforms.
Recent studies using the PacBio isoform sequencing (Iso-Seq) method demonstrated the advantages of obtaining high-quality, full-length transcript sequences for improving genome annotation , , identifying fusion cancer genes , and discovering novel alternative splicing patterns . Here we apply the Iso-Seq method from an Alzheimer brain RNA sample. The purpose of releasing the dataset to the public is to provide researchers with a full-length transcriptome reference from which they can develop bioinformatics tools and validate their own findings.
In our final “confident” dataset, we obtained 21,742 high-quality, full-length isoforms covering 9,313 non-overlapping loci ranging from 352 bp – 9,457 bp, with an average length of 3,400 bp (Fig. 1). The total percentage of consensus bases that disagreed with the hg38 genome is 0.036% substitutions, 0.08% insertions, and 0.08% deletions, bringing the overall concordance with hg38 to 99.8%. More than half of the transcribed loci have one observed isoform, while most of the rest have about two to five isoforms (Fig. 2). When compared with the reference transcript annotation Gencode v25, more than a third of the isoforms match a reference transcript completely, while the majority of isoforms are possible novel splice forms of a known gene. In addition to the stringent “confident” dataset, we are also releasing a larger, less stringent “promiscuous” dataset. Details on the difference between the two versions can be found in the download section.
Library Preparation and Sequencing
An Alzheimer’s Disease Brain total RNA sample was purchased from BioChain. First strand cDNA library was generated using Clontech SMARTer cDNA synthesis kit followed by size selection using the SageELFTM device by Sage Science, with lanes combined to create five size libraries that roughly correspond to 1-2 kb, 2-3 kb, 3-5 kb, 5-7 kb, and > 7kb libraries. Sequencing was done using P6-C4 chemistry and 3-hr movies for the 1-2 kb fraction and 4-hr movies for the remaining fractions. Sequencing was completed in 2015. Download details on the sample preparation procedure.
The standard Iso-Seq pipeline (ToFU version 2.2.3 or equivalent to SMRT Analysis 3.1; for detailed methods see ) was used to process the data. Iso-Seq classify generated 1,107,889 FLNC reads and 1,929,319 nFL reads. The reads were then used to generate high-quality, full-length isoforms using ICE followed by Quiver polishing (HQ Quiver isoform consensus). By definition, an HQ Quiver consensus sequence must have at least two supporting full-length reads and predicted accuracy of >= 99%. The HQ Quiver consensus sequences were then aligned to human reference genome hg38 to create a final “confident” dataset of unique isoforms. To create the larger “promiscuous” dataset, additional consensus results that contained only one supporting full-length read were added. For details on the bioinformatics analysis, please see the README file on the Download Page.
Figure 1. Length distribution of final, unique, full-length isoforms.
Number of isoforms: 21,742
Min-max length: 352 bp – 9457 bp
Average length: 3400 bp
Figure 2. Number of isoforms per loci. 21,742 isoforms were grouped into 9,313 non-overlapping strand-specific loci. The average number of isoforms per loci was 2.3.
We welcome researchers to download and use the dataset for their research. For citation of the dataset, please use:
The Alzheimer brain Iso-Seq dataset was generated by Pacific Biosciences, Menlo Park, California, and additional information about the sequencing and analysis is provided at https://downloads.pacbcloud.com/public/dataset/Alzheimer_IsoSeq_2016/. The data used in the present study was retrieved from PacBio’s online database at https://downloads.pacbcloud.com/public/dataset/Alzheimer_IsoSeq_2016/ (date of retrieval).
 B. Wang, E. Tseng, M. Regulski, T. A. Clark, T. Hon, Y. Jiao, Z. Lu, A. Olson, J. C. Stein, and D. Ware, “Unveiling the complexity of the maize transcriptome by single-molecule long-read sequencing,” Nat Comms, vol. 7, p. 11708, Jun. 2016.
 S. E. Abdel-Ghany, M. Hamilton, J. L. Jacobi, P. Ngam, N. Devitt, F. Schilkey, A. Ben-Hur, and A. S. N. Reddy, “A survey of the sorghum transcriptome using single-molecule long reads,” Nat Comms, vol. 7, p. 11706, Jun. 2016.
 J. L. Weirather, P. T. Afshar, T. A. Clark, E. Tseng, L. S. Powers, J. G. Underwood, J. Zabner, J. Korlach, W. H. Wong, and K. F. Au, “Characterization of fusion genes and the significantly expressed fusion isoforms in breast cancer by hybrid sequencing,” Nucleic Acids Research, vol. 43, no. 18, pp. e116–e116, Oct. 2015.
 D. I. Pretto, J. S. Eid, C. M. Yrigollen, H.-T. Tang, E. W. Loomis, C. Raske, B. Durbin-Johnson, P. J. Hagerman, and F. Tassone, “Differential increases of specific FMR1mRNA isoforms in premutation carriers,” J Med Genet, vol. 52, no. 1, pp. 42–52, Dec. 2014.
 S. P. Gordon, E. Tseng, A. Salamov, J. Zhang, X. Meng, Z. Zhao, D. Kang, J. Underwood, I. V. Grigoriev, M. Figueroa, J. S. Schilling, F. Chen, and Z. Wang, “Widespread Polycistronic Transcripts in Fungi Revealed by Single-Molecule mRNA Sequencing,” PLoS ONE, vol. 10, no. 7, p. e0132628, Jul. 2015.