20140711 4 e_tseng_ercc2.0_workshop
2 likes905 views
This document discusses technical variability in PacBio full-length cDNA sequencing (Iso-Seq). It summarizes the Iso-Seq experimental and informatics pipelines, and analyzes read count variation between technical replicates and tissues. While technical variation is minimal, amplification biases from different enzymes and detection limits remain areas for improvement. Combining Iso-Seq with short-read data may help address these challenges.
20140711 4 e_tseng_ercc2.0_workshop
- 1. FIND MEANING IN COMPLEXITY For Research Use Only. Not for use in diagnostic procedures. © Copyright 2014 by Pacific Biosciences of California, Inc. All rights reserved. Elizabeth Tseng / 2014.07.11 Staff Scientist Technical Variability in PacBio® Full-length cDNA (Iso-SeqTM) Sequencing
- 2. SampleNet: Iso-Seq Method with Clonetech® cDNA Synthesis Kit PacBio’s Iso-Seq™ Method for High-quality, Full-length Transcripts PolyA mRNA AAAAA AAAAA AAAAA AAAAA cDNA synthesis with adapters AAAAA TTTTT AAAAA TTTTT AAAAA TTTTT AAAAA TTTTT AAAAA TTTTT AAAAA TTTTT AAAAA TTTTT AAAAA TTTTT Size partitioning & PCR amplification SMRTbell™ ligation PacBio® RS II Sequencing Experimental Pipeline Informatics Pipeline Remove adapters Remove artifacts Clean sequence reads Reads clustering Isoform clusters Consensus calling Nonredundant transcript isoforms Quality filtering Final isoforms PacBio raw sequence reads 5’ primer 3’ primer Map to reference genome Experimental pipeline Informatics pipeline PacBio raw sequence reads a b AAAA AAAA AAAAA AAAAA AAAAA AAAAA AAAAA Size partitioning & PCR amplification cDNA synthesis with adapters SMRTbell ligation RS sequencing Remove adapters Remove artifacts Reads clustering Quality filtering Clean sequence reads Nonredundant transcript isoforms Final isoforms TTTT TTTT Consensus calling Isoform clusters Map to reference genome Evidence-based gene models polyA mRNA AAAA AAAA TTTT TTTT AAAA TTTT AAAA TTTT AAAA TTTT AAAA TTTT Evidenced-based gene models (AAA)n (TTT)n 1 2 3 4 5 6 7 8 9 10 (TTT)n (AAA)n Coding sequence polyA tail SMRT® adapter DevNet: Iso-Seq wiki page (AAA)nReads of Insert (AAA)n
- 3. Iso-Seq Full-length cDNA Library Protocol 3 polyA+ RNA Total RNA Optional Poly-A Selection Reverse Transcription (SMARTScribe RT) Full-‐length 1st Strand cDNA PCR Optimization Large-scale Amplification Amplified cDNA 1-‐2 kb 2-‐3 kb 3-‐6 kb Size Selection 1-‐2 kb 2-‐3 kb 3-‐6 kb Re-Amplification 1-‐2 kb 2-‐3 kb 3-‐6 kb SMRTbell™ Template Preparation 1-‐2 kb 2-‐3 kb 3-‐6 kb SMRT® Sequencing 3-‐6 kb Optional Size Selection
- 4. Iso-Seq Informatics Pipeline Per-molecule reads Clusters of transcript alignments using FL + nFL reads Transcript 1 Transcript 2 Transcript 3 Final transcript consensus Transcript 1 Transcript 2 Transcript 3 Full-length (FL) reads Non-FL reads Transcript 1 Transcript 2 Transcript 3 Isoform-level clusters
- 5. Key Features of Current Iso-Seq Bioinformatics • Non-redundant, full-length, transcript consensus sequences – No assembly – De novo – Achieves high-quality consensus (≥ 99%) – Universal PacBio features: robust to GC%, repeat structure, etc • Applications – Alternative splicing – Fusion transcripts – Alternative polyadenlyation – (possible w/ proper protocol) Alternative start sites
- 6. Disclaimer • Everything shown from now on are transcripts/isoforms, not genes • Data shown is preliminary, very unbaked • Concept Analysis
- 7. Count Information Associated with Each Unique Transcript Clusters of transcript alignments using FL + nFL reads Transcript 1 Transcript 2 Transcript 3 Final transcript consensus Transcript 1 Transcript 2 Transcript 3 Count matrix Transcript Count Norm_Count 1 2 3 … 8 5 7 … 0.08 0.05 0.07 …
- 8. Count Information from non-FL reads For non-FL reads: • If uniquely associated with a transcript, assume it is the transcript • If ambiguously associated, most likely because it’s a partial match • For now, weight of ambiguous nFL is just read _count = # of FL + # of unique nFL + weighted # of ambiguous nFL 1 Number of associated transcripts In current dataset, about 40-60% nFL reads partially match multiple isoforms (FL reads are always fully and uniquely associated)
- 9. Read Count Variation in Technical Replicates Rat Heart • Technical replicates (same starting RNA & protocol) • 3 size libraries (1 – 2 kb, 2 – 3 kb, 3 – 6 kb) • Runs from diff sizes pooled for bioinformatics pipeline Boxplot of log2 read counts Scatterplot of log2 read count for each transcript Rat Heart, technical replicates
- 10. Read Count Variation in Technical Replicates 10 Rat Lung, technical replicates All technical replicates were seq with total ~8 SMRT® Cells (low depth) Most NA transcripts are low counts
- 11. Choice of Chemistry Does Not Bias Sequencing 11 Rat Brain Same 3-size library (not technical replicate) • Sequenced with P4-C2 chemistry • Sequenced with P5-C3 chemistry However for longer (> 3 kb) transcripts, P5-C3 chemistry will increase chance of seeing FL reads
- 12. Choice of PCR Enzyme May Bias Amplification 12 Human Brain, 2 – 3 kb library Human Brain, 3 – 6 kb library
- 13. Current Iso-Seq Protocol Amplifies Sample Twice 13 polyA+ RNA Total RNA Optional Poly-A Selection Reverse Transcription (SMARTScribe RT) Full-‐length 1st Strand cDNA PCR Optimization Large-scale Amplification Amplified cDNA 1-‐2 kb 2-‐3 kb 3-‐6 kb Size Selection 1-‐2 kb 2-‐3 kb 3-‐6 kb Re-Amplification 1-‐2 kb 2-‐3 kb 3-‐6 kb SMRTbell™ Template Preparation 1-‐2 kb 2-‐3 kb 3-‐6 kb SMRT® Sequencing 3-‐6 kb Optional Size Selection
- 14. 2nd Amplification Does Not Introduce Strong Bias 14 FL Read Length Distribution Std. vs. skipping 2nd amp Std. vs. skipping 1st amp Skipping 1st amplification results in size selection of first-strand cDNA that may be hard to optimize
- 15. Expected Transcript Variability in Different Rat Tissues 15 Rat Heart vs Rat Lung Rat Heart vs Rat Brain Heart Lung Heart Brain
- 16. Conclusion • Technical variation not a big issue – If done with same library protocol – Different (PCR) enzymes bias amplification – Amplification can be tolerated if kept at reasonable # of cycles • Potential for DE – Still many unknown factors – Everything shown in previous slides merely “proof of concept” – With control comes better modeling 16
- 17. Looking Ahead 17 • Detection limit • Amplification bias – Adding control at known % – Factors: GC? Length? Enzyme? • Account for library pooling • Ambiguous mapping • Modeling bias • DE isoform detection • Combining short-read data Wet Lab Bioinformatics
- 18. For Research Use Only. Not for use in diagnostic procedures. Pacific Biosciences, the Pacific Biosciences logo, PacBio, SMRT, SMRTbell and Iso-Seq are trademarks of Pacific Biosciences in the United States and/or other countries. All other trademarks are the sole property of their respective owners.