Digital DNA/RNA sequencing enables highly accurate and sensitive biomarker detection and quantification

Transcription

1 Digital DNA/RNA sequencing enables highly accurate and sensitive biomarker detection and quantification Erwin Chen ( 陳立德 ) Technical Product Specialist QIAGEN Taiwan

2 Precision medicine: Right drug, right patient, right time and dose One size fits all does not work 2

3 DNA variants for precision medicine A C T G T A G C Mutations AGCTCGTTGCTCAGCTC Reference genome AGCTCGTTGCTCAGCGTTC Insertion AGCTC---GCTCAGCTC Deletion Indels Copy number variations RS 3

4 Actionable DNA variants for precision medicine Only a handful of mutations are actionable AGCTCGTTGCTCAGCTC Reference genome AGCTCGTTGCTCAGCGTTC Insertion AGCTC---GCTCAGCTC Deletion Mutations Indels Copy number variations Actionable DNA Variant BRAF V600E EGFR E Kinase domain mutation HER2 Disease Melanoma Lung adenocarcinomas IDC-Breast cancer Therapy Vemurafenib (PLX4032) Erlotinib / Gefitinib Trastuzumab RS 4

5 Actionable DNA variants for precision medicine How many? RS 5

6 Precision medicine for lung cancer Current lung cancer biomarker landscape EGFR (L858R) + KRAS (G12C) Unknown 37% KRAS 25% NRAS 0.2% MAP2K1 0.4% ERBB2 1% MET 2% PIKC3A 3% BRAF 3% EML4-ALK 6% EGFR 23% Response rates of >70% in patients with non-small cell lung cancer treated with either erlotinib or gefitinib Poor response rate in patients with non-small cell lung cancer treated with either erlotinib or gefitinib How many mutations to test for? How to test for these mutations Sequential testing Parallel testing Adapted from: Govindan, R. et al. (2012). Genomic landscape of non-small cell lung cancer in smokers and never-smokers. Cell 150, RS 6

7 Actionable DNA variants for precision medicine Targeted DNA sequencing delivers accurate information required for precision medicine Attribute / Parameter Whole Genome Sequencing Whole Exome Sequencing Targeted DNA Sequencing Benefits of Targeted DNA Sequencing: Information level 3 x 10 9 bps 5 x 10 7 bps 6 x 10 4 bps More relevant data Cost per sample $5000 $2000 $200 More cost effective Coverage achieved 30x 100x 1000x Detect low-frequency mutations DNA input 1 µg ng 10 ng Lower DNA requirements No. of samples multiplexed Higher multiplexing capabilities Clinical utility requires targeted analysis RS 7

8 Why choose targeted DNA sequencing? Targeted DNA sequencing limits the genes or targets to be sequenced Features Well-defined content Small target size More reads per sample Benefits Examine variants that matter Multiplex many samples to save money Detect low frequency variants RS 8

9 Targeted DNA Sequencing (TDS) Shrink the genome Sample KRAS EGFR IDH1 Insight Sample isolation Library construction & targeted enrichment NGS run Data analysis Interpretation Variants Report: gdna KRAS G12D EGFR T790M IDH1 R132H The principle of targeted enrichment is to simultaneously sequence millions of small DNA fragments that represent the region of interest RS 9

10 PCR is widely used in library construction for NGS Eric Samorodnitsky, etc. Hum Mutat 36: ,

11 Basic concepts before sequencing Biology is still biology Assay technology improvement will not change biological facts. Biological replicates still needed no matter how good an assay is. Real patient samples are still very important no matter how computation can help. The more the assay points the more biological variants NGS usually means more samples are needed Samples only provide the information they have Low quality sample (such as bad FFPE) will provide less information and/or more noise. Less than enough samples will not provide meaningful information Technology always has its limitation Sequencer Image analysis or ph analysis Materials prep Bioinformatics Highly rely on data processing accuracy and speed 11

12 QIAseq solutions to detect all Biomarkers using NGS Gene Expression QIAseq targeted RNA panels QIAseq targeted DNA panels Indels mirna expression QIAseq mirna sequencing system Biomarkers QIAseq targeted DNA panels Mutations Copy number variants Fusions QIAseq targeted RNAscan panels QIAseq targeted DNA panels 12

13 Why choose PCR-based targeted enrichment? It delivers unmatched specificity and uniformity (compared to capture-based methods) Features Offers specificity that beats capture-based approaches Benefits Lets you use sequencing capacity on regions targeted by the panel, with minimal offtarget sequencing Lets you achieve more uniform enrichment for more sequencing efficiency RS 13

14 The necessary evil: PCR amplification PCR amplification is required for target enrichment, but DNA dsdna PCR amplification & sequencing PCR and sequencing errors 14

15 Challenges of current DNA targeted sequencing approaches Inability to detect low-frequency mutations PCR and sequencing errors Limits sensitivity and accuracy of calling low-frequency variants o Doesn t let you confidently call variants down to 1% variant allele frequency (VAF) Inefficient enrichment and sequencing of GC-rich regions Suboptimal, GC-rich region-incompatible PCR chemistry Limits comprehensiveness of panel coverage o Doesn t let you efficiently sequence clinically-relevant genes such as CEBPA or CCND1 or clinically-relevant regions such as TERT promoter Suboptimal uniformity of enrichment and sequencing Conventional PCR protocols and two-primer amplicon design Increases variability in coverage across targeted genomic regions o o Causes you to over-sequence to accommodate the under-sequenced Doesn t let you call variants in low-depth regions Mainly due to inferior PCR amplification approaches 15

16 PCR amplification: The necessary evil PCR amplification is required but generates artifacts such as errors, duplicates, and bias False positives due to errors Inaccurate quantification due to duplicates Non-uniform coverage due to bias PCR amplification PCR amplification artifacts limit the NGS accuracy RS 16

17 Challenges of conventional targeted DNA sequencing PCR duplicates limit accurate quantification Conventional targeted DNA sequencing EGFR exon 21 5 reads OR library fragments that look exactly the same. Cannot tell whether they represent: 1. 5 unique DNA molecules, or 2. Quintuplets of the same DNA molecule (PCR duplicates) Quantification based on non-unique reads does not reflect quantities of original DNA molecules 17

18 Challenges of conventional targeted DNA sequencing PCR and sequencing errors (artifacts) limit variant calling accuracy Conventional targeted DNA sequencing EGFR exon 21 A mutation is seen in 1 out of 5 reads that map to EGFR exon 21. Cannot accurately tell whether the mutation is: 1. A PCR or sequencing error (artifact) / false positives, or 2. A true low-frequency mutations * Variant calling based on non-unique reads does not reflect the mutational status of original DNA molecules 18

19 Digitalize sequencing result with more accurate counting of original signal 19

20 What is a UMI (molecular barcode)? Tag (barcode) to identify unique DNA molecules TATCGTACAGAT (12 nucleotides long) Incorporate this random barcode (signature) into the original DNA molecules before amplification to preserve their uniqueness 20

21 Overcoming current challenges For optimal variant detection Current approach Challenges How QIAseq targeted DNA panels overcome challenges of current approaches Conventional targeted DNA sequencing for variant detection PCR and sequencing errors UMIs that enable digital sequencing to correct for PCR and sequencing errors Inefficient sequencing of GCrich regions Suboptimal uniformity of enrichment and sequencing Proprietary chemistry to efficiently sequence GC-rich regions SPE-based primer design to increase uniformity With the QIAseq targeted DNA panels, variant detection is done by analyzing unique DNA molecules instead of total reads 21

22 How are UMIs incorporated? Ligate molecularly-barcoded adapters to unique DNA molecules before amplification DNA dsdna Molecularly-barcoded adapter PCR amplification & sequencing TATCGTACAGAT Incorporate this random barcode (signature) into the original DNA molecules before amplification to preserve their uniqueness Correct for PCR duplicates & errors 22

23 Achieve accurate quantification with molecular barcodes Count and analyze single original molecules (not total reads) = digital sequencing Conventional targeted DNA sequencing EGFR exon 21 5 reads OR library fragments that look exactly the same. Cannot tell whether they represent: 1. 5 unique DNA molecules, or 2. Quintuplets of the same DNA molecule (PCR duplicates) UMIs before any amplification Digital sequencing with UMIs UMI 5 unique DNA molecules since 5 molecular barcodes are detected Quintuplets of the same DNA molecule (PCR duplicates) since 1 molecular barcode is detected 23

24 Achieve accurate variant calling with molecular barcodes Count and analyze single original molecules (not total reads) = digital sequencing Conventional targeted DNA sequencing EGFR exon 21 * A mutation is seen in 1 out of 5 reads that map to EGFR exon 21. Cannot accurately tell whether the mutation is: 1. A PCR or sequencing error (artifact) / false positives, or 2. A true low-frequency mutations UMIs before any amplification Digital sequencing with UMIs UMI * * False variant is present in some fragments carrying the same UMI True variant is present in all fragments carrying the same UMI 24

25 Advantage of Single Primer Extension (SPE) SPE Nested PCR Booster GSP No need for long fragments to give space for two primers apart from the fusion junction, so shorter fragments are better tolerated. This will help to deal with the challenge for FFPE samples in which many RNA fragments may not have enough length to cover both GS1 and GS2 sequence in competitor A panels. Competitor A panels might require longer fragments, which will reduce the efficiency of PCR and less spanning reads will cross fusion junctions One gene/target-specific primer Two gene/target-specific primers Flexibility in primer design Focus on original signal but not the amplification Generate relatively small library fragments to maintain compatibility with fragmented RNA (FFPE samples) 25

26 Enhanced design and chemistry for high specificity Standard: 96 primers designed with standard primer design algorithm, assay was run with standard PCR chemistry. QIAseq RNAscan/DNAseq: 96 SPE primers designed with optimized design algorithm, assay was run with optimized chemistry. Improved specificity with RNAscan while keeping high sensitivity and uniformity 26

27 QIAseq targeted DNA panels: Sample to Insight Panels, molecularly-barcoded adapters and data analysis algorithms Sample Sample isolation Library construction & targeted enrichment NGS run Data analysis Interpretation Insight Panels and molecularlybarcoded adapters Barcode-aware variant calling pipeline 27

28 1.5 Days QIAseq Targeted DNA Panel: Workflow DNA Enzyme-based random DNA fragmentation 5 5 End repair and A tailing 5 MB A 5 A 5 Adapter ligation / library construction (incorporation of adapters, molecular barcodes and sample indexes) MB Adapter IL-F MB Add GSPs and UP* RSP Target enrichment by SPE 5 Add indexes and UP* 5 IL-U Sequencing-ready library SIP Universal PCR amplification Sample indexing and amplification Lib quant* MB: Molecular barcode RSP: Region-specific primer FP: Forward primer UP: Universal primer SIP: Sample index primer *Preceded by bead cleanup 28

29 Data analysis with barcode-aware variant caller overview For QIAseq targeted panels Data upload to cloud ( Barcode-aware variant caller has been developed Caller is available on the cloud In conjunction with molecular barcodes incorporated in the workflow, the caller can confidently call low-frequency variants (down to 1% variant allele frequency, VAF ) Variant caller will do the following: Mapping Alignment Molecular barcode counting Variant / calling Variant / annotation variants based on public databases 29

30 The differences between Variant Analysis and QCI Interpret Variant A Variant B Variant d Filtered Variants Variant Analysis filters and annotates variants (IVA) Variant Analysis applies our knowledge base and user-configurable filter cascade to generate a list of meaningful variants from a broader VCF or VCFs Variant Analysis provides the user with analysis tools for one sample, pairs, trios or cohort studies. 30

31 QIAseq Targeted RNA Panels and mirna library kit for gene expression profiling 31

32 QIAseq solutions to detect all Biomarkers using NGS Gene Expression QIAseq targeted RNA panels QIAseq targeted DNA panels Indels mirna expression QIAseq mirna sequencing system Biomarkers QIAseq targeted DNA panels Mutations Copy number variants Fusions QIAseq targeted RNAscan panels QIAseq targeted DNA panels 32

33 Gene expression profiling II qrtpcr Hybridization Array (aka the reverse dot blot) Digital PCR Transcriptome NGS relative quantitation with high precision large dynamic range moderate assay throughput 384 in parallel low throughput singleplex assays high sample requirements relative quantification with medium precision compressed dynamic range extremely high assay throughput low sample throughput absolute quantification, high accuracy narrow dynamic range moderate assay throughput low sample throughput price per data point can be high relative quantitation with high precision high dynamic range extremely high assay throughput extremely low sample throughput price per sample very high 33 33

34 WTS - whole transcriptome sequencing Benefits Quantifies and characterizes all RNA Identification of alternative splicing events Detects expressed SNPS or mutations allele specific expression patterns Drawbacks Large computational requirements Massive amount of data generated Filtering, alignment, assembly, curation Aggressive normalization for quantification Cost; Only runs on HT instruments Limits accessibility to core labs Requires large read budget & money Limits sample numbers in studies 34 34

35 WTS - whole transcriptome sequencing Benefits Quantifies and characterizes all RNA Identification of alternative splicing events Detects expressed SNPS or mutations allele specific expression patterns But what if we are only interested in gene expression? WTS Benefits Quantifies and characterizes all RNA Identification of alternative splicing events Detects expressed SNPS or mutations allele specific expression patterns 35 35

36 QIAseq targeted RNA panels Digital sequencing by Molecular barcodes PCR duplicates and amplification bias are major issues in current RNAseq workflow, as they result in biased and inaccurate gene expression profiles PCR duplicates Amplification bias Ratio based on reads Gene A Sample 1 Ratio of original state of genes 4 1 Ratio of genes based on reads [reads (ratio)] 12 (2) Gene A Sample 2 6 (1) mrna cdna 36

37 QIAseq targeted RNA panels How does it work? How does it allow digital sequencing? Molecular barcodes allow the counting of original gene levels instead of PCR duplicates, thereby enabling digital sequencing and resulting in unbiased and accurate gene expression profiles Tag each gene with unique molecular barcodes Count unique barcodes, not reads Gene A Sample 1 Ratio of original state of genes 4 1 Ratio of genes based on barcodes 4 Gene A Sample 2 1 mrna cdna 37

38 QIAseq targeted RNA panels Workflow RNA sample cdna synthesis 6 hours Primer extension & Molecular barcoding (GSP1) 1 st -stage PCR (GSP2) FS2 GSP2 Second Primer for gene 1 (GSP2) GSP1 Barcoded Primer for gene 1 (GSP1) GSP1 MT MT RS2 RS2 2 nd -stage PCR Library construction & Sample indexing Library quantification 38

39 Take the guesswork out of your analysis Built-in controls, integrated with data analysis HKG 6 gdna assays control for gdna contamination in RNA samples mean barcodes per target calculated and mrnas expressed near this number are flagged during analysis as close to noise level 10 HKG assays available to normalize data to make sample-to-sample and runto-run comparisons possible Flexible use one, two, all, none or any other genes as normalizers in custom panels HKG suitability calculations built into secondary data analysis. 39

40 Some QIAseq RNA considerations Criteria Biological replicates Technical replicates Coverage across the transcript Essential for robustness of experimental design Generally not required Not important; we are counting genes by common regions. Stranded library prep Paired-end reads Not required, amplicons do not overlap lncrna Not required. 150-base single ended reads more than enough (platform independent) Role of sequencing depth Overall sequencing depth Capture enough unique barcodes of each transcript such that statistical inferences can be made (=>10 per gene) High enough to infer accurate statistics as determined by molecular barcoding - >1 reads per unique barcode 40

41 Unparalleled efficiency and flexibility vs PCR An example: 96 samples, 421 genes Parameter QIAseq targeted RNA panels RT-PCR Material required One pool of primers well plates Run time 14 hours for NextSeq run 310 hours (2 hours per plate) Hands-on time 3 hours (for 96 samples) 105 hours (one hour per plate) Cost per sample $65 (exclusive of sequencing run) $239 Sample 10 ng each sample 4 ug each sample 42

42 QIAseq solutions to detect all Biomarkers using NGS Gene Expression QIAseq targeted RNA panels QIAseq targeted DNA panels Indels mirna expression QIAseq mirna sequencing system Biomarkers QIAseq targeted DNA panels Mutations Copy number variants Fusions QIAseq targeted RNAscan panels QIAseq targeted DNA panels 43

43 mirnas: Master regulators of gene expression mirnas are ~21 nucleotide, small non-coding RNAs that are expressed in virtually all tissues. Changes in mirna expression can be correlated with gene expression changes in: Development Differentiation Signal transduction Infection Aging Disease Human Cancer: mirnas are associated with cell proliferation, resistance to apoptosis, invasiveness, and differentiation in cancer cells 44

44 What challenges were there with mirna sequencing? Sample input requirements: Kits require up to 1 µg of input and struggle on the lower input side May not be compatible or optimized for biofluids Data quality may be low/non-quantitative: On target mirna NGS reads account for only 20-30% of total reads Adapter dimers eat at your read budget Gel band excision is not an exact science (not 100% mirna) Biofluid samples may be largely contaminated with other RNA types similar in size Workflow can be tedious and unreliable Gel-based workflows take multiple days, unlikely to get on the sequencer in days Large sample numbers are tedious with gels Optimization from sample isolation not available mirna Sequencing Novel mirna discovery and sample screening Real-time PCR NGS validation and Expression analysis 45

45 QIAseq mirna overview What is the kit? mirna-focused next-generation sequencing library prep kit and integrated bioinformatics/data analysis solution Compatible with Illumina sequencers What can be done with the sequencing data? Differential expression calculations of mirna from highly multiplexed samples Novel mirna discovery What are distinguishing features of the prep kit? Gel-free, rapid workflow Broad RNA input: 1 ng to 500 ng No adapters dimers at any RNA input amount Library prep from serum, plasma, biofluids, cells and tissues (any species) Integrated Unique Molecular Index (UMI) technology Highly optimized chemistry All-in-one-box solution QIAseq mirna Library Kit: Unparalleled mirna-focused sequencing for robust mirna quantification and discovery 47

46 QIAseq mirna: Library construction 5 PO 4 3 mirna 5 PO 4 Pre-adenylated adapter 3 Step 1: 3 ligation 5 3 Step 2: 5 ligation 5 3 Elimination of adapter dimer from 3 sequencing library RT primer with UMI 5 Step 3: Reverse-transcription with Unique Molecular Index (UMI) assignment Step 4: cdna cleanup Universal For Rev with Index Step 5: Library amplification and Sample Index assignment Step 6: Library cleanup Library Pre-Seq QC, Determining Library Conc, Prep for Seq, Data Analysis 48

47 QIAseq mrna/mirna: Primary data analysis Free, easy to use primary data analysis: Adapter Trimming Read Mapping UMI Counting Species Human, mouse, rat or other (all of mirbase) Adaptor Trimming Remove the adapter sequences from the reads Read Mapping Identify the possible position of the read within the reference Align the read sequence to reference sequences UMI Counting Merges unique UMI with mapped reads 49

48 Small Molecule Application Data Primary Data Analysis for QIAseq Targeted RNA Sequencing QIAseq RNA Quantification - Read Details: Unique Captures per Target Gene molecular barcode count Differential gene expression inter- and intra-samples 50

49 QIAseq Targeted RNA/miRNA Secondary Data Analysis Export Data Fold Changes and p-value Tab 51

50 Secondary Data Analysis for QIAseq Targeted RNA Sequencing Scatter plot and clustergram (treated sample compared to Control) Scatter Plot Volcano Plot Clustergram 52

51 Ingenuity IPA analysis HDAC Mechanistic Network in HEK293T Cells Treated with Trichostatin A HDAC is predicted to be inhibited by Trichostatin A and drives a mechanistic network with 18 other regulators. Cell cycle NHR, proliferation Transcriptional activator 53

52 Sample to Insight: Integrated universal targeted NGS workflow To overcome NGS challenges Sample Insight Sample QC Library QC Variant confirmation Sample isolation Library construction & targeted enrichment NGS run Data analysis Interpretation Isolation of high-quality DNA samples Quantification of amplifiable (not total) amounts of DNA Turnaround time, and limited amounts of DNA Uniformity of enrichment Coverage of GCrich regions Platformdependent challenges Data processing & variant calling Clinical & biological interpretation of data RS 54

53 Questions? 55