1. NGS genetic analysis for National Lung Matrix Trial
Stratified Medicine technology hubs and Illumina
22 June 2015

2. Contents Overview of sample workflow in technology hubs
Timings
QC steps
Overview of the SMP2 NGS panel and analysis
Explanation of Nextera protocol
Overview of the validation at Illumina
Analysis / How aberrations are detected
How the aberrations will be reported
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

3. TH Operations Clinical Hub Technology Hub Matrix Trial / CRUK XML XML
Sample prep and assessment
Clinical Hub
XML receipt and logging in
Allocation of test
Macro-dissection and DNA extraction
Testing
Technology Hub
Receipt of results
Monthly KPIs to CR-UK
Archive results to FTP
Patient enters appropriate trial arm
Matrix Trial / CRUK
XML
You will hear quite a lot about this entire process today. This is an OVERVIEW.
All three THs have established very similar operations, but not identical. In each case utilising their existing diagnostic workflows and expertise – would have been very difficult to establish from “scratch”
Sample journey: XML & request receipt – blood and tumour samples (not always at the same time) – booking-in using usual sample reception, checks of sample identifiers and assignment of tests. Sorting out of discrepancies……
Macrodissection and DNA extraction (Cardiff: 2 FTE+) – huge time resource, but essential for good quality analyses. No easy automation!!
Molecular analysis – organised by gene, NOT tumour. In Cardiff, team of 5 or 6 GTs. All analyses checked by two. Integrated into main lab work (e.g. KRAS, EGFR…) NGS developed throughout and then launched in June 2013 for final ~ samples. (NGS development expected from TSB).
Results recombined by tumour. Checking, checking, checking. Transcription errors avoided. Can not be reported until all results ready.
Result returned to CH by XML
KPI data direct to CR-UK: No. of samples, failures, RTs, clinical trials – major resource required (not funded)
XML
XML
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

4. Sample processing timings at the TH
Pre-testing
Post-testing
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

5. Pre-testing Day 1 Day 2 Day 3 Day 4 Day 5-10 Paired sample receipt
Booked in
Triaged
Assign extraction / send for quantification
Day 2
DNA extraction
Day 3
DNA quantification
>50ng – proceed to testing
<50ng – send Pre-Test QC fail report for tumour and bank blood DNA
Day 4
Select next 5 pairs ready for testing
Prepare worksheets, etc
Day 5-10
Nextera enrichment
MiSeq run
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

6. QC step Work done by Illumina to identify a QC step
Three samples run at different concentrations.
Lower than 30ng input DNA, no sequencing data is generated.
50ng input DNA to the NGS analysis is optimal.
50ng
30ng
20ng
10ng
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

7. Factors affecting transitioning from Day 3 to Day 4
Batching
5 pairs processed per NGS run
10 pairs processed max per week (2 MiSeq runs) (40 samples per month)
Sample numbers
Too few – waiting for samples to activate a batch
Too many – samples in a queue waiting to start testing
Timings
Protocol for enrichment is complex – 5 day protocol
Safe stopping points that are built into a working week
Sample pair that is QC ready on a Thursday will wait until following Wednesday to start testing
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

8. Post-testing Day 11 Day 12 Day 13 Day 14 Day 15
Raw data retrieved from MiSeq
Scientist 1 processes data and performs 1st analysis
Day 12
Scientist 2 performs 2nd check
Scientist 3 validates test result entry onto laboratory database
Day 13
Reports written
Day 14
Reports checked and authorised by senior scientist
XML generated
Day 15
XML reports available for retrieval by CH from sFTP site
From start of test it takes ~10 days to turnaround a sample. The whole process from sample receipt should take ~15 days but because of the issues highlighted in the previous slide this is variable
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

9. Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

10. Nextera hybridisation
CR-UK NGS Panel 2
Nextera hybridisation
28 gene
Must have matched blood sample and tumour % information for analysis stage
Detects SNVs, insertions/deletions, CNV, translocations
Developed in partnership with Illumina
Panel linked to Matrix Trial
Increased gene spectrum
Increased mutation spectrum
AKT1
ALK
BRAF
CCND1
CCND2
CCND3
CCNE1
CDK2
CDK4
CDKN2A
EGFR
FGFR2
FGFR3
Her2*
HRAS
KRAS
MET
NF1
NRAS
NTRK1
PIK3CA
PTEN
RB1
RET
ROS1
STK11
TSC1
TSC2
Single test that requires less DNA, analyses more genes, all types of mutational events, better quality result
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

11. Nextera library prep
Patient DNA is processed to generate a sequencing library
collection of DNA fragments derived from the patient
DNA fragments are flanked by adaptor sequences
Allow fragments to ‘stick’ to the flow cell surface
Transposons fragment and tag the DNA with primers
Indexes are added to each sample so they can be separated
Uses an enrichment or capture approach
positively select fragments of interest from your sequencing library

12. SMP2 Nextera NGS panel validation overview
Validation was performed at Illumina
Assessment of how panel works
All types of variant were validated, but not CNV as no clinical samples were available
Used sample previously tested by THs in their routine clinical pathways
Cell line DNA provided by Horizon Discovery
translocations
known variant frequency
Technology transfer to TH
TH ran 3 Nextera panels in house
Panel 1, cell lines
Panel 2, previously analysed blood and tumour
Panel 3, ‘real’ SMP2 samples
No issues with the transfer of the wet protocol.
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

13. Validation conclusions
Combining the results of the validation for SNV and indels; we know with 95% certainty that the TH can detect SNV and indels at >10% overall allele frequency in at least 95% of cases.
CNV are detectable in samples with high tumour percentage and if the CNV is large.
Confident in calling >5 copy number increases
TH will request a retrospective FISH slide to confirm low level or suspected copy number variants.
Homozygous deletions are not detectable in samples with <60% of neoplastic nuclei, because of contamination from normal tissue.
The TH will request a retrospective FISH slide to confirm the deletion is homozygous.
Translocations detected but caution required with poorer quality samples where less read depth achieved
Prospective validation by FISH analysis, through analysis of a number of translocation negative samples.
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

14. How aberrations are detected
Development of analysis tool by the THREE TH
4 stages to analysis 1
Raw Data output from MiSeq run 2
Analysis of raw data using commercially available (Illumina Variant Studio), Open source tools (Manta), or manually 3
Collation of detected variants
Excel spreadsheet containing algorithms developed by TH based bioinformaticians
Information on variants detected and coverage achieved across the panel test 4
Assessment of pathogenicity / eligibility for Matrix Trial
Scientific assessment
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

15. Detecting SNV and indels is automated.
Patient raw data from MiSeq (blood and FFPE)
MiSeq reporter
Windows
BWA alignment to human genome
(generate BAM files) 2 1
Flag and remove duplicate reads that span chromosomes
Coverage calculator
(FFPE)
Somatic variant caller SNV and indels in FFPE and blood
(generates gVCF files)
Min 3 events
Manta variant calling
Structural changes
(FFPE)
Min 3 events
Using coverage at each base per gene calculates;
Average depth per gene
Min and max depth per gene
Mean depth per gene
Variant studio
Germline variants removed from FFPE 2
Germline variants removed from FFPE
Virtual machine
Unix
Pass filter 2 2 3
Excel
SNV calling;
Filter known SNP out
Filter 10% allele freq and min 10 reads
Manual raw data check if needed
CNV calling;
Plot graph of mean coverage per gene/mean coverage per sample
Filter gains >5 fold
Filter loss <0.5 fold
Structural variant calling;
See if Manta has pulled out a variant.
WT calling;
Add tumour %
Look at % of bases in each gene coverage to required depth
Consider hotspot genes
Pass or fail gene
Bioinformatics pipeline has required significant development by THs with Illumina
Detecting SNV and indels is automated.
CNV and translocation detection is not automated as there is no commercially available software.
Detecting CNV will always be complex with NGS due to contamination of normal tissue. 4
Compare to tier variant list from Pharma
Report in XML

16. How aberrations will be reported
General rules
Variants are being reported in an XML format.
Split into 3 tiers by the pharmaceutical partners.
Tier 1= trial eligible
Tier 2= trial eligible
Tier 3= not trial eligible
These lists will be maintained throughout the programme through quarterly meetings to look at evidence for moving variants between tiers
Confidence scoring system developed to answer the question of how confident can you be that in a given tumour sample no variants have been missed that are above 10% frequency. This considers;
Minimum sequencing coverage across the region of interest
Tumour % of sample
Can only be applied to SNVs and in/dels
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

17. Tumour %, depth of coverage, WT confidence, variant frequency are all linked
The number of times a DNA base or gene is sequenced is called the depth/coverage or number of reads/read depth
This is what makes NGS a quantitative assay
100% tumour material has a KRAS c.35G>A p.(Gly12Asp) present at 20%
In 100 reads at base c.35, 75 reads =G and 25 reads = A
NGS allows us to look for low level variants ie ‘needle in a haystack’
Based on validation, test sensitivity set at 10%
Confidently call variants where we detect 10 reads in 100
100% tumour material where we have 100 uniform reads at any given site we can be confident that we have not missed any variant that was present above 10%
Most material sent for testing will be less than 100% tumour even if macrodissected so this will either
Decrease sensitivity of detection if keep a given read depth eg 100 reads
Maintain sensitivity by increasing read depth
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

18. Tumour % information critical for analysis
Variant present at 10% frequency – what % of reads would be mutant
Mutant reads
10 mutant reads minimum so minimum read depth required is 5%
0.5%
1/200
2000
10% 1%
1/100
1000
20% 2%
1/50
500
40% 4%
1/25
250
50%
1/20
200
100%
1/10
100
Tumour % information critical for analysis
Really important to define this to the nearest 10%.
Coverage calculator works out % of bases within gene covered to the depth required
Gene is passed or failed based on criteria below

19. How aberrations will be reported
Wild type samples or failed samples
Wild type samples will say how confident the TH are this is a true WT for SNV and indels.
Samples that fail either NGS or QC step are eligible for a Matrix Trial re- biopsy.
TH will never say no translocation or no copy number variation as still developing confidence value for this.
Test Result
Test Report
Wild type
No variant detected
High/medium confidence
Gene test failed
No result
Fail. Repeat sample requested if available.
QC step failed
Not tested
Failed QC step- insufficient sample. Repeat sample requested if available.
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

20. How aberrations will be reported
Variants detected
The general format for variants that will be in the ‘Test Result’ field are;
Type of variant
Test Result
Single nucleotide variants and small indels
(Mutation/Sequence change/In dels)
c.codon number and base change p.(amino acid number and change)
Translocation
Gene_Gene_exons
Copy number variant- deletions
Whole gene deletion homozygous
Whole gene deletion heterozygous
Exons deletion homozygous
Exons deletion heterozygous
Copy number variant - amplification
Whole gene amplification ± confirmed by FISH
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

21. How aberrations will be reported
Variants detected
The ‘Test Report’ field will contain text around the variant tier, therefore whether the variant makes the patient potentially trial eligible.
There will also be text to say if a FISH confirmation confirmed the variant.
If FISH slides were requested by the TH but could not be obtained, this will be in the comments field of the XML.
Test Report
Meaning- trial eligibility
Tier 1
Potentially trial eligible
Tier 2
Tier 3
Confirmed by FISH
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

22. How aberrations will be reported
Examples of reporting – WHOLe panel
Gene
Test Status
Test Result
Test Report
AKT1
Success
c.49G>A p.(Glu17Lys)
Tier 1 AKT1 variant
ALK
EML4-ALK
Tier 3 ALK translocation
BRAF
c.1799T>A p.(Val600Glu)
Tier 3 BRAF variant
CCND1
No variant detected
Medium confidence Wild Type
CCND2
Complete Fail
No result
Fail. Repeat sample requested if available.
CCND3
CCNE1
CDK2
CDK4
Amplification
CDKN2A
Whole gene deletion homozygous
Tier 1 CDKN2A homozygous deletion confirmed by FISH
EGFR
c.2235_2249del15 p.(Glu746_Ala750del) 
Tier 3 EGFR 14bp duplication
FGFR2
FGFR2-TACC2
Tier 3 FGFR2 translocation
FGFR3
Her2
Exon 5 and 6 deletion heterozygous
Tier 3 Her2 exon 5 and 6 heterozygous deletion
HRAS
High confidence Wild Type
KRAS
MET
Exon 5 deletion heterozygous
Tier 3 MET exon 5 heterozygous deletion
NF1
Partial fail
c.135T>A p.(Asn45Lys)
Tier 2 NF1 variant
NRAS
c.183A>C p.(Gln61His)
Tier 1 NRAS variant
NTRK1
CD74-NTRK1.C3N13
Tier 3 NTRK1 translocation
PIK3CA
c.1624G>A p.(Glu542K) and c.1616C>G p.(Pro539Arg)
Tier 1 and tier 2 PIK3CA variant
PTEN
c.106G>C p.(Gly36Arg)
Tier 2 PTEN variant
RB1
RET
Partial Fail
Low confidence Wild Type
ROS1
CD74-ROS1_C6:R34
Tier 1 ROS1 translocation
STK11
c.523_528del6 (delAAGGAC) p.(Lys175_Asp176del)
Tier 2 STK11 6bp deletion
TSC1
c.2647G>A p.(Ala883Thr)
Tier 2 TSC1 variant
TSC2
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

23. How aberrations will be reported
Monday 22 June 2015
Lung Cancer Research Stratified Medicine Educational Event- Birmingham

24. Stats so far … 147 samples received
48 (33%) failed before testing - not enough DNA
78 reported full 28 gene panel
9/78 (12%) all genes failed post testing
21 in progress
Detected sequence variants 49/78 patients (63%)
78 sequence variants detected in 49 patients
39/49 patients with Tier 1 mutations
Some with multiple Tier 1 mutations

25. 78 Variants Variants detected in 17/28 genes on panel AKT1 ALK BRAF 2
CCND1 2
CCND2
CCND3
CCNE1 1
CDK2
CDK4 2
CDKN2A 6
EGFR 8
FGFR2 2
FGFR3
Her2*
HRAS
KRAS 15
MET 2
NF1 10
NRAS
NTRK1
PIK3CA 10
PTEN 1
RB1 7
RET
ROS1 2
STK11 6
TSC1 1
TSC2 1
Variants detected in 17/28 genes on panel

26. Thank you cruk.org CRUK SMP2 Technology Hubs Illumina
Birmingham- Mike Griffiths, Jennie Bell, Fiona MacDonald, Pauline Rehal, Alessandro Rettino, Sam Clokie
Cardiff- Rachel Butler, Ian Williams, Michelle Wood, Helen Roberts
RMH- David Gonzalez de Castro, Lisa Thompson, Keeda Dover, Brian Walker, Lisa Grady
Illumina
David McBride
Mark Ross