a next-generation sequencing assay to estimate tumor mutation load at > 5% allelic frequency
TRANSCRIPT
Thermo Fisher Scientific • 5791 Van Allen Way • Carlsbad, CA 92008 • www.lifetechnologies.com
Ruchi Chaudhary1, Dinesh Cyanam, Vinay Mittal, Nick Khazanov, Paul Williams, Warren Tom, Janice Au-Young, Seth Sadis, and Fiona Hyland - 1Thermo Fisher Scientific × 200 Oyster Point Blvd, South San Francisco, CA 94080
A Next-Generation Sequencing Assay to Estimate Tumor Mutation Load at > 5% Allelic Frequency
Immunotherapies have shown anti-cancer effects in melanoma, NSCLC, and bladder cancer. High tumor mutation load is associated with positive responses from immune checkpoint inhibitors. However, current methods to estimate tumor mutation load often have high infrastructure needs, and require large amounts of DNA.
Herein, we characterize the ability of a targeted panel with low sample input requirements to estimate mutation load from tumor research samples.
A simple workflow has been developed on the Ion Torrent sequencing platform with an AmpliSeq panel to estimate per Mb somatic mutation burden from FFPE and fresh frozen tumor research samples. This solution will advance research in immuno-oncology.
INTRODUCTION & METHOD
RESULTS
CONCLUSIONS
Oncomine Tumor Mutation Load (TML) Assay Workflow
For Research Use Only. Not for use in diagnostic procedures.
To measure the limits of variant detection, we used a custom DNA sample that was developed using AcroMetrix™ MegaMix™ Technology in a Genome in a Bottle GM24385 background. Mutations were injected at frequency (dilution) from 50% to 5%. Each dilution contains 339 common tumor mutations interrogated by our assay after germline filtering. TML assay
Ø Achieves > 82% sensitivity (Figure 3A) and > 93% positive predictive value (PPV, i.e., true variants w.r.t. all variant calls) at and above 15% dilutions (Figure 3B), showing the strength of the assay at 30% tumor content.
Ø Is optimized to be highly specific in variant detection for estimating accurate somatic mutation load.
Ø Is highly reproducible as seen from close estimates from library replicates.
Figure 3A. Sensitivity of the assay in detecting true variants at different dilutions
Figure 3B. PPV of the assay in detecting true variants at different dilutions
We tested performance on varying sequencing depths through down-sampling FFPE and control (50% dilution based on AcroMetrix™ MegaMix™ in background GM24385). We ran both samples about 1 sample per chip to obtain deep sequencing BAMs. We down-sampled BAMs to 50% of initial size interactively, and ran our analysis workflow on each resulting BAM, to find
Ø Reduction in mutation load with decreasing read depth on FFPE sample (Figure 5A).
Ø Slight increase in sensitivity (Figure 5B) and PPV (Figure 5C) with increasing read depth in control.
Separation of Mutation Load High and Low Samples
Figure 5A. Mutation count per Mb for each down sample of initial FFPE sample
Figure 5B. Sensitivity on varying average read depths of control
Figure 5C. PPV on varying average read depths of control
Run Sequence
Prepare Library
SampleInput AnalyzePrepare
Template
Oncomine Tumor
Mutation Load Assay
FFPE IR 5.6Ion S5™ System (540)
Ion Chef™ System
• 2 pool assay• Manual and
automated library workflows
• gDNAextracted from FFPE tissue
• Ion 540 Kit -Chef
• Ion 540 chip –multiplex up to 8 unique libraries per chip
• Ion 540 Kit -Chef
• Automated uploading and analysis
Assay SpecificationsPanelq 409 key cancer genes (Oncogenes
& Tumor Suppressor Genes)q 1.7 Mb genomic footprintLibraryq Ion AmpliSeq™ Library Kit Plusq 20 ng input DNA requirement Analysis Workflowq SNV calling at > 5% allelic
frequency q Germline variant filteringq Detailed result reporting
Sensitivity/Specificity of Variant Detection in Control Sample
Comparison with Tumor-Normal Analyses
Figure 1A (Left). Whole Exome Sequencing (WES) data of 21,056 samples downloaded from COSMIC v80. Somatic mutations were restricted to Oncomine TML panel targets. Somatic mutations in exomes strongly correlated with mutation counts by Oncomine TML with r2 = 0.968, demonstrating suitability of panel size and appropriateness of its targets.
TML
Mut
atio
n C
ount
Figure 2. Correlating somatic mutation count by TML panel with that of tumor-normal analysis
Analysis Result Report
Figure 1B (Right). Clinical trial, WES data for 110 melanoma subjects treated with ipilimumab (anti-CTLA4) was downloaded from Van Allen et. (2015 Science 350:207-211) with response status. Somatic mutation counts were restricted to Oncomine TML panel targets. Significant difference (through Mann–Whitney Exact test) in mutation counts of responders and non-responders observed with p = 0.00498, demonstrating prognostic capability of Oncomine TML assay.
TML
Mut
atio
n C
ount
In Silico Analyses
Figure 6. Analysis result report of an FFPE lung tumor sample, first page (above) and second page (right), containing analysis settings, sample information, QC metrics, and analysis results displaying allele ratio distribution, substitution type and context information of somatic mutations. Notice 47% somatic mutations are consistent with smoking damage in this lung tumor.
To test the ability of Oncomine TML assay in counting only somatic variants (i.e., removing germline variants and systematic noise), tumor-normal analysis was performed on colon and lung tumors with matched normal samples. Oncomine TML assay ran on only tumor samples. We observed
Ø Mutation count by Oncomine TML assay strongly correlates with that of tumor-normal analysis with r = 0.9233 (Figure 2).
r = 0.9233
Performance on Varying Sequencing Depths
We ran Oncomine TML assay on a batch of 7 CRC tumor samples in which 3 had known MSI high and 4 MSI low status, to learn
Ø TML assay successfully stratify mutation high and low samples as MSI high correlates to mutation high in CRC tumor (Figure 4).
Figure 4. Separating MSI high and low samples