CLIA Hurdles for NGS — CAP Perspective - Cancer Diagnosis ...

hordeprobableBiotechnology

Oct 4, 2013 (3 years and 6 months ago)

98 views

Shashikant Kulkarni, M.S (Medicine)., Ph.D., FACMG

Associate Professor of Pediatrics, Genetics, Pathology and Immunology

Medical Director of Genomics and Pathology Services



High throughput


T
est
many genes at once


Systematic, unbiased mutation detection


All mutation
types


Single nucleotide variants (SNV), copy number
alteration (CNA)
-
insertions, deletions and
translocations


Digital readout of mutation frequency


Easier to detect and quantify mutations in a
heterogeneous
sample


Cost effective
precision

medicine


“Right drug at right dose to the right patient at
the right time”



Cancer genomes are extremely complex and
diverse


Mutation frequency


Degree of variation in cancer cells compared to
reference genome


Copy number/
ploidy


Most tumors are
aneuploid


Bioinformatic software assume diploid status


Genome structure


Genomic alterations in cancer found at low
-
frequency


Samples vary in quantity, quality and purity from
constitutional samples


Quantity


Limiting for biopsy specimens


Whole genome amplification not ideal


Quality


Most biopsies are formalin fixed, require special protocols


Often include necrotic, apoptotic cells


Purity (tumor heterogeneity)


Admixture with normal cells (need pathologists to ensure test is
performed on DNA from tumor cell)


Within cancer heterogeneity (different clones)



FFPE (formalin
-
fixed, paraffin
-
embedded)
samples


Age, temperature, processing


Fresh tissues


Not ideal without accompanying pathology
investigation and marking of tumor cell population to
guard against dilution effect on total DNA extracted


Fine needle biopsies


Very few cells available


NGS methods will need to work by decreasing
minimum inputs of DNA




Goals


High throughput, cost effective multiplexed
sequencing
assay with
deep coverage


Target
clinically actionable regions in clinically
relevant time


Challenges


Huge infrastructure costs


Bioinformatic
barriers


Solution


Leverage expertise and resources across
Pathology, Bioinformatics and
Genetics

From “soup to nuts”

Test overview

Genes

Disease

ALK

Lymphoma, Lung

BRAF

Brain, Colon, Lung, Melanoma, Thyroid

CEBPA

AML

CTNNB1

Colon, Desmoid Tumor, Liver, Lung, Prostate, Renal, Thyroid

CHIC2

Myeloid Neoplasms w/Eosinophilia

CSF1R

AML, GIST

DNMT3A

AML

EGFR

Colon, Lung

FLT3

AML

IDH1

AML, Brain

IDH2

AML, Brain

JAK2

Myeloproliferative Neoplasms

KIT

AML, GIST, Systemic Mastocytosis

KRAS

Colon,
Endometrium
, Lung, Melanoma, Pancreatic, Thyroid

MAPK1(ERK)

Lung, Melanoma

MAPK2(MEK)

Lung, Melanoma

MET

Lung, Melanoma

MLL

AML

NPM1

AML

NRAS

Colon, Lung, Melanoma, Pancreatic, Thyroid

PDGFRA

GIST, Sarcoma

PIK3CA

Colon, Lung, Melanoma, Pancreatic

PTEN

Brain, Endometrium, Melanoma, Ovarian, Prostate, Testis

PTPN11

JMML, MDS

RET

MEN2A/2B (adrenal), Thyroid

RUNX1

AML

TP53

Colon, Lung, Pancreatic

WT1

AML, Renal,
Wilms

Tumor

Exons +/
-

200 bp, plus 1000 bp +/
-

each gene

AUG

STOP

TSS

poly(A)

promoter

splice signals


Capture efficiency and coverage


Overall and by gene


Specimen type differences


Fresh
-
frozen vs. FFPE specimens


Detection of single nucleotide variants (SNVs)


Methods


Filters


Detection of indels and other mutation
types


Methods

HapMap

samples

Known
genotypes

lung
adenocarcinomas

Known
genotypes

frozen
DNA sample

+

FFPE
DNA sample

Library prep, target enrichment

Multiplex
sequencing

Analysis

(coverage and comparison with genotypes)

Coverage

Capture baits

Target region

1000x

500 bp

Coverage

1000x

Capture baits

Target region

500 bp

Good coverage of EGFR

Poor coverage of CEBPA

NA19129 coverage distribution by gene

(black bar = median; box = 25
-
75%ile)

*

*

Capture for genes with poor coverage
have been redesigned

Tumor 1 normalized coverage, by gene

(solid = frozen, hatched = FFPE)

Only minor differences are apparent
between fresh
-
frozen and FFPE data

http://www.cdc.gov/genomics/gtesting/ACCE/

Accuracy

Degree of agreement between the nucleic acid sequences
derived from the assay and a reference sequence

Precision

Repeatability

degree to which the same sequence is derived in
sequencing multiple

reference samples, many times.
Reproducibility

degree to which the same sequence is derived
when sequencing is performed by multiple operators and by
more than one instrument.

Analytical
Sensitivity

The likelihood that the assay will detect a sequence variation, if
present, in the targeted

genomic region.

Analytical
Specificity

The probability that the assay

will not detect a sequence
variation, if none are present, in the targeted genomic region.

Diagnostic
Specificity

The probability

that the assay will not detect a clinically
relevant sequence variation, if none are present, in the targeted
genomic region.

Test Type

Definitions

Inter
-
Tech (Stringent)

The technicians performing the run were different, but

the
experiment and lanes were the same.

Inter
-
Tech (Relaxed)

The technicians performing the run were different for each
comparison. We did not control for the experiment

or lane.

Intra
-
Tech

The technician performing the run was the same. The
experiment was different.

Inter
-
Lane (All)

The lanes are different. These experiments, the techs were
different

in two, and the same in two.

Inter
-
Lane &

Intra
-
Tech

The

lanes are different. In these experiments, the techs were
the same.

Intra
-
Lane &

Inter
-
Tech

The lanes are the same. In these experiments, the techs were
different.

98.1%

97.9%

97.1%

98.6%

98.7%

98.4%

90.0%
92.0%
94.0%
96.0%
98.0%
100.0%
102.0%
Inter-Tech
(Stringent)
Inter-Tech
(Relaxed)
Intra-Tech
Inter-Lane (All)
Inter-Lane &
Intra-Tech
Intra-Lane &
Inter-Tech
Percent Agreement

Variability Method

Reproducibility

Major
barriers

for
clinical

implementation

of NGS

Data tsunami

1. Need expertise in Biomedical Informatics

2. Need clinical grade “user
-
friendly
-
soup to nuts” software solution

3. Hardware

Informatics pipeline workflow

Patient

Physician

Sample

Order

Sequence

Tier 1:

Base Calling

Alignment

Variant Calling

Tier 2:

Genome Annotation

Medical Knowledgebase

Tier 3: Clinical Report

EHR

Order Intake


Patient samples accessioned in Cerner
CoPath


Gene panels ordered through
CoPath


Orders received will initiate workflow


HL7

Order Intake

Tier 1 Informatics


Optimized pipelines using several base
callers, aligners, and variant calling
algorithms to meet CAP/CLIA standards


Easily customizable and updateable


Facilitates new panel introduction and the rapid
delivery of novel analytical tools and pipelines


Seamless to the clinical
genomicist

Tier 1 Informatics

Cancer specific analysis pipeline

HiSeq

MiSeq

Novoalign
TM


GATK/
Samtools

Pindel

Breakdancer

SLOPE

Merged

VCF file

Tier 2 Informatics


Deliver a clinical grade variant database
that meets CAP/CLIA standards


Requires combined expertise of
informaticians and clinical
genomocists/pathologists


Future interoperability with
(inter)national variant databases that
meet CAP/CLIA standards

Tier 2 Informatics

Tier 3 Informatics

EGFR (L858R)

Response rates of >70% in patients

with non
-
small cell lung cancer

treated with either
erlotinib

or
gefitinib

KRAS (G12C)

Poor response rate in patients

with non
-
small cell lung cancer

treated with either
erlotinib

or
gefitinib

+

Tier 3 Informatics: Variant
classificaiton

Clinical NGS process map

Conclusions


Cancer
NGS gene
panel helps in
multiplexing key actionable genes for a cost
effective, accurate and sensitive assay


Targeted
cancer panel
are useful
for

drug
repurposing


of small molecule
inhibitors


Clinical validation of NGS assays in cancer
is complex and labor intensive but basic
principles remain


Bioinformatic

barriers are the most
challenging

Karen Seibert,
John
Pfiefer
, Skip Virgin,

Jeffrey
Millbrandt
, Rob
Mitra
, Rich Head

Rakesh

Nagarajan

and his
Bioinf
. team

David Spencer,
Eric
Duncavage
, Andy
Bredm
.

Hussam

Al
-
Kateb
, Cathy Cottrell

Dorie

Sher
, Jennifer
Stratman

Tina Lockwood
, Jackie Payton

Mark Watson, Seth Crosby, Don Conrad

Andy Drury,
Kris
Rickoff
, Karen Novak

Mike Isaacs and his IT Team

Norma Brown, Cherie Moore, Bob
Feltmann

Heather Day, Chad
Storer
, George
Bijoy

Dayna

Oschwald
,
Magie

O
Guin
, GTAC team

Jane Bauer and
Cytogenomics

&
Mol

path
team

MANY MORE!