This document summarizes the analysis of variant calling using trio information from the NA12878 pedigree. It finds that jointly analyzing related samples using the RTG Family caller, which leverages Mendelian inheritance patterns and shared haplotypes, improves genotype calls by reducing errors compared to singleton calling. The trio analysis corrects over 60x more Mendelian inconsistency errors compared to singleton calling and identifies variants that segregate in offspring as expected for heterozygous and homozygous-alternate sites.

1. ©
2013
Real
Time
Genomics,
Inc.
NA12878
Trio/Pedigree
Analysis
Francisco
M.
De
La
Vega,
D.Sc.
VP
Genome
Science

2. ©
2013
Real
Time
Genomics,
Inc.
Leveraging trio information
• GiaB has selected reference materials in the form of father,
mother, offspring trios
• The goal was to leverage the Mendelian inheritance patterns
to:
– Identify variant genotype errors that are inconsistent with
Mendelian inheritance
– Remove these errors from the reference baseline calls
• However, if variant identification methods don't use directly
pedigree information and jointly analyze the trio alignments,
an opportunity to improve the genotype calls would be
missed
• We focused on using the RTG Family caller to better leverage
the shared information in the trios and improve the call set,
whilst reducing Mendelian inconsistent genotype errors

3. ©
2013
Real
Time
Genomics,
Inc.
C
AA
A
A
A
A
A
A
A
A
A
A
A A/Genotype:
A A
CA
C
C
A
A
A
A
A /Genotype: C
C
A /Genotype:
AC
C
C
|
||
Variant calling can be improved by jointly
analyzing related samples
Shared
haplotypes

4. ©
2013
Real
Time
Genomics,
Inc.
C
AA
A
A
A
A
A
A
A
A
A
A
A A/Genotype:
A A
CA
C
C
A
A
A
A
A /Genotype: C
C
A /Genotype:
AC
C
C
|
||
Variant calling can be improved by jointly
analyzing related samples
Mendelian
variant
segregaJon
Shared
haplotypes

5. ©
2013
Real
Time
Genomics,
Inc.
Mendelian inconsistency
C
C
/Genotype: C
C
C
C
C
C
C
A
A
A
A A/Genotype: (Low QV)
C
A
A
A
A
A
A /Genotype:
C
C
C
A
A A
CC
AC
|
||

6. ©
2013
Real
Time
Genomics,
Inc.
Joint trio analysis corrects Mendelian errors
C
/Genotype: C
C
C
C
C T
G
G
G
C T
C T
C T
C
A
A
A
A
A
Genotype:
C
A / C
G
G
G
G
G
G
G
A
A
A
Genotype: (Good QV)
C T
C T
C T
C T
A / C
G
G
G
A A
CC
AC
|
||

7. ©
2013
Real
Time
Genomics,
Inc.
NA12878 calls from trio calling
• Comparing offspring variants from singleton vs
pedigree calling
– Both showing good quality metrics
• Using family information more good calls can be
made and dubious calls are downgraded
NA12878
Call
set SNVs Indels MNPs
SNV
Het/Hom Ti/Tv
%
dbSNP
(r129)
RTG
single
3,329,797 558,242 31,070 1.55
2.11
90.8%
RTG
trio
3,363,619 595,030 33,686 1.57
2.11
90.4%
GATK/VQSR
3,263,289 610,837 N/A 1.51
2.09
91.7%
Variant
StaBsBcs
Data:
WGS
2x100bp
>50X
Illumina
PlaJnum
Genomes
data
(ENA
Acc.
No.
ERP001960).
RTG
AVR
score
cut-­‐off
0.15;
GATK
v1.7
&
BWA
0.6.1.
142,848
68,000
Family
Singleton
3,849,457
NA12878
NA12891 NA12892

8. ©
2013
Real
Time
Genomics,
Inc.
NA12878 vs reference datasets
NA12878
Call
set
1kP
OMNI
Poly
(TP%)
1kP
OMNI
Mono
(FP%)
Get-­‐RM¶
(TP
%)
GiaB
(TP%)
GiaB-­‐BED
(TP%)
RTG
single
97.5%
0.10%
97.4%
N/A
N/A
RTG
trio
97.5%
0.24%
97.0%
90.5%
94.1%
GATK/VQSR
97.8%
0.17%
87.8%
88.4%
92.5%
§
RelaJve
to
dbSNP
137;
StaJsJcs
for
SNVs
only.
¶Get-­‐RM
consistent
high-­‐quality
variants;
n=498
NA12878
NA12891 NA12892
– 1000 Genomes Illumina OMNI SNP array
• Polymorphic sites – TP proxy
• Monomorphic sites – FP proxy
– Get-RM high confidence call set
– GiaB high confidence calls in BED region

9. ©
2013
Real
Time
Genomics,
Inc.
ROC Trio calls vs. GiAB baseline (BED)
RTG
snpsimeval
tool;
SNV/indel/MNP;
zygosity
match

10. ©
2013
Real
Time
Genomics,
Inc.
ROC Trio calls vs. GiaB baseline
RTG
snpsimeval
tool;
SNV/indel/MNP;
zygosity
match

11. ©
2013
Real
Time
Genomics,
Inc.
ROC Trio calls vs. CGI baseline
RTG
snpsimeval
tool;
SNV/indel/MNP;
zygosity
match

12. ©
2013
Real
Time
Genomics,
Inc.
Mendelian inconsistency errors
RTG family caller reduces Mendelian Inheritance Errors over 60X vs. RTG
singleton calling (over 70X vs. GATK/VQSR)
Log
Counts
of
MIE
1
10
100
1000
10000
100000
1000000
RTG
single
RTG
trio
GATK/VQSR
335,625
4,870
351,904

13. ©
2013
Real
Time
Genomics,
Inc.
Pattern #1: Heterozygous variant
TrioCalling
NA12878
NA12892NA12891
NA12877
NA12889 NA12890
NA12879 NA12880 NA12881 NA12882 NA12883 NA12884 NA12885 NA12886 NA12887 NA12888 NA12893
0/1
0/10/0
0/0 0/0 0/00/0 0/00/1 0/1 0/10/10/1

14. ©
2013
Real
Time
Genomics,
Inc.
Segregation of heterozygous variants
0
20,000
40,000
60,000
80,000
1
2
3
4
5
6
7
8
9
10
11
SNV
count
#
of
offspring
segregaBng
SNV
0
100
200
300
400
500
1
2
3
4
5
6
7
8
9
10
11
MNP
count
#
of
offspring
segregaBng
MNP
0
2,000
4,000
6,000
8,000
10,000
1
2
3
4
5
6
7
8
9
10
11
indel
count
#
of
offspring
segregaBng
indel
0
20,000
40,000
60,000
80,000
100,000
1
2
3
4
5
6
7
8
9
10
11
Variant
count
#
of
offspirng
segregaBng
All
Variants
SegregaJon
of
NA12878
heterozygous
variants
called
as
family,
GQ>50,
homozygous
reference
in
other
parent.

15. ©
2013
Real
Time
Genomics,
Inc.
Pattern #2: Homozygous-alt variant
TrioCalling
NA12878
NA12892NA12891
NA12877
NA12889 NA12890
NA12879 NA12880 NA12881 NA12882 NA12883 NA12884 NA12885 NA12886 NA12887 NA12888 NA12893
0/1
1/10/0
0/1 0/1 0/10/10/10/1 0/1 0/1 0/1 0/1

16. ©
2013
Real
Time
Genomics,
Inc.
Segregation of homo-alt variants
0
20,000
40,000
60,000
80,000
100,000
120,000
1
2
3
4
5
6
7
8
9
10
11
SNV
count
#
of
offspring
segregaBng
SNV
0
100
200
300
400
500
600
700
1
2
3
4
5
6
7
8
9
10
11
MNP
count
#
of
offspring
segregaBng
MNP
0
2,000
4,000
6,000
8,000
10,000
12,000
1
2
3
4
5
6
7
8
9
10
11
indel
count
#
of
offspring
segregaBng
indel
0
20,000
40,000
60,000
80,000
100,000
120,000
1
2
3
4
5
6
7
8
9
10
11
Variant
count
#
of
offspring
segregaBng
All
Variants
SegregaJon
of
NA12878
homozygous
alternaJve
variants
called
as
family,
GQ>50,
homozygous
reference
in
other
parent.

17. ©
2013
Real
Time
Genomics,
Inc.
False positive estimate by segregation
GT
Type
All
variants
SNV
MNP
indel
Het
TP
(10-­‐11)
123672
110262
693
12717
FP
(1-­‐8)
1901
1000
47
854
FP%
1.40%
0.88%
1.42%
5.67%
Homo-­‐alt
TP
(2-­‐10)
373260
329642
2258
41360
FP
(1,11)
4457
3672
36
749
FP%
1.18%
1.10%
1.57%
1.78%
Overall
TP
496932
439904
2951
54077
FP
6358
4672
83
1603
Overall
FP%
1.26%
1.05%
2.74%
2.88%

18. ©
2013
Real
Time
Genomics,
Inc.
Data imputation by pedigree caller
• For genomes with no data use population priors
– With care can iterate over offspring then each of parents
independently
– Avoid exponential explosion so can do whole extended
family in one calling step

19. ©
2013
Real
Time
Genomics,
Inc.
Imputation of family members with no data
Simulated
data
True
PosiJves
False
PosiJves
1
offspring
2
offspring
4
offspring
4
offspring
+
father

20. ©
2013
Real
Time
Genomics,
Inc.
ROC vs NA12878 imputed baseline
RTG
snpsimeval
tool;
SNV/indel/MNP;
zygosity
match

21. ©
2013
Real
Time
Genomics,
Inc.
de novo mutation identification
Call
set
de
novo
candidates
de
novo
germline*
de
novo
somaBc*
TP/FP
Singleton
calls 16,902 49
(100%)
941
(99%)
1:17
Trio
calls 2,205 49
(100%)
941
(99%)
1:2.2
de
novo
MutaBon
Accuracy
(NA12878)
*SensiJvity
vs.
Conrad
et
al.
(2011)
validated
dataset
of
germline
and
somaJc
cell
line
de
novo
mutaJons.
– Uses the parental genomes to identify & score de novo
mutations in offspring
– Greater than 7X improvement in precision to find de novo
mutations vs. naïve methods
NA12878
NA12891 NA12892

22. ©
2013
Real
Time
Genomics,
Inc.
Status
• Working through the complete trio datasets for
producing joint pedigree calls for NA12878 trio
– Aiming for a trio call set and another that
includes full Platinum pedigree data
– There is disproportionally more data for
NA12878 than her parents or offspring
• Comprehensive segregation analysis that
includes all Mendelian patterns
• Phasing analysis to identify variants that are
inconsistent with transmitted phases

23. ©
2013
Real
Time
Genomics,
Inc.
Issues
• How to integrate pedigree calls with other data?
– Variants that segregate appropriately
candidates for inclusion in baseline
– Variants that don’t segregate appropriately
candidates for removal of baseline
– Improvement of baseline genotypes using
pedigree-based genotypes
• Use of the imputed NA12878 baseline
• Creation of a more inclusive baseline for ROC
curves to compare new methods and select
thresholds

24. ©
2013
Real
Time
Genomics,
Inc.
Acknowledgements
• RTG team at Hamilton, New Zealand
– Led by John Cleary, CTO
• RTG team at San Bruno, CA
– Sahar Malakshah
– Minita Shah
– Brian Hilbush
• Michael Eberle, Illumina, Inc. – Platinum Data
• Justin Zook, NIST
• 1000 Genomes Project
©
2013
Real
Time
Genomics,
Inc.
All
rights
reserved.
US
Patent
7,640,256.
Other
patents
pending.
For
research
use
only.
Not
for
diagnosJc
applicaJons.