Polygenic Risk of Breast Cancer in Latinx Populations · Building a Polygenic Risk Score SNP Minor...
Transcript of Polygenic Risk of Breast Cancer in Latinx Populations · Building a Polygenic Risk Score SNP Minor...
Polygenic Risk of Breast Cancer in Latinx Populations
Elad Ziv, MD
Department of Medicine
Helen Diller Family Comprehensive Cancer Center
Institute for Human Genetics
University of California, San Francisco
Disclosures
No conflicts to disclose
Genetic effects on Breast Cancer
Common variants, modest effects
Very rare variant very large odds ratios
Relatively rare variants large effects
Increase risk by 1.05 to 1.3-fold; lifetime risk goes from 12% to 12-13% (some women’s risk will go down), at least 180 of these in large studies of European populations
Increase risk by 2-5 fold or lifetime risk goes from 12% to 20-30
Lifetime risk goes from 12% to 50-80%BRCA1,
BRCA2
PALB2
ATM, CHEK2
>180 SNPs
GWASFamily studies / Sequencing
Building a Polygenic Risk Score
SNP Minor allele frequency
OR per allele Gene
rs35054928 0.4 1.27 FGFR2
rs4784227 0.24 1.23 TOX3
rs62355902 0.16 1.18 MAP3K1
rs3757322 0.32 1.14 ESR1
rs9397437 0.07 1.32 ESR1
7.69 x Risk
Polygenic Risk
Currently some companies provide Polygenic Risk Scores for women who test negative for known susceptibility genes.
Scores are provided only to women of European ancestry.
Genetic Ancestry in Latinos
Bryc et al PNAS 2010
Building a Polygenic Risk Model in Latinas
Methods
SNP selection: SNPs 5x10-8 from prior studies for overall breast cancer (not ER+
or ER-neg BC)
PRS model construction:Use a Bayesian approach and calculate Likelihood ratio for BC for each genotype
Assume SNPs are independent
Approach automatically corrects for allele frequency differences
Lu et al AJHG 2008
Michalidou, et al Nature 2017 Fejerman, et al Nature Comm 2014
StudiesRegion Study Name Sample Sizes Age Family History Estrogen Receptor Status
Positive Negative Unknown
SF Bay Area SFBCS/NC-BCFR Controls 589 53 (11) 55 (9)* NA
Cases 942 50 (11) 190 (20)* 593 (63)§ 230 (19) 119 (13)
Kaiser RPGEH Controls 3563 55 (13) 211 (6)‡ NA
Cases 222 57 (10) 38 (17)‡ 161 (73) 29 (13) 32 (14)
LA County MEC Controls 1469 67 (8) 141 (10)* NA
Cases 532 66 (8) 73 (14)* 303 (57) 108 (20) 121 (23)
COH/CCGCRN Controls 305 52 (11) 26 (9)† NA
Cases 1039 43 (9) 348 (33)† 585 (56) 233 (22) 221 (21)
Mexico CAMA Controls 702 52 (9) 27 (4) NA
Cases 709 52 (10) 50 (7) 116 (16) 52 (7) 541 (76)
COLUMBUS Controls 453 35 (12) 34 (8)‡ NA
Cases 481 57 (13) 23 (5)‡ 140 (29) 41 (9) 300 (62)
Colombia COLUMBUS Controls 768 64 (10) ND NA
Cases 954 52 (10) 49 (5)† 354 (37) 177 (19) 423 (44)
Peru PEGEN-BC Controls 85 ND ND NA
Cases 818 50 (11) 54 (7) 548 (67) 246 (30) 24 (3)
All Controls 7934 57 (13) 494 (6) NA
Cases 5697 52 (12) 825 (14) 2800 (49) 1116 (20) 1781 (31)
Genetic Ancestry
SFBACS/NCBCFR (SF Bay Area) Kaiser RPGEH (SF Bay Area) CCGCRN (LA) MEC (LA County)
CAMA (Mexico) COLOMBUS (Mexico) COLOMBUS (Colombia) PEGEN (Peru)
Risk Models Discrimination: separation of cases from controls
Calibration: Centering of cases and controls at correct risk of disease
Good Calibration, Good Discrimination
Poor Calibration, Poor Discrimination Good Calibration, Poor Discrimination
Poor Calibration, Good Discrimination
PRS Discrimination
Shieh, Fejerman et al, JNCI 2019
180-SNP PRS* 71-SNP PRS†
Controls Cases OR (95% CI)‡ P-trend§ Controls
Cases OR (95% CI)‡ P-trend§
Continuous (per standard deviation)
7629 4658 1.58 (1.52 to 1.64) 7934 5697 1.51 (1.46 to 1.57)
Percentiles of PRS <0.001 <0.001
<10 763 192 0.44 (0.37 to 0.53) 794 276 0.54 (0.46 to 0.63)
10-20 763 233 0.54 (0.46 to 0.64) 793 347 0.68 (0.59 to 0.79)
20-30 763 321 0.74 (0.64 to 0.87) 794 377 0.74 (0.64 to 0.85)
30-40 762 352 0.82 (0.70 to 0.95) 793 429 0.84 (0.73 to 0.97)
40-60 1526 863 1 (referent) 1587 1023 1 (referent)
60-70 764 505 1.1 (1.02 to 1.34) 793 649 1.27 (1.11 to 1.45)
70-80 763 572 1.33 (1.16 to 1.52) 793 701 1.37 (1.21 to 1.56)
80-90 762 744 1.73 (1.51 to 1.97) 793 827 1.62 (1.43 to 1.83)
>90 763 876 2.03 (1.79 to 2.31) 794 1068 2.09 (1.85 to 2.35)Latinas: AUC ROC = 0.63 (0.62 to 0.64) for 180 SNP model0.61 (0.61-0.62 ) for 71 SNP model
European Ancestry AUC ROC = 0.63 (0.629-0.651) for ~313 SNP model
0.615 ( 0.608-0.616) for 75 SNP model
Mavaddat ,et al, AJHG 2019
Mavaddat ,et al, JNCI 2015
PRS by Ancestry
180 SNP PRS 71 SNP PRS
Controls Cases AUROC (95% CI)‡ P odds ratio (95% CI) Controls Cases AUROC (95% CI) P odds ratio (95% CI)‖
All 7622 4658 0.63 (0.62 to 0.64) 1.58 (1.52 to 1.64) 7927 5697 0.61 (0.61 to 0.62) 1.51 (1.46 to 1.56)
By Quartiles of Indigenous AncestryQ1
<29% 2349 721 0.63 (0.61 to 0.66) 0.56 1.67 (1.52 to 1.83) 2455 951 0.64 (0.62 to 0.66) 0.02 1.68 (1.55 to 1.83)Q2,
29 - 42% 2049 1021 0.61 (0.59 to 0.63) 1.51 (1.39 to 1.64) 2117 1289 0.60 (0.58 to 0.62) 1.44 (1.34 to 1.55)
Q3, 42– 54% 1820 1250 0.63 (0.61 to 0.65) 1.57 (1.45 to 1.69) 1869 1537 0.62 (0.60 to 0.63) 1.52 (1.41 to 1.63)
Q4, >55% 1404 1666 0.63 (0.61 to 0.65) 1.56 (1.45 to 1.68) 1486 1920 0.61 (0.59 to 0.63) 1.46 (1.36 to 1.56)
Model Calibration
Shieh, Fejerman et al, JNCI 2019
Can we improve: Signals at
Other Known Loci
0
2
4
6
8
10
-lo
g1
0(p−
valu
e)
0
20
40
60
80
100
Recom
bin
atio
n ra
te (c
M/M
b)
rs4849887
0.2
0.4
0.6
0.8
r2
TMEM185B
RALB
INHBB LINC01101
121 121.1 121.2 121.3 121.4 121.5
Position on chr2 (Mb)
Plotted SNPs
0
2
4
6
8
10
-lo
g1
0(p−
valu
e)
0
20
40
60
80
100
Reco
mbin
atio
n ra
te (cM
/Mb)rs2981582
0.2
0.4
0.6
0.8
r2
FGFR2 ATE1
123.1 123.2 123.3 123.4 123.5 123.6
Position on chr10 (Mb)
Plotted SNPs
0
2
4
6
8
10
-lo
g1
0(p−
va
lue)
0
20
40
60
80
100
Recom
bin
atio
n ra
te (c
M/M
b)
rs4808801
0.2
0.4
0.6
0.8
r2
PIK3R2
IFI30
MPV17L2
RAB3A
PDE4C
LOC729966
KIAA1683
JUND
MIR3188
LSM4
PGPEP1
GDF15
MIR3189
LRRC25
SSBP4
ISYNA1
ELL
FKBP8
KXD1
UBA52
C19orf60
CRLF1
TMEM59L
KLHL26
CRTC1
18.3 18.4 18.5 18.6 18.7 18.8
Position on chr19 (Mb)
Plotted SNPs
Future goal is to capitalize on loci such as these, but need adequate sample sizes to avoid overfitting
Variants in Intermediate Penetrance Genes in
Latinx Populations
Variants of Uncertain Significance (VUS)
Caswell-Jin Genetics in Medicine 2017
VUS in BRCA1 & BRCA2
VUS remain a significant challenge in interpretation of genetic tests
Reclassifying VUSs depends on data and lags in non-European populations
Eggington, ClinGenet 2015
Variants of Unknown Significance in gene panels
Pathogenic and likely Pathogenic Mutations in Known Breast Cancer Susceptibility Genes
Gene No.
frameshift
No.
nonsense
No.
missense
No.
splicing
Total No.
variants
ATM 3 1 2 6
BRIP1 2 2
CDH1 1 1
CHEK2 1 1 17 1 20
NF1 1 1
PALB2 14 3 1 18
PTEN 1 1
TP53 2 2
21 5 19 6 51 (4.8%)
Weitzel et al , Cancer, 2019
1054 Latinas were sequenced. Included women with breast cancer & family history (age<60 1st degree relative) or age<50 or bilateral breast cancer AND BRCA1/2 negative
Pathogenic Mutations in Known Breast Cancer Susceptibility Genes
Weitzel et al , Cancer, 2019
Gene Variant No. Alleles in
cases
/ Total
Chromosomes
in Cases (%)
No. Alleles in
Exac /
Total Chrom
ExAC144(%)
OR (95% CI) P Value
CHEK2 c.707T>C: pL236P 11/ 1104 (1.32) 35/11206 (0.31) 3.2 (1.5-6.5) 0.002
PALB2 c.2167_2168del:
p.M723fs
9/ 1104 (0.85) 5/11216 (0.045) 12.2 (3.1-50.8) 0.0001
PALB2 c.2411_2412del: p.
S804fs
3/1104 (0.28) 1/11202 (0.009) 30.4 (2.4-1582.5) 0.0027
PRS Going Forward
Gravel et al PNAS 2011
Common variants, modest effects
Relatively rare variants large effects
BRCA1, BRCA2
PALB2
ATM, CHEK2
>180 SNPs
Summary• PRS for breast cancer works well in Latinas (approximately
same as in Caucasians)
• Modest (non-significant) attenuation by ancestry
• Intermediate penetrance genes: PALB2 and CHEK2 have recurrent (founder) mutations
• Going forward, prediction from rare variants will need larger sample sizes in minority populations, complemented by functional studies
Acknowledgements
UCSF
Laura Fejerman
Yiwey Shieh
Katie Marker
Donglei Hu
Scott Huntsman
UC Davis
Luis Carvajal-Carmona
Paul Lott
Ana Estada-Florez
Guadalupe Polanco-Echeverry
City of Hope
Susan Neuhausen
Jeffrey Weitzel
USC
Chris Haiman
Stanford
Esther John
INSP Mexico
Gabriela Torres-Mejia
Kaiser
Larry Kushi
University of Tolima, Colombia
Magdalena Echeverry
Mabel E. Bohorquez
IMSS, Mexico
Javier Torres Juan
Carlos Martínez-Chéquer
INEN, Peru
Tatianna Vidaurre
Sandro CasavilcaZambrano