University of Southern California Dissertations and Theses

The impact of global and local Polynesian genetic ancestry on complex traits in Native Hawaiians

(USC Thesis Other)

The impact of global and local Polynesian genetic ancestry on complex traits in Native Hawaiians

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content 1
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians

The Impact of Global and Local Polynesian Genetic Ancestry
on Complex Traits in Native Hawaiians

By Hanxiao Sun

___________________________________________________________________________

A Thesis Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(BIOSTATISTICS)

May 2019

Copyright 2019 Hanxiao Sun

2
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Table of Contents
Acknowledgements ..................................................................................................................3
List of Tables/ Figures ..............................................................................................................4
Abstract ....................................................................................................................................5
Introduction ..............................................................................................................................5
Methods ....................................................................................................................................7
Results ....................................................................................................................................11
Discussion ..............................................................................................................................30
References ..............................................................................................................................31
Abbreviation ...........................................................................................................................33

3
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Acknowledgements

I would like to convey my deepest appreciation to my master thesis committee: Dr. Charleston
Chiang, Dr. Kristine Monroe and Dr. Xuejuan Jiang. Specially, I would like to express my deep
gratitude to Prof. Charleston Chiang, my thesis supervisor, for his patient guidance, enthusiastic
encouragement and constructive critiques of this work.

Besides, I wish to thank Dr. Meng Lin for her assistance with determining the unrelated NH
in the section of identifying NH reference panel. Additionally, I would like to thank Dr. Minhui
Chen and PhD candidate Soyoung Jeon for their help when I was preparing my thesis in the center
of Genetic Epidemiology.

Finally, I would express my gratitude to those who help me and support me throughout my
graduate study.

4
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
List of Tables/ Figures

1. Fig 1. The distribution of imputation score .............................................................................. 8
2. Table 1. Summary of date preprocessing and covariates involved in PAGE study ................ 10
3. Fig 2. Output of ADMIXTURE in unsupervised mode in k=4 run ........................................ 12
4. Fig 3. Output of ADMIXTURE in supervised mode in k=4 turn ........................................... 12
5. Table 2. Summary of statistics of traits and covariates involved ........................................... 12
6. Table 3. Summary of regressions of traits of interests ........................................................... 14
7. Table 4. Statistics of coefficients in the regression of BMI .................................................... 15
8. Table 5. Summary of regressions of BMI among subgroups ................................................. 15
9. Table 6. Statistics of coefficients in the regression of BMI among subgroups ...................... 16
10. Table 7. Statistics of coefficients in the regression of glucose ............................................... 17
11. Table 8. Statistics of coefficients in the regression of insulin ................................................ 18
12. Table 9. Statistics of coefficients in the regression of HDL ................................................... 19
13. Table 10. Statistics of coefficients in the regression of LDL ................................................. 19
14. Table 11. Statistics of coefficients in the regression of triglycerides ..................................... 20
15. Table 12. Statistics of coefficients in the regression of total cholesterol ................................ 20
16. Table 13. Statistics of coefficients in the regression of type 2 diabetes ................................. 21
17. Table 14. Statistics of coefficients in the regression of obesity .............................................. 22
18. Table 15. Summary of regressions of obesity among subgroups ........................................... 22
19. Table 16. Statistics of coefficients in the regression of obesity among subgroups................. 22
20. Fig 4. QQ plot on chromosome 6 for type 2 diabetes ............................................................ 23
21. Fig 5. Admixture mapping for type 2 diabetes on chromosome 6 ......................................... 24
22. Fig 6. Admixture mapping for triglycerides on chromosome 22 ........................................... 25
23. Table 17. Summary of statistics of significant region ............................................................ 26
24. Table 18. Summary of possible influencing SNPs from GWAScatalog ................................. 26
25. Table 19. Summary of allele frequency difference between NH and EUR, EAS .................. 27
26. Fig 7. Admixture mapping on chromosome 6 before and after adjusted by rs9362054 ......... 28
27. Fig 8. Admixture mapping before and after adjusted by all potential confounding SNPs ..... 29

5
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Abstract
Epidemiologic studies have shown that Native Hawaiians exhibit an excess risk for obesity,
Type 2 diabetes, cardiovascular diseases and several common cancers compared to European or
Asian Americans living in the Hawaiian islands today. However, like many non-European or
indigenous populations, Native Hawaiians are understudied, particularly from a genetic standpoint.
To fill this gap, we took advantage of the availability of genome-wide array data on nearly 4,000
Native Hawaiians from the Multiethnic Cohort to study the risk related to their genetic ancestry for
these complex diseases and the associated quantitative traits. Specifically, we first identified a
subset of Native Hawaiians with extremely high proportion of Polynesian ancestry to construct an
internal reference panel to help us estimate the genomic proportion of Polynesian ancestry (“global
ancestry”), as well as Polynesian ancestry along each chromosome (“local ancestry”), in the
remaining 3,428 unrelated individuals. We then tested the associations between proportion of global
Polynesian ancestry and both binary (obesity, type 2 diabetes, prostate cancer, breast cancer) and
quantitative (BMI, glucose, insulin, HDL, LDL, triglyceride, total cholesterol) traits. After
adjusting for available covariates on socioeconomic status, we found that per 10% increase in
Polynesian genetic ancestry, there is a 17% increase in the odds of being obese (P < 2 ×10
-16
), a
0.076 s.d. increase in BMI (P < 2 ×10
-16
) and a 0.046 s.d. increase in HDL (P = 8.2 ×10
-6
). We also
found weaker nominal associations between proportion of Polynesian genomic ancestry and insulin
and total cholesterol. To examine regions along the genome that might harbor Polynesian-specific
variants associated with risk of disease or biomarkers, we conducted admixture mapping on these
traits and identified genomic regions at approximately 57.4Mb-70.3Mb on chromosome 6 and
22.4Mb-29.0Mb on chromosome 22 to be putatively associated with Type 2 diabetes and
triglycerides, respectively; neither of the admixture signals are explained by known GWAS variants
reported in these regions, suggesting potentially new regions of interest specific to Native
Hawaiians.
Key words:
Native Hawaiian; genome-wide association study; admixture mapping; BMI; type 2 diabetes; HDL;
insulin; triglycerides; minor populations

Introduction
Native Hawaiians are the second fastest growing ethnic group in the U.S., growing 40% from
the 2000 to 2010 U.S. census. Moreover, Native Hawaiians display alarming rates of obesity,
coronary heart disease, diabetes, cardiovascular disease, cancers, and other related chronic health
conditions
1-5
. Epidemiological studies have shown that 49% of adult Native Hawaiians are obese,
compared to 21% of European Americans
2
, and still had > 2x higher odds of being obese after
adjusting for socioeconomic status
5
. In addition, Native Hawaiians are ~2-3 times more likely to
develop type 2 diabetes and ~30% more likely to develop breast cancer than their European
American counterparts, even after adjusting for common modifiable risk factors such as BMI and
socioeconomic covariates
3,4
. Similarly, Native Hawaiians are ~1.7 times more likely to develop
cardiovascular disease than European Americans
6
, and cardiometabolic risk factors such as
hypertension has been shown to be correlated with genealogical estimates of proportion of Native
6
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Hawaiian ancestry
7
. Taken together, these observations suggest that there may be systematic
differences in number and frequency of risk alleles at known genetic loci that at least partly explain
the differences between Native Hawaiians and European ancestry populations. Yet, despite
awareness and efforts to include more non-European populations in genomic studies, indigenous
populations such as Native Hawaiians remain understudied
8,9
.
To date, the best genetic and epidemiological resource available to study the Native Hawaiians,
in terms of both the scale of the cohort and the quality of the genetic data, is the Native Hawaiian
cohort in the Multiethnic Cohort (MEC). The MEC is a prospective epidemiological cohort
of >215,000 individuals spanning five ethnicities, including biospecimen samples on > 70,000
individuals: >10,000 African Americans, >13,000 European Americans, >23,000 Japanese
Americans, > 16,000 Latino Americans, and > 5,300 Native Hawaiians. Over 3,900 Native
Hawaiian samples were genotyped on the MEGA array
18
as part of the Population Architecture
using Genomics and Epidemiology (PAGE) study, which is a consortium aggregating nearly 50,000
non-European individuals for a genome-wide association study (GWAS) aimed to discover novel
loci associated with complex traits and diseases that would have been missed in European cohorts
19
.
Yet even in an effort to specifically study non-European populations such as in PAGE, Native
Hawaiians are buried as a small contributing component of a much larger consortium driven by
other ethnic populations such as Latinos or Asians. Native Hawaiian specific effects would likely
be overshadowed due to their relatively small sample size, unless examined in stratified analyses.
Historically, Native Hawaiians experienced a unique population history. The ancient
Polynesians originated from East Asia, likely through Taiwan, and successively settled major
landmasses of the Philippines, Near Oceania and Remote Oceania. They reached the Hawaii
archipelagos approximately 1200-2000 years ago
10-12
and remained isolated there until 1778, when
they encountered Western explorers and novel infectious agents that decimated the Native
Hawaiian population before they rebounded over the last couple of centuries. It is also only during
the 18
th
and 19
th
centuries that the Native Hawaiians became admixed with European and East
Asian immigrants to the island, as well as with African American immigrants in the more recent
time. Consistent with this history, it has been estimated that on average, the Native Hawaiians
derive 78% of their ancestry from Polynesian origin, 11.5% from Europeans, and 7.8% from Asians,
with very small amount from Africans. The 2010 U.S. census data suggest that currently there are
at most only approximately 1.2 million individuals in the U.S. who derive some proportion of their
ancestry from Native Hawaiians, accounting for about 0.4% of the U.S. population. In contrast,
recent GWAS of European populations are exceeding 1 million research participants (e.g. see ref
11
). It is thus obvious that the major obstacle to study the genetic architecture of complex traits in
Native Hawaiians or any other indigenous populations is the inability to recruit large sample sizes
to overcome the multiple testing burden in a standard GWAS design. Alternative approaches will
be needed.
One alternative approach to identify novel loci associated with complex traits and diseases is
through admixture mapping. Admixture mapping is designed to locate genetic loci with excess
ancestry associated to a phenotype of interest, based on the assumption that causal variants leading
to increased risk or trait values occur more frequently on chromosomal segments inherited from
7
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
the ancestral population that has higher disease risk or larger average trait values
14
. Critically,
because admixture mapping aims to associate locus-specific ancestry to disease risk, and each
ancestral haplotype blocks are expansive in recently admixed populations, the total number of
independent tests performed is orders of magnitudes smaller than a typical GWAS. In Latinos, the
effective number of genome-wide tests has been estimated to be 1041, leading to a genome-wide
threshold of 4.8x10
-5
(as opposed to 5x10
-8
in GWAS)
15
. Therefore, at the cost of variant-level
resolution, admixture mapping is more powered to identify loci where the causal alleles show large
frequency discrepancies in the ancestral populations or where multiple causal alleles aggregate in
a region. This technique has previously been used in African-American and Latinos to identify
novel genomic regions associated with phenotypes such as asthma, blood cell traits, and breast and
prostate cancer (reviewed in ref
16,17,25
).
Therefore, to begin filling the gap missing in the genetic understanding of disease risks in
Native Hawaiians, in this study, we tested the association of both global (genomic) and local
Polynesian ancestry to complex traits and diseases in Native Hawaiians from MEC previously
genotyped as part of the PAGE consortium. We focused on obesity, type 2 diabetes and breast cancer,
as well as related quantitative or cardiometabolic phenotypes that have reported excess risk in
Native Hawaiians. To do so, we first conducted genetic ancestry analysis to identify an internal
Native Hawaiian reference panel (those that derived over 90% of their ancestry from Polynesian
ancestors) to use as imputation and ancestry references, since there is no appropriate representative
currently available. We find that obesity (and BMI as quantitative trait) and HDL are strongly
associated with proportion of Polynesian ancestry among Native Hawaiians, even after correcting
for available measures of socioeconomic status. We also found regions on chromosome 6 and on
chromosome 22 that appear to be implicated in admixture mapping to be associated with type 2
diabetes and triglyceride levels. Signals from these regions could not be explained by known
GWAS variants in proximity.

Methods
Data cohort description. In this study, we used genetic data from Native Hawaiian individuals from
the Multiethnic Cohort (MEC) project
18
. Of the Native Hawaiians (NH), approximately 5,300 have
biospecimen samples, and approximately 3,900 were genotyped as part of the PAGE consortium
19

using the MEGA array. Genetic variants from 3940 Hawaiian individuals were prepared for later
steps, as well as corresponding BMI, glucose, insulin, HDL, LDL, triglycerides (TG), total
cholesterol (TC), type 2 diabetes, prostate cancer and breast cancer status, and other relevant
indicators.
For reference panels of other global populations, we used individuals sequenced in the 1000
Genomes Project. All European except Finnish, all East Asian, Yoruba and Luhya were extracted
from 1000 Genomes Project, contributing to 404 European ancestry, 504 Asian ancestry and 207
African ancestry, respectively.

Imputation. To improve the span and coverage of genetic variation across the genome, genetic data
from NH were imputed using the Sanger Imputation Server
20.
Pre-imputation quality controls were
8
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
done as part of the PAGE consortium, resulting in 156,791,234 SNPs and 3940 NH individuals for
imputation. Pre-phasing and imputation were completed by EAGLE+PBWT with the reference
panel from 1000 Genomes Phase 3
21,22
.
A summary of the number of SNPs based on their imputation INFO score at the score interval
of 0.1 is shown in Figure 1. The imputed data was first filtered according to its imputation INFO
score, which less than 0.4 was discarded. Duplicated SNPs and indels were then dropped. To obtain
a population of unrelated individuals, KING
23
was used to identify the kinship between individuals
and we removed individuals such that there is no first-degree relationship in our cohort. Alleles
with minor allele frequency under 0.05, were discarded from both the imputed data and 1000
Genomes. In total, we retained 5,617,838 SNPs in 3482 NH for analysis in this thesis.

Fig 1. The distribution of imputation score

Identifying NH reference panel. Previous investigation of Native Hawaiian ancestry suggested
that Native Hawaiians are admixed with recent European and East Asian ancestry
11
. In order to call
local ancestry, we needed a good reference panel for the Polynesian component of the Native
Hawaiian ancestry. Such an external reference panel does not exist. Therefore, we identified among
NH individuals those with the largest amount of global Polynesian ancestry using ADMIXTURE
to construct an internal reference panel of Polynesian ancestry. Using the combined NH and 1000
Genomes data, we first pruned SNPs with r
2
> 0.1 (using window sizes of 50 SNPs with steps of 5
SNPs across the genome) by PLINK. We then ran ADMIXTURE in unsupervised mode on the
prepared merged dataset. We found that at k=4, the NH individuals exhibited known component of
ancestry from European, East Asian and African, as well as a component of ancestry that is unique
to the NH, presumed to be Polynesian. We then identified the individuals with the Polynesian
9
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
component of ancestry over 0.9 and defined them as the reference individuals for Polynesian
ancestry.

Global and local ancestry calling. To obtain global ancestry estimates for each NH individual, we
combined the imputed NH reference individuals with 1000 Genomes Europeans, East Asians, and
Africans, and pruned the SNPs in the same manner as above. We then ran ADMIXTURE now in
supervised mode using the newly constructed reference panel. We then used the estimated
Polynesian ancestry to represent the proportion of global ancestry derived from Polynesian ancestry
for each of the non-reference NH individuals. To call local ancestry segments for each NH
individual, we used the same reference panel, without pruning of SNPs by r
2
, and used RFMix
24
to
generate local ancestry call.

Association between binary and quantitative traits and Polynesian ancestry. Because Native
Hawaiians are known to have elevated risk in obesity, type 2 diabetes, breast cancer, and
cardiovascular disease, it could represent the propensity of developing certain diseases or having
certain traits is correlated with the proportion of NH global ancestry. Therefore, we examined the
association of global Polynesian ancestry with binary and quantitative traits related to these diseases.
Specifically, we examined the following quantitative traits/biomarkers: BMI, glucose, insulin, HDL,
LDL, triglyceride, total cholesterol. We also examined the following binary disease traits: prostate
cancer, breast cancer, type 2 diabetes and obesity.
We tested the association with quantitative traits using linear regression, after adjusting for
relevant covariates. As quantitative traits sometimes violate the assumption of normality in
dependent variable and heteroscedasticity in linear regression, quantitative traits were first adjusted
for common, trait-specific covariates and the residuals were inversed normalized prior to applying
linear regression. Categorical traits were examined by logistic regression, also adjusting for
covariates. That is:

For quantitative traits,
𝑡𝑟𝑎𝑖𝑡𝑠 = 𝛼 + 𝛽 1
· 𝑐𝑜𝑣 1
+ 𝛽 2
· 𝑐𝑜𝑣 2
+ ⋯ + 𝛽 𝑛 · 𝑐𝑜𝑣 𝑛
𝑖𝑛𝑣𝑒𝑟𝑠𝑒𝑑 𝑛𝑜𝑟𝑚𝑎𝑙𝑖𝑧𝑒𝑑 𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙𝑠 = 𝛼 + 𝛽 · 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦

For categorical traits,
𝑙𝑜𝑔𝑖𝑡 (𝑡𝑟𝑎𝑖𝑡𝑠 ) = 𝛼 + 𝛽 · 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 + 𝛽 1
· 𝑐𝑜𝑣 1
+ 𝛽 2
· 𝑐𝑜𝑣 2
+ ⋯ + 𝛽 𝑛 · 𝑐𝑜𝑣 𝑛

Since the association of global ancestry and disease traits/biomarkers can be confounded by
social factors, covariates related to sociological economic status (SES) that were available in MEC
were added into the models. We first tested a full model
19
that does not contain the global ancestry
estimate but included all of the covariates as well as pairwise interaction terms in a multivariate
regression model (Table 1). Covariates that were not significantly associated with the outcome were
dropped to obtain a reduced model. We then tested the association with global ancestry.

10
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Table 1. Summary of data preprocessing and covariates involved in PAGE study
19

Trait Data management
Glucose
Remove non-fasting glucose and glucose from pregnant women and patients with type
2 diabetes and that greater than 7 mmol/L. The rest of glucose is statistically associated
with age, BMI, gender, education, smoking status and age×gender.
Insulin
Remove non-fasting insulin and insulin from pregnant women and patients with type 2
diabetes and those of whom glucose is greater than 7 mmol/L. Then log-transformed.
The log-transformed insulin is statistically associated with age, BMI, gender, education,
smoking status and age×gender.
HDL
Drop 5.9 from HDL, if fibrates related medication was received. Remove HDL from
pregnant women and those who hadn’t fasted over 8 hours before blood draw. The
modified HDL is statistically associated with age at taking lipid medications, gender and
education.
LDL
Add 40.1 to LDL, if fibrates related medication was received. Remove LDL from
pregnant women and those who hadn’t fasted over 8 hours before blood draw and that
greater than 400 mg/dL. The modified LDL is statistically associated with age at taking
lipid medications, gender and education.
TG
Add 57.1 to TG, if fibrates related medication was received. Remove TG from pregnant
women and those who hadn’t fasted over 8 hours before blood draw and that greater
than 3000 mg/dL. The modified TG is statistically associated with age at taking lipid
medications, gender and education.
TC
Add 46.1 to TC, if fibrates related medication was received. Remove TC from pregnant
women and those who hadn’t fasted over 8 hours before blood draw. The modified TC
is statistically associated with age at taking lipid medications, gender and education.
Type 2
diabetes
Remove diabetes cases from pregnant women and those who have diagnosed as type 1
diabetes, cases with age under 20 and controls with glucose over 7 mmol/L. type 2
diabetes is statistically associated with age, gender, BMI and education

Admixture mapping. To identify local genomic segments in which the ancestry is associated with
trait of interest, we conducted admixture mapping. Briefly, case-control (CC) and Case-only (CO)
models were the two main approaches. Assume Y is the vector of binary indicator variable of
disease status, 𝑄 𝑖𝑙𝑘 , 𝐿 𝑖𝑙𝑘 refer to the global ancestry and the local ancestry at locus l from ancestry
k for the i
th
individual, respectively. As 𝐴 𝑖𝑙𝑘 = 𝑄 𝑖𝑙𝑘 − 𝐿 𝑖𝑙𝑘 and A lk is the vector of observed
deviations for N individuals, CC and CO model could be expressed as
25

CC: 𝑨 ℓ𝑘 = 𝛼 + 𝛽 𝑐𝑐
𝒀 + 𝜖
CO: 𝑨 ℓ𝑘 = 𝛽 𝑐𝑜
𝒀 + 𝜖

Both CC and CO model are designed to detect whether there exists a significant deviation
between local ancestry and global ancestry. As CO model assumes that controls only contribute to
noise, i.e. the average global ancestry is equivalent to the average local ancestry at certain loci in
11
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
the unaffected population, E (Q
̅
ℓ𝑘 𝑈𝑛𝑎𝑓𝑓 − L
̅
ℓ𝑘 𝑈𝑛𝑎𝑓𝑓 ) = 0. The difference between the two is that
CC model also compares the distinctions between cases and controls, whereas CO model only tests
the significance of the deviation in cases. So CO model is usually more powerful on account of less
error source introduced, but it is also exposed to higher false-positive rate if the assumption is
violated. On the other hand, CC model is more generally robust, especially in scenarios where the
assumption is violated, but with less detecting power
25
.
We also applied regression methods in admixture mapping, which allows additional covariates
being included, such as other components of ancestry in a multi-ancestry admixed population to
reduce the impact from other ancestries. By controlling global ancestry from other ancestries, for
quantitative traits, we test:

𝑖𝑛𝑣𝑒𝑟𝑠𝑒 𝑛𝑜𝑟𝑚𝑎𝑙𝑖𝑧 𝑒𝑑 𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙𝑠 = 𝛼 + 𝛽 · 𝑙𝑜𝑐𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝑁𝐻
+ 𝛽 1
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝑈𝑅 +
𝛽 2
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝐴𝑆 + 𝛽 3
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐴𝐹𝑅

While for categorical traits, we test:

𝑙𝑜𝑔𝑖𝑡 (𝑡𝑟𝑎𝑖𝑡𝑠 ) = 𝛼 + 𝛽 · 𝑙 𝑜𝑐𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝑁𝐻
+ 𝛽 1
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝑈𝑅 + 𝛽 2
·
𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝐴𝑆 + 𝛽 3
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐴𝐹𝑅

For regions we detected to be significantly associated using admixture mapping, we checked
whether a previously known variant from GWAS for the same trait could drive this signal.
Therefore we identified known variants from GWAS catalog
26
located within ±1 MB of the top
admixture mapping region. We then included genotypes of these variants in the regression model
as additional covariates. CC, CO and regression models were modified accordingly:

CC: 𝑨 ℓ𝑘 = 𝛼 + 𝛽 𝑐𝑐
𝒀 + 𝑆𝑁𝑃 𝑖 / ∑ 𝑆𝑁𝑃 𝑖 𝑚 𝑖 =1
+ 𝜖
CO: 𝑨 ℓ𝑘 = 𝛽 𝑐𝑜
𝒀 + 𝑆𝑁𝑃 𝑖 / ∑ 𝑆𝑁𝑃 𝑖 𝑚 𝑖 =1
+ 𝜖

For quantitative traits, regression model would be:

𝑟𝑎𝑛𝑘𝑒𝑑 𝑛𝑜𝑟𝑚𝑎𝑙𝑖𝑧𝑒𝑑 𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙 = 𝛼 + 𝛽 · 𝑙𝑜𝑐𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 + 𝛽 1
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝑈𝑅 + 𝛽 2
·
𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝐴𝑆 + 𝛽 3
· 𝑔𝑙𝑜𝑏𝑎 𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐴𝐹𝑅 + 𝑆𝑁𝑃 𝑖 / ∑ 𝑆𝑁𝑃 𝑖 𝑚 𝑖 =1

For categorical traits, it would be:

𝑙𝑜𝑔𝑖𝑡 (𝑡𝑟𝑎𝑖𝑡𝑠 ) = 𝛼 + 𝛽 · 𝑙𝑜𝑐𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 + 𝛽 1
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝑈𝑅 + 𝛽 2
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐸𝐴𝑆 +
𝛽 3
· 𝑔𝑙𝑜𝑏𝑎𝑙 𝑎𝑛𝑐𝑒𝑠𝑡𝑟𝑦 𝐴𝐹𝑅 + 𝑆𝑁𝑃 𝑖 / ∑ 𝑆𝑁𝑃 𝑖 𝑚 𝑖 =1

Results
Identifying NH reference panel. As part of the PAGE consortium, Japanese, African Americans,
12
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
and Native Hawaiians from MEC were uniformly genotyped on the MEGA array. We merged these
MEC individuals with 1000 Genomes populations and performed unsupervised ADMIXTURE
analysis. In the k=4 run, the blue, green, purple components represented European, East Asian, and
African components of ancestry, respectively. The red component, found predominantly in Native
Hawaiians, thus likely represent the Polynesian component of ancestry of NH (Figure 2). There
were 177 individuals with the new component over 0.9 in their global ancestry, and they were
considered as NH reference panel.

Fig 2. Output of ADMIXTURE in unsupervised mode in k=4 run

Global ancestry calling. ADMIXTURE was performed in supervised mode on 214,555 SNPs after
pruning, and global ancestry of 3428 unrelated, non-reference NH individuals were called. The
average global ancestry of NH is composed of 28.31% EUR, 26.82% EAS, 43.70% NH and 1.17%
AFR (Figure 3).

Fig 3. Output of ADMIXTURE in supervised mode in k=4 turn

Association between traits and global ancestry. Descriptive statistics of relevant traits and
covariates are presented in Table 2.

Table 2. Summary of statistics of traits and covariates involved
Quantitative traits
Traits Mean Std.
BMI/m
2
·kg
-1
29.04 5.95
Glucose/mmol ·L
-1
4.954 1.238
Insulin/pmol ·L
-1
8.331 7.431
HDL/mg ·dL
-1
42.04 15.29
LDL/mg ·dL
-1
128.0 36.89
TG/mg ·dL
-1
121.8 79.84
13
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
TC/mg ·dL
-1
193.9 39.68

Categorical traits
Traits Level Number
Prostate cancer
0 = undiagnosed 1529
1 = diagnosed 69
Breast cancer
0 = undiagnosed 1855
1 = diagnosed 112
Type 2 diabetes
0 = undiagnosed 1849
1 = diagnosed 1382
Obesity
0 = undiagnosed 2686
1 = diagnosed 877

Quantitative covariates
Covariate Mean Std.
Age 54.27 7.64

Categorical covariates
Covariate Level Number
Sex
1 = Male 1598
2 = Female 1967
Education
1: ≤8
th
grade 248
2: high school 1265
3: some college or vocational school 1203
4: college graduates 825
Cigs_per_day
1: 0 cigarettes smoked per day 1532
2: ≤5 cigarettes smoked per day 227
3: 6-10 cigarettes smoked per day 451
4: 11-20 cigarettes smoked per day 764
5: 21-30 cigarettes smoked per day 364
6: ≥31cigarettes smoked per day 174

For each trait we tested, we first evaluated a full model that does not include the global ancestry
estimate, but contains all relevant covariates and their interactions in a multivariate regression
model. We then retained a subset of the covariates that are significantly associated with the trait in
a reduced model (eg. Equation 1 in Table 3). For quantitative traits, we adjusted for these covariates
and generated inverse normalized residuals to test for association with global Polynesian ancestry
(eg. Equation 2 in Table 3); for categorical traits we included the global ancestry estimate along
with the reduced model of the covariates. In summary, we found that global Polynesian ancestry is
strongly associated with BMI, obesity, and HDL, and nominally associated with insulin, type 2
14
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
diabetes, and total cholesterol. For BMI, type 2 diabetes and obesity, increased proportion of global
Polynesian ancestry is associated with an increased risk, consistent with excess risks of these
diseases in Native Hawaiians. Increased Polynesian ancestry is also associated with decreased HDL
level, consistent with an excess risk of cardiovascular diseases among Native Hawaiians. On the
other hand, increased Polynesian ancestry is associated with increased insulin level, and decreased
total cholesterol, which are inconsistent with the excess risk of diabetes and cardiovascular disease,
suggesting that these nominally significant signals should be further investigated in larger samples.
Note that no significant multivariate models were generated for prostate cancer and breast cancer
in the first step, and thus the full model was tested. We were also likely to severely underpower the
testing for association with breast cancer and prostate cancer, due to the small number of cases.
Each association will be elaborated next.

Table 3. Summary of regressions of traits of interests
Quantitative traits
Traits Equation 1: regression with covariates
Equation 2:
regression with
global ancestry
P of global
ancestry
BMI
(Overall)
bmi=31.9259-0.0387×age+12.4894×t2d-
1.3226×sex-1.0167×edu2-
0.1984×age:t2d+1.5449 ×t2d:sex
resid=-
0.3525+0.7643×NH
<2×10
-16

Glucose
gluc=4.0513+0.0052×age+0.0182×bmi-
0.0754×sex
resid=0.0712-
0.1660×NH
0.1680
Insulin
ln(insulin)=1.0681-
0.0016×age+0.0529×bmi-
0.4090×sex+0.0092 ×age:sex
resid=-
0.1307+0.3043×NH
0.0111
HDL
hdl=23.1582+0.1258×age_at_lipid+7.1324
×sex
resid=0.2011-
0.4566×NH
8.21×10
-6

LDL
ldl=159.6125-
0.6538×age_at_lipid+7.5304×sex
resid=0.0447-
0.1014×NH
0.325
TG
tg=160.8338-0.5182×age_at_lipid-
2.8280×sex
resid=-
0.0337+0.0765×NH
0.4560
TC
tc=211.7619-
0.5816×age_at_lipid+13.7749×sex
resid=0.1248-
0.2834×NH
0.0036

Categorical traits
15
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Traits Equation 1: regression with covariates
P of global
ancestry
Type 2
diabetes
Logit(t2d)=-7.0967+0.4000 ×NH+0.0619 ×age+0.1269 ×bmi-
{
0.3643 × 𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 , 𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 = 2
0.5855 × 𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 , 𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 = 3
0.6411 × 𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 , 𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 = 4

0.0178
Obesity
(Overall)
Logit(obesity)=-1.6717+1.5946 ×NH-0.0151 ×age+4.1249 ×
t2d+0.3846 ×cig2-0.0526 ×age:t2d
<2×10
-16

BMI (Body Mass Index). BMI is statistically correlated with age, gender, education status and type
2 diabetes status, age ×type 2 diabetes status and gender ×type 2 diabetes status (Table 4). Given
that age, gender and type 2 diabetes status significantly interact with each other, we also conducted
stratified analysis among different subgroups (Table 5, Table 6).
Generally, BMI will increase ~0.0764 unit of standard deviation with per 10% increase in
global Polynesian ancestry. In stratified analysis, BMI is associated with global Polynesian ancestry
in all strata; BMI will increase 0.073, 0.094, 0.061, 0.091 unit per 10% increase in Polynesian
ancestry for male type 2 diabetes, male non-type 2 diabetes, female type 2 diabetes and female non-
type 2 diabetes, respectively.

Table 4. Statistics of coefficients in the regression of BMI
Equation 1: linear regression between BMI and covariates
variables estimate std. error t p R
2

intercept 31.9259 1.0488 30.441 <2× 10
-16

0.1452
age -0.0387 0.0177 -2.192 0.0285
t2d_stat_new 12.4894 1.621 7.705 1.74× 10
-14

sex -1.3226 0.2617 -5.054 4.57× 10
-7

edu2
1
-1.0167 0.2338 -4.348 1.42× 10
-5

age:t2d_stat_new -0.1984 0.0267 -7.435 1.34× 10
-13

t2d_stat_new:sex 1.5449 0.3994 3.869 1.12× 10
-4

Equation 2: linear regression between rank-based inversed residual and global ancestry
intercept -0.3525 0.0377 -9.359 <2× 10
-16

0.03354
NH_prop 0.7643 0.0725 10.546 <2× 10
-16

1
: edu2 was created based on the original education levels, which is a binary variable. Level 0 in edu2
categorized level 1,2,3 in the origin education status together, while level 1 represented the original
education level 4.

Table 5. Summary of regressions of BMI among subgroups
Traits Equation 1: regression with Equation 2: P of global
16
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
covariates regression with
global ancestry
ancestry
BMI
Male &
t2d
bmi=42.4760-0.2009×age-0.4685×edu2
resid=-
0.3615+0.7323×NH
2.77×10
-6

Male &
non-t2d
bmi=30.9073-0.0452×age-0.8864×edu2
resid=-
0.4100+0.9376×NH
7.79×10
-11

Female &
t2d
bmi=46.3793-0.2638×age-1.2313×edu2
resid=0.3080+0.6114×
NH
1.10×10
-4

Female
&non-t2d
bmi=29.1497-0.0347×age-1.3531×edu2
resid=-
0.3932+0.9147×NH
1.35×10
-11

Table 6. Statistics of coefficients in the regression of BMI among subgroups
Male & type 2 diabetes
Equation 1: linear regression between BMI and covariates
Variables estimate std. error t p R
2

Intercept 42.476 1.6025 26.506 <2×10
-16

0.0764

Age -0.2009 0.0281 -7.155 2.35×10
-12

edu2 -0.4685 0.4866 -0.963 0.336
Equation 2: linear regression between rank-based inversed residual and global ancestry
Intercept -0.3615 0.0859 -4.209 2.94×10
-5

0.0345
NH_prop 0.7323 0.1548 4.731 2.77×10
-6

Male & non-type 2 diabetes
Equation 1: linear regression between BMI and covariates
Variables estimate std. error t p R
2

Intercept 30.9073 1.3312 23.218 <2×10
-16

0.0121 Age -0.0452 0.0244 -1.854 0.0641
edu2 -0.8864 0.3864 -2.294 0.0221
Equation 2: linear regression between rank-based inversed residual and global ancestry
Intercept -0.41 0.0712 -5.76 1.21×10
-8

0.0524
NH_prop 0.9369 0.1421 6.595 7.79×10
-11

Female & type 2 diabetes
Equation 1: linear regression between BMI and covariates
Variables estimate std. error t p R
2

Intercept 46.3793 1.8077 25.657 <2×10
-16

0.0862 Age -0.2638 0.0319 -8.282 5.75×10
-16

edu2 -1.2313 0.6639 -1.854 0.0641
17
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Equation 2: linear regression between rank-based inversed residual and global ancestry
Intercept -0.308 0.0872 -3.531 4.40×10
-4

0.0202
NH_prop 0.6114 0.1573 3.888 1.10×10
-4

Female & non-type 2 diabetes
Equation 1: linear regression between BMI and covariates
Variables estimate std. error t p R
2

Intercept 29.1497 1.2258 23.78 <2×10
-16

0.0127 Age -0.0347 0.0227 -1.526 0.1274
edu2 -1.3531 0.3914 -3.457 5.68×10
-4

Equation 2: linear regression between rank-based inversed residual and global ancestry
Intercept -0.3932 0.0649 -6.057 1.39×10
-9

0.0426
NH_prop 0.9147 0.1338 6.839 1.35×10
-11

Glucose. We found that glucose is statistically associated with age, BMI and gender. After adjusting
for these covariates, we did not find any significant association between global ancestry and glucose
levels (Table 7).

Table 7. Statistics of coefficients in the regression of glucose
Full model: linear regression between glucose and covariates
Variables estimate std. error t p R
2

Intercept 3.8596 0.4944 7.806 1.24×10
-14

0.1452
Age 0.0088 0.0084 1.045 0.2963
Bmi 0.0177 0.0038 4.668 3.37×10
-6

Sex 0.053 0.2785 0.190 0.8429
education
2 vs 1 0.019 0.0974 0.195 0.8455
3 vs 1 -0.0113 0.0974 -0.117 0.9072
4 vs 1 -0.0366 0.1001 -0.366 0.7143
cigs_per_day
5 vs 0 0.1123 0.0795 1.411 0.1584
8 vs 0 0.0237 0.0618 0.383 0.7017
15.5 vs 0 0.0973 0.0506 1.921 0.0549
25.5 vs 0 -0.0248 0.0684 -0.363 0.7166
31 vs 0 0.1468 0.1044 1.406 0.1599
age:sex -0.0025 0.0052 -0.485 0.6278

Reduced model: linear regression between glucose and covariates
Variables estimate std. error t p R
2

Intercept 4.0513 0.1866 21.716 <2×10
-16

0.0257 Age 0.0052 0.0025 2.050 0.0406
Bmi 0.0182 0.0037 4.869 1.26×10
-6

18
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Sex -0.0754 0.038 -1.985 0.0474
Equation 2: linear regression between rank-based inversed residual and global ancestry
Intercept 0.0712 0.0586 1.215 0.225
0.0015
NH_prop -0.1660 0.1202 -1.381 0.168

Insulin. We found that log-insulin is only correlated with age, gender, BMI and age ×gender in this
case, and there is a marginally significant association between global ancestry and log-insulin level.
Generally, ln(insulin) will increase ~0.0304 unit of standard deviation with per 10% increase of
global ancestry from NH (Table 8).

Table 8. Statistics of coefficients in the regression of insulin
Full model: linear regression between ln(insulin) and covariates
Variables estimate std. error t p R
2

Intercept 1.4243 0.4376 3.255 0.0012
0.1759
Age -0.0195 0.0075 -2.607 0.0092
Bmi 0.0523 0.0034 15.451 <2×10
-16

Sex -0.5099 0.2475 -2.060 0.0394
education
2 vs 1 -0.106 0.0869 -1.22 0.2228
3 vs 1 -0.1209 0.0869 -1.391 0.1644
4 vs 1 -0.1746 0.0892 -1.957 0.0506
cigs_per_day
5 vs 0 0.1578 0.0714 2.210 0.0273
8 vs 0 0.0653 0.0546 1.196 0.232
15.5 vs 0 -0.0043 0.0452 -0.096 0.9233
25.5 vs 0 -0.0338 0.0610 -0.553 0.5800
31 vs 0 0.1279 0.0922 1.387 0.1656
age:sex 0.0108 0.0046 2.346 0.0191

Reduced model: linear regression between ln(insulin) and covariates
Variables estimate std. error t p R
2

Intercept 1.0681 0.4079 2.619 0.0089
0.1666
Age -0.0016 0.0073 -2.124 0.0338
Bmi 0.0529 0.0033 15.858 <2×10
-16

Sex -0.409 0.2444 -1.674 0.0944
age:sex 0.0092 0.0045 2.023 0.0433
Equation 2: linear regression between rank-based inversed residual and global ancestry
Intercept -0.1307 0.0584 -2.238 0.0254
0.0050
NH_prop 0.3043 0.1197 2.543 0.0111

HDL. We found that HDL is only correlated with age at taking lipid medications and gender in this
case, and there is a statistically significant association between global ancestry and HDL level.
19
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Generally, HDL will decrease ~0.0457 unit of standard deviation with per 10% increase of global
ancestry from NH (Table 9).

Table 9. Statistics of coefficients in the regression of HDL
Full model: linear regression between HDL and covariates
variables estimate std. error t p R
2

intercept 19.4778 3.7131 5.246 1.75×10
-7

0.0599
age_at_lipids 0.1464 0.0458 3.197 0.0014
sex 7.2096 0.7251 9.943 <2×10
-16

education
2 vs 1 1.6882 1.7148 0.985 0.325
3 vs 1 2.5348 1.7219 1.472 0.1412
4 vs 1 3.0789 1.7697 1.74 0.0821

Reduced model: linear regression between HDL and covariates
variables estimate std. error t p R
2

intercept 23.1582 3.0758 7.529 8.18×10
-14

0.0577 age_at_lipids 0.1258 0.0444 2.834 0.0046
sex 7.1324 0.7184 9.928 <2×10
-16

Equation 2: linear regression between rank-based inversed residual and global ancestry
intercept 0.2011 0.0509 3.95 8.14×10
-5

0.0114
NH_prop -0.4566 0.1021 -4.473 8.21×10
-6

LDL. We found that LDL is only correlated with age at taking lipid medications and gender in this
case, and there isn’t a statistically significant association between global ancestry and LDL level
(Table 10).

Table 10. Statistics of coefficients in the regression of LDL
Full model: linear regression between LDL and covariates
variables estimate std. error t p R
2

intercept 148.1484 9.0622 16.348 <2×10
-16

0.0349
age_at_lipids -0.5814 0.1118 -5.199 2.24×10
-7

sex 7.7124 1.7707 4.355 1.41×10
-5

education
2 vs 1 4.7659 4.1763 1.141 0.254
3 vs 1 9.5296 4.1935 2.272 0.0232
4 vs 1 6.5174 4.3109 1.512 0.1308

Reduced model: linear regression between LDL and covariates
variables estimate std. error t p R
2

intercept 159.6125 7.5508 21.138 <2×10
-16

0.0309
age_at_lipids -0.6538 0.109 -6.000 2.40×10
-9

20
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
sex 7.5304 1.7629 4.272 2.05×10
-5

Equation 2: linear regression between rank-based inversed residual and global ancestry
intercept 0.0447 0.0514 0.870 0.385
5.64×10
-4

NH_prop -0.1014 0.103 -0.985 0.325

TG (triglycerides). We found that TG is only adjusted with age at taking lipid medications and
gender in this case, and there isn’t a statistically significant association between global ancestry
and TG level (Table 11).

Table 11. Statistics of coefficients in the regression of triglycerides
Full model: linear regression between TG and covariates
variables estimate std. error t p R
2

intercept 172.6882 19.9441 8.659 <2×10
-16

0.0046
age_at_lipids -0.5401 0.2457 -2.199 0.028
sex -2.5975 3.8938 -0.667 0.505
education
2 vs 1 -12.9997 9.2142 -1.411 0.158
3 vs 1 -8.864 9.2538 -0.958 0.338
4 vs 1 -13.9002 9.509 -1.462 0.144

Reduced model: linear regression between TG and covariates
variables estimate std. error t p R
2

intercept 160.8338 16.584 9.698 <2×10
-16

0.0030 age_at_lipids -0.5182 0.2392 -2.166 0.0304
sex -2.828 3.8735 -0.73 0.4654
Equation 2: linear regression between rank-based inversed residual and global ancestry
intercept -0.0337 0.0511 -0.659 0.51
3.21×10
-4

NH_prop 0.0765 0.1025 0.746 0.456

TC (total cholesterol). We found that TC is only correlated with age at taking lipid medications and
gender in this case, and there is a marginally significant association between global ancestry and
TC level. Generally, TC will decrease ~0.0283 unit of standard deviation with per 10% of increase
of global ancestry from NH (Table 12).

Table 12. Statistics of coefficients in the regression of total cholesterol
Full model: linear regression between TC and covariates
variables estimate std. error t p R
2

intercept 199.2207 9.6597 20.624 <2×10
-16

0.0499
age_at_lipids -0.4936 0.119 -4.148 3.51×10
-5

sex 14.0703 1.8859 7.461 1.36×10
-13

education 2 vs 1 3.6468 4.4628 0.817 0.4139
21
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
3 vs 1 10.2459 4.4819 2.286 0.0224
4 vs 1 6.4313 4.6056 1.396 0.1628

Reduced model: linear regression between TC and covariates
variables estimate std. error t p R
2

intercept 211.7619 8.0436 26.327 <2×10
-16

0.0441 age_at_lipids -0.5816 0.116 -5.013 5.90×10
-7

sex 13.7749 1.8787 7.332 3.46×10
-13

Equation 2: linear regression between rank-based inversed residual and global ancestry
intercept 0.1248 0.051 2.447 0.0145
0.0036
NH_prop -0.2834 0.1023 -2.772 0.0056

Type 2 diabetes. We found that type 2 diabetes isn’t correlated with gender in this case, and there
is a marginally significant association between global ancestry and type 2 diabetes. The odds ratio
of developing type 2 diabetes will increase ~0.0599 unit with per 10% increase of global ancestry
from NH (Table 13).

Table 13. Statistics of coefficients in the regression of type 2 diabetes
Full model: logistics regression between type 2 diabetes and covariates
Variables estimate std. error t p
Intercept -6.98 0.4753 -14,685 <2×10
-16

NH_prop 0.3994 0.1689 2.365 0.018
Age 0.0617 0.0054 11.463 <2×10
-16

BMI 0.1266 0.0077 16.5 <2×10
-16

Sex -0.0602 0.0785 -0.767 0.4434
education
2 vs 1 -0.3637 0.1659 -2.192 0.0284
3 vs 1 -0.5893 0.1688 -3.491 0.0005
4 vs 1 -0.6476 0.1772 -3.654 0.0003

Reduced model: logistics regression between type 2 diabetes and covariates
Variables estimate std. error t p
Intercept -7.0967 0.4507 -15.745 <2×10
-16

NH_prop 0.4000 0.1689 2.369 0.0178
Age 0.0619 0.0054 11.514 <2×10
-16

BMI 0.1269 0.0077 16.563 <2×10
-16

education
2 vs 1 -0.3643 0.1658 -2.197 0.0281
3 vs 1 -0.5855 0.1687 -3.47 0.0005
4 vs 1 -0.6411 0.177 -3.622 0.0003

Obesity. Obesity is defined as BMI is over 32. All population is stratified into obesity and non-
22
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
obesity groups. Obesity is correlated with age, type 2 diabetes status and smoking status and age ×
type 2 diabetes, and after adjusting for these factors there is a statistically significant association
between global ancestry and obesity (Table 14). The odds of obesity will increase 0.083 with per
10% increase of global Polynesian ancestry. Moreover, the odds will increase 0.079, and 0.088 unit
with per 10% increase of Polynesian ancestry for those with and without type 2 diabetes,
respectively (Table 15, Table 16).

Table 14. Statistics of coefficients in the regression of obesity
Variables estimate std. error t p
Intercept -1.6717 0.495 -3.377 0.0007
NH_prop 1.5946 0.1818 8.769 <2×10
-16

Age -0.0151 0.0092 -1.628 0.1035
t2d_stat_new 4.1249 0.6713 6.145 8.00×10
-10

cig2
1
0.3846 0.1848 2.081 0.0374
age:t2d_stat_new -0.0526 0.0124 -4.244 2.20×10
-5

1
: cig2 was created based on the original smoking levels, which is a binary variable. Level 0 in cig2
categorized level 1,2,3,4,5 in the original smoking status together, while level 1 represented the original
cigarettes consume level 6.

Table 15. Summary of regressions of obesity among subgroups
Traits Equation 1: regression with covariates
P of global
ancestry
Obesity
T2d
Logit(obesity)=2.5657+1.3148 ×NH-0.0669 ×age+0.2427 ×
cig2
5.03×10
-8

Non-t2d
Logit(obesity)=-1.7997+1.9472 ×NH-0.0162 ×age+0.6125 ×
cig2
1.20×10
-12

Table 16. Statistics of coefficients in the regression of obesity among subgroups
Type 2 diabetes
Variables estimate std. error t p
Intercept 2.5657 0.4647 5.521 3.38×10
-8

NH_prop 1.3148 0.2412 5.450 5.03×10
-8

Age -0.0669 0.0082 -8.127 4.40×10
-16

cig2 0.2427 0.2319 1.047 0.295

Non-type 2 diabetes
Variables estimate std. error t p
Intercept -1.7997 0.5057 -3.559 3.72×10
-4

NH_prop 1.9472 0.274 7.106 1.20×10
-12

23
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
Age -0.0162 0.0093 -1.733 0.0832
cig2 0.6125 0.2917 2.100 0.0358

Admixture mapping. We examined the associations between local ancestry and quantitative traits
using linear regression. For categorical traits we tested additionally the CC and CO models in
addition to the regression models. We used a p-value threshold of 10
-6
as the threshold for genome-
wide significance in admixture mapping. The significance threshold for admixture mapping should
be higher than the commonly used 5x10
-8
value in GWAS because of correlation between ancestry
of individual SNPs across the genome. We thus adopted the significance threshold from previous
Latino American admixture mapping studies
14
, which, like Native Hawaiians, is a multi-ancestry
admixed population that became admixed during the colonial era.
To illustrate the difference between CO, CC, and regression model, we examined QQ plot for
admixture mapping of type 2 diabetes on chromosome 6 (Figure 4). Although the CO model
implied that plenty of SNPs may be associated to type 2 diabetes, the QQ plot showed a unrealistic
departure of nearly all SNPs from the null distribution of p-values, suggesting that there is a high
false-positive rate with CO models. This can be seen further in Figure 5C. There are multiple peaks
where the local ancestry among cases significantly deviates from the genome-wide average, leading
to many significant associations in a CO model. However, comparing the local ancestry estimate
among controls, those peaks all correspond to a similar peak among controls, thus implying the
deviation is not related to case status. For CC and regression models, most loci do not show
significant associations with type 2 diabetes, as expected. We thus focused on interpreting the
results based on CC and regression models.
Across the 7 quantitative traits and 4 categorical traits we tested via admixture mapping, we
found two regions significantly associated with the traits of interest: one region on chromosome 6
for type 2 diabetes detected by CC model and one region on chromosome 22 for triglycerides
detected by regression model (Figure 5a, 6a).

Fig 4. QQ plot on chromosome 6 for type 2 diabetes
a) QQ plot of CO model, b) QQ plot of CC model, c) QQ plot of regression mode

24
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians

Fig 5. Admixture mapping for type 2 diabetes on chromosome 6
a) Genome-wide Manhattan plot for type 2 diabetes, b) -log 10 p-value on chromosome 6, c) average
25
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
local ancestry among diagnosed and undiagnosed population on chromosome 6

Fig 6. Admixture mapping for triglycerides on chromosome 22
a) Genome-wide Manhattan plot for triglycerides, b) -log 10 p-value on chromosome 22

For type 2 diabetes, the significant region locates on chr6:56423870-chr6:70343136. 2 peaks
surpass the threshold of significance level continuously, for which the top signal band is at
chr6:57095205-chr6:57661007 (Figure 5b). There is a second region nearby, at chr6:87144649-
chr6:87647591, that nearly surpassed the genome-wide threshold (Table 17). We searched the
GWAS catalog and identified 5 SNPs previously identified to be associated with type 2 diabetes: 3
in the first region and 2 in the second region. One SNP, rs115395006, was not imputed in our data.
We thus examined whether the remaining 4 SNPs could explain the admixture signal we detected
(Table 18).
In the 4 SNPs examined, only rs9362054 showed a large allele frequency difference between
populations (Table 19, 66.84% difference between NH and EUR). We tested if this variant could
26
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
explain the admixture signal by including the genotype of this variant in the admixture model.
However, the significance level only slightly decreased in this critical region (Figure 7, P-value
increased from 1.2683 ×10
-6
to 3.2890 ×10
-6
). This suggests that rs9362054 cannot explain the
admixture signal we detected, suggesting there may be additional variants with large frequency
differences in this region that may contribute to type 2 diabetes risk. We also tested a model where
all 4 SNPs were included. In this case, we find the significance level of p-value actually slightly
increased in the top signal band (Figure 8). Therefore, there may be a Native Hawaiian-specific
association between local ancestry and type 2 diabetes in these regions on chromosome 6 and it is
worth future follow-up.
For triglycerides, the significant region is located at chr22:22394144-chr22:29011279, and the
top signal band is at chr22:21648225-chr22:23013355. However, we found no variant associated
with triglyceride level in the GWAS catalog located in this region. Thus, this region could be a
newly discovered region related to triglycerides.

Table 17. Summary of statistics of significant region
Type 2 diabetes
order
Start
position/Mb
Start genetic
position
End
position/Mb
End genetic
position
Maximum
p-value
1 80.11905 chr6:56423870 80.16702 chr6:56701129 6.0116
2 80.34789 chr6:57095205 80.59808 chr6:57661007 6.7747
3 81.51103 chr6:65245009 81.86123 chr6:65761684 6.3875
4 82.87077 chr6:67040286 82.91880 chr6:67149979 6.1629
5 83.10806 chr6:67555387 83.21798 chr6:68547602 6.3316
6 83.87756 chr6:68547602 84.48833 chr6:69295030 6.6145
7 84.82111 chr6:70104848 84.89333 chr6:70343136 6.3843
8 95.75998 chr6:87144649 95.90955 chr6:87647591 5.8968

Triglycerides
order
Start
position/Mb
Start genetic
position
End
position/Mb
End genetic
position
Maximum
p-value
1 16.58746 chr22:22394144 16.84205 chr22:29011279 6.7414

Table 18. Summary of possible influencing SNPs from GWAScatalog
27-30

Variant
Genetic
position
1

P Influenced traits
Mapped
gene
Exists in
imputed
data
rs1048886 chr6:71289189 3.00×10
-8

type 2 diabetes
mellitus
Y
rs79976124 chr6:66618657 2.00×10
-6

type 2 diabetes
mellitus
NUFIP1P,
ADH5P4
Y
rs10498828 chr6:65533066 9.00×10
-6
type 2 diabetes EYS Y
27
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
mellitus
rs115395006 chr6:87312930 4.00×10
-6

Coronary artery
calcified
atherosclerotic
plaque score in type
2 diabetes
AL353133.1,
RN7SL643P
N
rs9362054 chr6:85178268 1.00×10
-6

Diabetic retinopathy
in type 2 diabetes
LINC01611 Y
1
: genetic architecture is based on hg19

Table 19. Summary of allele frequency difference between NH and EUR, EAS
NH vs EUR
Variants
Physical
position
1

NH EUR Absolute
value of allele
difference
Difference
rate/%
2

Allele
1
Allele
2
Allele
1
Allele
2
rs1048886
chr6:
71289189
A:
0.8418
T:
0.1582
A:
0.8540
T:
0.1460
0.0122 1.44
rs79976124
chr6:
66618657
G:
0.7994
A:
0.2006
G:
0.7129
A:
0.2871
0.0866 12.14
rs10498828
chr6:
65533066
C:
0.7938
T:
0.2062
C:
0.7822
T:
0.2178
0.0116 1.48
rs9362054
chr6:
85178268
T:
0.2797
C:
0.7203
T:
0.4666
C:
0.5334
0.1869 66.84

NH vs EAS
Variants
Physical
position
1

NH EAS Absolute
value of allele
difference
Difference
rate/%
2

Allele
1
Allele
2
Allele
1
Allele
2
rs1048886
chr6:
71289189
A:
0.8418
T:
0.1582
A:
0.9167
T:
0.0833
0.0749 8.89
rs79976124
chr6:
66618657
G:
0.7994
A:
0.2006
G:
0.9226
A:
0.0774
0.1232 15.41
rs10498828
chr6:
65533066
C:
0.7938
T:
0.2062
C:
0.8899
T:
0.1101
0.0961 12.11
rs9362054
chr6:
85178268
T:
0.2797
C:
0.7203
T:
0.3065
C:
0.6935
0.0269 9.61
1
: genetic architecture is based on hg19
2
: difference rate %=absolute value of allele frequency difference ÷min(allele frequency) ×100%

28
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians

Fig 7. Admixture mapping on chromosome 6 before and after adjusted by rs9362054
a) comparison of CC model, b) comparison of regression model

29
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians

Fig 8. Admixture mapping before and after adjusted by all potential confounding SNPs
a) comparison of CC model, b) comparison of regression model, c) comparison between CC and
30
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
regression adjusted model

Discussion
This study aimed to fill in the gap in genetic research by focusing on studying the association
of genetic ancestry, both globally and locally, to diseases in which Native Hawaiians showed excess
risk and to related quantitative traits and biomarkers. We found that Native Hawaiians derived on
average 43.70% of their genomic ancestry from Polynesian, with the remaining roughly equally
split between Europeans and East Asians. We found that global Polynesian ancestry from NH is
positively associated with BMI, HDL, type 2 diabetes and obesity, and is negatively associated with
insulin and TC. The association for BMI, HDL, and obesity are strong (P-value <2 ×10
-16
, 8.21 ×10
-
6
, <2 ×10
-16
), while the association for other traits are nominal that would not have survived a
Bonferroni correction for the multiple traits that we tested. Moreover, from examination of
association between local ancestry and these quantitative and categorical traits, we found that the
local Polynesian ancestry from NH appears to be associated with type 2 diabetes on chromosome
6 and triglyceride level on chromosome 22. In both of these regions we showed that known GWAS
variants for these traits could not have driven the signal we detected in admixture mapping,
suggesting that these regions should be targeted for replication and further detailed investigation in
the future.
For type 2 diabetes, we detected both a nominal association of global ancestry and disease
status (P-value = 0.0178) and a significant association of local ancestry on 80.12Mb-84.89Mb on
chromosome 6 and disease status (P-value = 1.6800 ×10
-7
). However, we noted that the associations
of local ancestry should not be implied that there is an association of global ancestry expected. To
be more precise, the association of global ancestry would provide a genome-wide association
between genotypes and phenotypes, but it doesn’t mean that it bears an inevitable link with the
associations of local ancestry, and vice versa. Because most SNPs only have a minor effect, it is
rare to discover a locus with a huge impact. A significant association of global ancestry may be
attributed to many SNPs that each exerts a small effect but together provides a signal. On the other
hand, there is a huge diversity in SNP effects. Although sometimes we can detect several significant
loci, their effects may be counteracted, which makes the global ancestry not associated with the
phenotype. Furthermore, an association to global ancestry may still be driven by cultural or other
non-genetic factors associated with global ancestry that could not be properly controlled. For
example, socioeconomic status (SES) is a common confounder, associated with both proportion of
global Polynesian ancestry as well as chance of disease risk. We attempted to correct for the effect
of SES in our study by including variables such as education level and smoking status, but they
may be imperfect proxies. As some quantitative traits are correlated with income
33,34
, it is better to
include covariates related to income in the study, which is unavailable. In the future, it may be
fruitful to explore a neighborhood SES index as covariate in the global ancestry analysis. That is,
we could link each individual to the neighborhood he or she lives in, and use the median income
level in the neighborhood as a proxy for SES in the model.
Moreover, our study can still be further improved or strengthened in a number of ways. First,
we will need to increase the sample size to improve the statistical power. Currently, QQ plots show
31
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
that there may be a deficiency in statistical power for most CC and regression models, particularly
for breast cancer and prostate cancer where the number of cases are few (112 cases for breast cancer,
69 cases for prostate cancer). Second, to better leverage both the increase power in a CO model and
the robustness in a CC model, an appropriate weighing between the two models may achieve the
best result. Such an approach is in development, such as the Bayesian Model Averaging (BMA)
method under development by Lilit Chemenyan and David Conti. This approach would generate a
composite model based on both CC and CO models, in which the coefficients are determined by
Bayesian Theory. Once it is released we will apply it to the admixture mapping study conducted
here to see if we can detect additional signals.
There are also some technical aspects of our study design that could be scrutinized further.
First, we applied the genome-wide significance threshold of 10
-6
in our admixture mapping study.
We derived this value based on a previous study of admixture mapping in Latino Americans, who
used the number of tests to determine this threshold. We reasoned that this was a reasonable
threshold to use, because both the Latinos and Native Hawaiians demonstrated multi-way
admixture in their past, both derived a large part of their primary ancestry from an indigenous
population (the Native Americans and the Polynesians) in similar proportion (~50% in Latinos and
~45% in Native Hawaiians), and both became admixed during the encounter with Europeans during
the colonial era. Nevertheless, the LD structure may be different between Native Hawaiians and
Latinos, and thus we should further investigate the appropriate significance threshold to use in
admixture mapping in Native Hawaiians. Secondly, we also used the genetic map from
1000Genomes to call local ancestry using RFMix, instead of a Native Hawaiian specific genetic
map. In our preliminary testing we found that there are minimal impacts if different maps were
used in RFMix, and the accuracy of local ancestry calls are more influenced by the appropriateness
of reference panel. Moreover, comparisons of genetic map between very divergent European and
African populations showed similarities at large scales (1 Mb)
31,32
, suggesting that calling local
ancestry using RFMix, where chromosomal segments are usually on the order of 1 Mb should be
less impacted by fine-scale differences in the genetic map used.
Finally, this study also raises additional questions for future investigations. Specifically, it is
common that there is a high prevalence of disease like obesity throughout Polynesia. It is thus
worthwhile to investigate whether the high prevalence of obesity and type 2 diabetes in the Native
Hawaiians result from a common evolutionary pathway originating from the Polynesian islands.
Moreover, there is a question whether adaptation to Hawaii islands would exacerbate their risk for
these traits, as a recently identified variant in the C R EB RF gene in Samoans were shown to be
associated with BMI and obesity and show genetic pattern consistent with positive selection
35
.
There may be more of these signals in Native Hawaiians that have not yet been found.

References
1. Braden, K.W. & Nigg, C.R. Modifiable Determinants of Obesity in Native Hawaiian and Pacific
Islander Youth. Hawaii J Med Public Health 75, 162-71 (2016).
2. Madan, A. et al. More than black and white: differences in predictors of obesity among Native
Hawaiian/Pacific Islanders and European Americans. Obesity (Silver Spring) 20, 1325-8 (2012).
32
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
3. Maskarinec, G. et al. Diabetes incidence based on linkages with health plans: the multiethnic cohort.
Diabetes 58, 1732-8 (2009).
4. Pike, M.C. et al. Breast cancer in a multiethnic cohort in Hawaii and Los Angeles: risk factor-
adjusted incidence in Japanese equals and in Hawaiians exceeds that in whites. Cancer Epidemiol
Biomarkers Prev 11, 795-800 (2002).
5. Singh, G.K. & Lin, S.C. Dramatic Increases in Obesity and Overweight Prevalence among Asian
Subgroups in the United States, 1992-2011. ISRN Prev Med 2013, 898691 (2013).
6. Tung, W.C. & Barnes, M. Heart Diseases Among Native Hawaiians and Pacific Islanders. Home
Health Care Management and Practice 26, 110-113 (2014).
7. Grandinetti, A. et al. Relationship of blood pressure with degree of Hawaiian ancestry. Ethn Dis
12, 221-8 (2002).
8. Claw, K.G. et al. A framework for enhancing ethical genomic research with Indigenous
communities. Nat Commun 9, 2957 (2018).
9. Popejoy, A.B. & Fullerton, S.M. Genomics is failing on diversity. Nature 538, 161-164 (2016).
10. Diamond, J.M. Taiwan's gift to the world. Nature 403, 709-10 (2000).
11. Kim, S.K. et al. Population genetic structure and origins of Native Hawaiians in the multiethnic
cohort study. PLoS One 7, e47881 (2012).
12. Skoglund, P. et al. Genomic insights into the peopling of the Southwest Pacific. Nature 538, 510-
513 (2016).
13. Jansen, P.R. et al. Genome-wide analysis of insomnia in 1,331,010 individuals identifies new risk
loci and functional pathways. Nat Genet 51, 394-403 (2019).
14. Shriner D. Overview of admixture mapping. Curr Protoc Hum Genet. 2013;Chapter 1:Unit 1.23.
15. Pino-Yanes M, Gignoux CR, Galanter JM, et al. Genome-wide association study and admixture
mapping reveal new loci associated with total IgE levels in Latinos. J Allergy Clin Immunol.
2014;135(6):1502-10.
16. Cheng CY , Kao WHL, Patterson N, Tandon A, Haiman CA, et al. (2009) Admixture Mapping of
15,280 African Americans Identifies Obesity Susceptibility Loci on Chromosomes 5 and X. PLOS
Genetics 5(5): e1000490. https://doi.org/10.1371/journal.pgen.1000490
17. Torgerson DG, Gignoux CR, Galanter JM, Drake KA, Roth LA, Eng C, et al. Case-control
admixture mapping in Latino populations enriches for known asthma-associated genes. J Allergy
Clin Immunol. 2012;130(1):76–82. e12. Epub 2012/04/17
18. Kolonel LN, Henderson BE, Hankin JH, et al. A multiethnic cohort in Hawaii and Los Angeles:
baseline characteristics. Am J Epidemiol 2000;151:346-57
19. G. Wojcik, et al. The PAGE Study: How Genetic Diversity Improves Our Understanding of the
Architecture of Complex Traits. bioRxiv 188094; doi: https://doi.org/10.1101/188094
20. S. McCarthy et al. (2016) A reference panel of 64,976 haplotypes for genotype imputation, Nature
Genetics. 48(10):1279-83
21. P. Loh, et al. (2016) Reference-based phasing using the Haplotype Reference Consortium panel,
Nature Genetics. 48(11):1443-1448
22. R. Durbin (2014) Efficient haplotype matching and storage using the positional Burrows-Wheeler
transform (PBWT), Bioinformatics. 30(9):1266-72
33
The Impact of Global and Local Polynesian Ancestry on Complex Traits in Native Hawaiians
23. Manichaikul A, Mychaleckyj JC, Rich SS, Daly K, Sale M, Chen WM (2010) Robust relationship
inference in genome-wide association studies. Bioinformatics 26(22):2867-2873
24. Maples BK, Gravel S, Kenny EE, Bustamante CD. RFMix: a discriminative modeling approach
for rapid and robust local-ancestry inference. Am J Hum Genet. 2013;93(2):278-88.
25. D. Conti, personal communication
26. Buniello A, et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies,
targeted arrays and summary statistics 2019. Nucleic Acids Research, 2019, Vol. 47 (Database
issue): D1005-D1012.
27. Sim X, Ong RTH, Suo C, Tay WT, Liu J, et al. (2011) Transferability of Type 2 Diabetes Implicated
Loci in Multi-Ethnic Cohorts from Southeast Asia. PLOS Genetics 7(4): e1001363.
https://doi.org/10.1371/journal.pgen.1001363
28. Imamura M, Takahashi A, Yamauchi T, et al. Genome-wide association studies in the Japanese
population identify seven novel loci for type 2 diabetes. Nat Commun. 2016;7:10531. Published
2016 Jan 28. doi:10.1038/ncomms10531
29. Scott RA, Scott LJ, Mägi R, et al. An Expanded Genome-Wide Association Study of Type 2
Diabetes in Europeans. Diabetes. 2017;66(11):2888-2902.
30. Divers J, Palmer ND, Langefeld CD, et al. Genome-wide association study of coronary artery
calcified atherosclerotic plaque in African Americans with type 2 diabetes. BMC Genet.
2017;18(1):105. Published 2017 Dec 8. doi:10.1186/s12863-017-0572-9
31. D. Wegmann, et al. Recombination rates in admixed individuals identified by ancestry-based
inference. Nat Genet. 2011 Jul 20; 43(9):847-53.
32. Hinch AG, Tandon A, Patterson N, et al. The landscape of recombination in African Americans.
Nature. 2011;476(7359):170-5. Published 2011 Jul 20. doi:10.1038/nature10336
33. Morgenstern M,et al. Relation between socioeconomic status and body mass index: evidence of an
indirect path via television use. Arch Pediatr Adolesc Med. 2009;163(8):731-8.
34. Dinsa GD, et al. Obesity and socioeconomic status in developing countries: a systematic review.
Obes Rev. 2012;13(11):1067-79.
35. Minster, R., et al. (2016). A thrifty variant in CREBRF strongly influences body mass index in
Samoans. Nature Genetics, 48(9), pp.1049-1054.

Abbreviation
AFR: African EAS: East Asian
EUR: European NH: Native Hawaiian
TC: total cholesterol TG: triglycerides
T2d: type 2 diabetes

Abstract (if available)

Abstract Epidemiologic studies have shown that Native Hawaiians exhibit an excess risk for obesity, Type 2 diabetes, cardiovascular diseases and several common cancers compared to European or Asian Americans living in the Hawaiian islands today. However, like many non-European or indigenous populations, Native Hawaiians are understudied, particularly from a genetic standpoint. To fill this gap, we took advantage of the availability of genome-wide array data on nearly 4,000 Native Hawaiians from the Multiethnic Cohort to study the risk related to their genetic ancestry for these complex diseases and the associated quantitative traits. Specifically, we first identified a subset of Native Hawaiians with extremely high proportion of Polynesian ancestry to construct an internal reference panel to help us estimate the genomic proportion of Polynesian ancestry (“global ancestry”), as well as Polynesian ancestry along each chromosome (“local ancestry”), in the remaining 3,428 unrelated individuals. We then tested the associations between proportion of global Polynesian ancestry and both binary (obesity, type 2 diabetes, prostate cancer, breast cancer) and quantitative (BMI, glucose, insulin, HDL, LDL, triglyceride, total cholesterol) traits. After adjusting for available covariates on socioeconomic status, we found that per 10% increase in Polynesian genetic ancestry, there is a 17% increase in the odds of being obese (P < 2×10⁻¹⁶), a 0.076 s.d. increase in BMI (P < 2×10⁻¹⁶) and a 0.046 s.d. increase in HDL (P = 8.2×10⁻⁶). We also found weaker nominal associations between proportion of Polynesian genomic ancestry and insulin and total cholesterol. To examine regions along the genome that might harbor Polynesian-specific variants associated with risk of disease or biomarkers, we conducted admixture mapping on these traits and identified genomic regions at approximately 57.4Mb-70.3Mb on chromosome 6 and 22.4Mb-29.0Mb on chromosome 22 to be putatively associated with Type 2 diabetes and triglycerides, respectively

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Native American ancestry among Hispanic Whites is associated with higher risk of childhood obesity: a longitudinal analysis of Children’s Health Study data

PDF

Understanding acute lymphoblastic leukemia in different ethnic groups in the United States

PDF

A global view of disparity in imputation resources for conducting genetic studies in diverse populations

PDF

The role of genetic ancestry in estimation of the risk of age-related degeneration (AMD) in the Los Angeles Latino population

PDF

Shortcomings of the genetic risk score in the analysis of disease-related quantitative traits

PDF

A genome wide association study of multiple sclerosis (MS) in Hispanics

PDF

Association of single nucleotide polymorphisms in GCK, GCKR and PNPLA3 with type 2 diabetes related quantitative traits in Mexican-American population

PDF

Diet quality and pancreatic cancer incidence in the multiethnic cohort

PDF

The impact of perceived parental and self-reported stress on BMI and body composition in young adults

PDF

Identifying genetic, environmental, and lifestyle determinants of ethnic variation in risk of pancreatic cancer

PDF

An assessment of impact of early local progression on subsequent risk for the treatment failure in adolescent and young adult patients with non-metastatic osteosarcoma

PDF

The environmental and genetic determinants of cleft lip and palate in the global setting

Asset Metadata

Creator Sun, Hanxiao (author)

Core Title The impact of global and local Polynesian genetic ancestry on complex traits in Native Hawaiians

School Keck School of Medicine

Degree Master of Science

Degree Program Biostatistics

Publication Date 11/09/2019

Defense Date 05/09/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag admixture mapping,BMI,genome-wide association study,HDL,insulin,minor populations,Native Hawaiian,OAI-PMH Harvest,triglycerides,type 2 diabetes

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Chiang, Charleston (committee chair), Jiang, Xuejuan (committee member), Monroe, Kristine (committee member)

Creator Email 3130100966@zju.edu.cn,hanxiaos@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-156877

Unique identifier UC11660743

Identifier etd-SunHanxiao-7434.pdf (filename),usctheses-c89-156877 (legacy record id)

Legacy Identifier etd-SunHanxiao-7434.pdf

Dmrecord 156877

Document Type Thesis

Format application/pdf (imt)

Rights Sun, Hanxiao

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA