Available via license: CC BY 4.0
Content may be subject to copyright.
J. Pers. Med. 2022, 12, 1974. https://doi.org/10.3390/jpm12121974 www.mdpi.com/journal/jpm
Article
The Penn Medicine BioBank: Towards a Genomics-Enabled
Learning Healthcare System to Accelerate Precision Medicine
in a Diverse Population
Anurag Verma
1,2,
*, Scott M. Damrauer
2,3,4
, Nawar Naseer
2
, JoEllen Weaver
2
, Colleen M Kripke
1,2
,
Lindsay Guare
5
, Giorgio Sirugo
1,2
, Rachel L. Kember
6
, Theodore G Drivas
1
, Scott M. Dudek
3
, Yuki Bradford
3
,
Anastasia Lucas
3
, Renae Judy
4
, Shefali S. Verma
5
, Emma Meagher
1,2
, Katherine L. Nathanson
1,7
,
Michael Feldman
5
, Marylyn D. Ritchie
3
, Daniel J. Rader
1,2,3,
* and The Penn Medicine BioBank †
1
Department of Medicine, Division of Translational Medicine and Human Genetics, University of
Pennsylvania, Philadelphia, PA 19104, USA
2
Institute for Translational Medicine and Therapeutics, University of Pennsylvania,
Philadelphia, PA 19104, USA
3
Department of Genetics, University of Pennsylvania, Philadelphia, PA 19104, USA
4
Department of Surgery, Division of Vascular Surgery and Endovascular Therapy, University of
Pennsylvania, Philadelphia, PA 19104, USA
5
Department of Pathology, University of Pennsylvania, Philadelphia, PA 19104, USA
6
Department of Psychiatry, University of Pennsylvania, Philadelphia, PA 19104, USA
7
Abramson Cancer Center, Perelman School of Medicine, University of Pennsylvania,
Philadelphia, PA 19104, USA
* Correspondence: anurag.verma@pennmedicine.upenn.edu (A.V.); ra[email protected] (D.J.R.)
† A full list of contributions from Penn Medicine BioBank Team is provided in the supplement.
Abstract: The Penn Medicine BioBank (PMBB) is an electronic health record (EHR)-linked biobank
at the University of Pennsylvania (Penn Medicine). A large variety of health-related information,
ranging from diagnosis codes to laboratory measurements, imaging data and lifestyle information,
is integrated with genomic and biomarker data in the PMBB to facilitate discoveries and transla-
tional science. To date, 174,712 participants have been enrolled into the PMBB, including approxi-
mately 30% of participants of non-European ancestry, making it one of the most diverse medical
biobanks. There is a median of seven years of longitudinal data in the EHR available on participants,
who also consent to permission to recontact. Herein, we describe the operations and infrastructure
of the PMBB, summarize the phenotypic architecture of the enrolled participants, and use body
mass index (BMI) as a proof-of-concept quantitative phenotype for PheWAS, LabWAS, and GWAS.
The major representation of African-American participants in the PMBB addresses the essential
need to expand the diversity in genetic and translational research. There is a critical need for a
“medical biobank consortium” to facilitate replication, increase power for rare phenotypes and var-
iants, and promote harmonized collaboration to optimize the potential for biological discovery and
precision medicine.
Keywords: genomics; electronic health records; biobank; PMBB; precision medicine; learning health
system
1. Introduction
Precision medicine incorporates clinical, environmental, lifestyle, family, and ge-
nomic data to tailor disease management and optimize disease prevention and health
maintenance. Since the completion of the human genome project, large-scale genomic in-
formation linked to individual-level phenotype data has fueled biomedical discovery and
has become an integral component of precision medicine. Large amounts of clinical data
Citation: Verma, A.; Damrauer,
S.M.; Naseer, N.; Weaver, J.; Kripke,
C.M.; Guare, L.; Sirugo, G.; Kember,
R.L.; Drivas, T.G.; Dudek, S.M.; et al.
The Penn Medicine BioBank:
Towards a Genomics-Enabled
Learning Healthcare System to
Accelerate Precision Medicine in a
Diverse Population. J. Pers. Med.
2022, 12, 1974. https://doi.org/
10.3390/jpm12121974
Academic Editor: Michael F. Murray
Received: 8 October 2022
Accepted: 19 November 2022
Published: 29 November 2022
Publisher’s Note: MDPI stays neu-
tral with regard to jurisdictional
claims in published maps and institu-
tional affiliations.
Copyright: © 2022 by the authors. Li-
censee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and con-
ditions of the Creative Commons At-
tribution (CC BY) license (https://cre-
ativecommons.org/licenses/by/4.0/).
J. Pers. Med. 2022, 12, 1974 2 of 16
are generated daily in clinical care and stored in the electronic health record (EHR). Link-
ing the phenomic data encompassed in the EHR with biospecimens and genomic data
from appropriately consented individuals at scale represents a tremendous opportunity
for biomedical discovery and precision medicine. Penn Medicine, part of the University
of Pennsylvania, is a large integrated academic health system with six hospitals and ten
multispecialty centers that serve south-central Pennsylvania, south-central New Jersey,
and northern Delaware. The Penn Medicine BioBank (PMBB) was launched with the in-
tent of harnessing clinical data for discovery, creating a genomics-enabled learning
healthcare system [1], and facilitating precision medicine for disease prevention and per-
sonalized therapy. Patients are enrolled under a single IRB-approved protocol that ena-
bles the acquisition of biospecimens, generation of genomic and multi-omic data, linkage
to clinical information included in the EHR, and permission to recontact participants for
future studies and/or the return of clinically relevant results. The scientific goal of the
PMBB is to promote the integration of clinical and genomic data to power biomedical dis-
covery and precision medicine. This report describes the operations and architecture of
the PMBB and summarizes the information on the first ~170,000 participants recruited.
2. Materials and Methods
2.1. Planning and Development of PMBB
Recognizing the need for access to large numbers of appropriately consented and
well-characterized human biospecimens to conduct translational research, in 2008, Penn
Medicine established an IRB protocol and process for consenting patients and obtaining
blood for genomic and biomarker research. In 2013, after a strategic planning process
identified a pressing institutional mandate for an expanded biobank resource, the Penn
Medicine BioBank (PMBB) was formally constituted, funded, and launched under the In-
stitute of Translational Medicine and Therapeutics (ITMAT) in order to ensure it was in-
tegrated with critical infrastructure relevant to clinical and translational research and pre-
cision medicine.
The PMBB protocol was established as an institutional umbrella protocol under
which any registered patient of Penn Medicine aged 18 or older was eligible, with no ex-
clusions except an inability to provide informed consent. The core features of the consent
include: (1) provision of a blood sample for biobanking and broad use for data generation,
including genomic data and permission to bank any other residual tissues obtained in the
context of clinical care; (2) permission to access data from the EHR for the purpose of
research; and (3) permission to recontact participants for potential future studies or to
return results.
2.2. Patient-Participant Recruitment
Initially, PMBB enrollment was accomplished through face-to-face encounters with
clinical research coordinators (CRCs) in outpatient clinical areas, prioritizing locations
where procedures that involved access to blood samples (phlebotomy labs, imaging pro-
cedures involving IV placement, cardiac catheterization labs, etc.) were being performed.
After the onset of the COVID-19 pandemic, in August of 2020, the PMBB transitioned to
remote recruitment efforts to prioritize patient-participant and staff safety by initiating an
electronic consent and enrollment process through REDCap, a secure web platform for
building and managing online databases and surveys. Simultaneously, a process for con-
sent utilizing the EHR (PennChart, Epic) was developed, which was initially done in per-
son at the time of patient check-in, and then also expanded to include consent at the time
of pre-check-in through their myPennMedicine online patient portal, available through
web browsers and mobile devices. Eligible patients scheduled for an upcoming outpatient
office visit at one of the UPHS clinic sites actively recruiting (Figure 1E) receive the PMBB
consent form as part of their electronic pre-visit check-in process and have the option to
complete the form online through this portal. The three-page consent form includes a link
J. Pers. Med. 2022, 12, 1974 3 of 16
to the PMBB website, the PMBB email address, and the PMBB enrollment telephone num-
ber, which is staffed by CRCs on weekdays from 7:30 a.m. to 5 p.m. local time to answer
questions potential participants may have as they are completing the consent procedure.
A small percentage (~5%) of patients who either are not eligible to receive the online pre-
check-in (e.g., certain surgical patients), or patients who skip the consent form during their
online pre-check-in, are consented in person by registration desk staff when the patient
physically reports for their appointment. PMBB brochures, with basic information about
the PMBB, the PMBB website link, and contact information, are also available for registra-
tion desk staff to distribute to patients during the consent process. The website also con-
tains a short video for patients explaining how PMBB participants contribute to scientific
research.
Figure 1. Recruitment and Demographics. (A) Distribution of enrollment through paper and elec-
tronic consent. (B) Cumulative numbers of participants consented and biospecimen sample collec-
tion. (C) Distribution of participants by age group and self-reported sex. (D) Distribution of partic-
ipants by self-reported race. (E) Density of recruitment around the six clinical sites of UPHS in Penn-
sylvania and New Jersey: Hospital of the University of Pennsylvania, Penn Presbyterian Medical
Center, Pennsylvania Hospital, Chester County Hospital, Lancaster General Health, Princeton
Health.
A) B)
C) D)
E)
J. Pers. Med. 2022, 12, 1974 4 of 16
2.3. Sample Collection
The PMBB patient participants consent to the collection of identifying information
(e.g., name, date of birth, medical record number, and contact information), information
from medical records (e.g., test results, medical procedures, medical diagnosis and proce-
dure codes, lab values, images such as X-rays, and medicines), blood samples (up to four
tablespoons), urine, saliva and/or respiratory specimens, and residual samples from clin-
ical pathology. There is no limit on the length of time samples may be kept in the biobank.
All blood samples are banked as whole blood, plasma, serum, buffy coat, and DNA for
future studies following stringent standard operating procedures. Samples are collected
in sterile vacutainer tubes barcoded with an identification number and scanned into a
system for sample tracking. Using the institution’s adopted laboratory information sys-
tem, LabVantage (LabVantage Solutions, Inc., Somerset, NJ, USA), biospecimens are pro-
cessed and tracked with time-date stamps to document processing and freezing times
throughout the laboratory workflow. Sample inventory is robustly supported with real-
time, adaptive sample pull lists and images of sample pull locations, as well as simulta-
neous creation of distribution boxes and decrement of sample aliquots.
Prior to the COVID-19 pandemic, blood samples were collected from patients at the
point of enrollment by CRCs cross-trained as certified phlebotomists. With the transition
to electronic consenting, the consenting and sample collection processes have been decou-
pled. For sample collection, an automated weekly report is generated containing a list of
consented patients for whom a blood sample is absent who have an upcoming phlebot-
omy appointment the following week at select Penn Medicine sites. Every Friday, three
PMBB physicians place the electronic lab order for PMBB blood draws in the EHR of these
patients. When they report to their phlebotomy appointment, the phlebotomist adds on
the PMBB blood order to the patient’s existing orders and collects the blood sample. One
6 mL EDTA tube is collected per patient. All blood samples are transported to the central-
ized clinical laboratory and logged in prior to being transferred to the PMBB core labora-
tory, where the blood is processed and stored following standard operating procedures.
Additionally, the PMBB banks residual biospecimens and tissues (for example,
blood, urine, cerebrospinal fluid, or tissue collected as part of clinical care) when available
as fresh, frozen or fixed dependent upon the tissue histology, following standard proce-
dures. Residual tissues are released by the Department of Pathology after examination if
the specimen or tissue is determined to be in excess of that required for patient care, or for
tissue or bodily fluid that would otherwise be waste material.
2.4. Genomic Data Generation
2.4.1. Genotyping and Imputation
DNA samples on approximately 44,000 PMBB participants have been genotyped to
date on an Illumina Global Screening Array v.2.0 (GSAv2) by the Regeneron Genomics
Center (RGC). The genotyping array chip has a backbone of 654,027 genetic markers as
well as additional ancestry informative markers. In addition, approximately 8595 of the
44,000 PMBB participants were also genotyped in the Center for Applied Genomics (CAG)
at the Children’s Hospital of Philadelphia on the GSAv1 and GSAv2 genotyping array.
After performing sample-level quality control (QC), genotype imputation was performed
using Eagle2 [2] and Minimac [3] software on the TOPMed Imputation Server [3]. Impu-
tation was performed for all autosomes, with TOPMed version R2 on a GRCh38 reference
panel. Cosmopolitan post-imputation QC included imputation score filtering (R2 > 0.7),
removal of palindromic variants, biallelic variant check, sex check, genotype call rate
(>99%) and sample call rate (>99%) filtering, minor allele frequency filtering (MAF > 1%),
and a Hardy–Weinberg equilibrium test (p-value > 1 × 10−08). We generated PCAs to adjust
for population structure and to identify genetically informed ancestry (GIA) from EIGEN-
SOFT [4].
J. Pers. Med. 2022, 12, 1974 5 of 16
2.4.2. Whole Exome Sequencing
Whole exome sequencing (WES) has been performed on approximately 44,000 par-
ticipants to date by the RGC. DNA samples were processed with the custom IDT xGen v1
exome capture platform and sequenced on the Illumina NovaSeq 6000 system on S4 flow
cells. Sequence alignment, variant identification, and genotype assignment were per-
formed using a WeCall variant caller. Sample level QC steps were then applied and sam-
ple sex errors, high rates of heterozygosity/contamination (D-stat > 0.4), low sequence cov-
erage (less than 85% of targeted bases achieving 20X coverage), or genetically identified
sample duplicates, were excluded. Additional filters were applied to pVCFs. Any SNV
genotype with a read depth of less than seven reads (DP < 7) was changed to a no-call.
After the application of the DP genotype filter, only the SNV variant sites that met at least
one of the following two criteria were retained: (1) at least one heterozygous variant gen-
otype with an allele balance ratio greater than or equal to 15% (AB ≥ 0.15); (2) at least one
homozygous variant genotype. The same filtering was applied to INDEL variants but
with an INDEL depth filter of DP < 10 and an INDEL allele balance cutoff of AB >= 0.20.
Multi-allelic variant sites in the PVCF file were normalized by left-alignment and repre-
sented as bi-allelic. The variant frequency data for exome sequences and imputed data are
available here: https://pmbb.med.upenn.edu/allele-frequency (accessed on 20 November
2022). The PMBB has a return of actionable results program for genomic findings that have
a potential impact on participant health; this program is beyond the scope of this manu-
script and will be described in a separate manuscript.
2.5. Clinical Data and Clinical Informatics Core
Clinical data are obtained through multiple sources, including a questionnaire com-
pleted at the time of recruitment and the Penn Medicine Clinical Data Warehouse and
PennG&P (Penn Genotype & Phenotype) platform. PennG&P contains over 5.6 million
patient records and other discrete clinical information amalgamated from 12 different
source systems throughout the enterprise. PennG&P is based on a standard research data
model called the Observational Medical Outcomes Partnership (OMOP), Common Data
Model (CDM) [5], which is used worldwide by the Observational Health Data Sciences
and Informatics (OHDSI) research consortium. It uses standardized language from na-
tional coding systems, such as SNOMED, LOINC [6], and RxNorm [7], for consistent
terms and the labeling of information. Additionally, the PMBB maps International Classi-
fication of Diseases (ICD)-9 and ICD-10 codes to 1866 discrete disease traits using Phecode
groupings [8].
2.6. Access to Data and Biospecimens
In keeping with the expansive mission of the PMBB, data and biospecimens are avail-
able to investigators throughout the Perelman School of Medicine and Penn Medicine for
a broad range of research. External collaborations, including those with other academic
institutions as well as biopharma, are encouraged and proceed through scientific collabo-
ration with identified local Penn investigators. All research studies must have study spe-
cific IRB approval because the umbrella PMBB IRB protocol covers only sample and data
acquisition, processing, storage, and dissemination. Researchers request access to data
and biospecimens using a simple REDCap project proposal intake form which is then re-
viewed by the PMBB Steering Committee. Proposals are evaluated for scientific merit and
priority, as well as the efficient use of data and biospecimens, as well as the ability to
impact the care provided by Penn Medicine clinicians. The PMBB Steering Committee
provides feedback to the investigator with either approval to move forward or with con-
cerns to be addressed. This REDCap project also includes a Data Access Agreement form
that must be signed by investigators prior to gaining access to any PMBB data. Per the
terms of this agreement, the sharing of PMBB data with additional collaborators, whether
J. Pers. Med. 2022, 12, 1974 6 of 16
they are internal or external to Penn, must be handled by the PMBB and not the investi-
gators themselves, to maintain the confidentiality and integrity of any protected health
information (PHI) included within PMBB datasets.
Standardized EHR clinical data are deidentified and provided to approved investi-
gators in a secure computing environment. For assistance with the creation of more com-
plicated phenotypes, researchers have access to the Clinical Informatics Core (CIC), a
shared resource that is managed by the Institute for Biomedical Informatics in collabora-
tion with the PMBB. The CIC is staffed by clinical data scientists with expertise in data
extraction, natural language processing (NLP), and data analysis.
3. Results
3.1. Enrollment
During the initial phase of recruitment from 2008 to 2013, 13,366 Penn Medicine pa-
tients were enrolled (Figure 1A,B). In 2013, recruitment was expanded, resulting in a
steady increase in enrollment to ~71,000 participants by the end of 2019 (Figure 1A,B). In
2020, the transition to electronic consenting triggered a rapid expansion in PMBB enroll-
ment, with the total number of participants more than doubling between 2020 and 2022.
There were 174,712 total participants enrolled in the PMBB as of September 2022 (Table 1,
Figure 1A,B). Presently, nearly 3500 new participants are being enrolled weekly across
two Penn Medicine hospitals that are actively recruiting through all their ambulatory
sites, and recruitment at the other four Penn Medicine hospitals is targeted to begin in
2023. The PMBB currently has obtained and processed blood biospecimens from approx-
imately 50% of enrolled participants; the recent shift to an electronic consent process has
resulted in enrollment outpacing sample collection (Figure 1B). Active processes are un-
derway to obtain biospecimens on the remainder of enrolled individuals. The goal is to
enroll >1 million Penn Medicine patient participants, with >90% providing a blood sample
for DNA and biomarker studies.
The PMBB participant population currently represents approximately 2.5% of active
Penn Medicine patients. Similar to the general Penn Medicine patient population, a
slightly higher percentage of PMBB participants are female (55.9%) as compared to male
(44.1%) (Table 1, Figure 1C). The age distribution of PMBB participants also tracks with
that of Penn Medicine patients, with participants ranging in age from 18 years to >100
years (Table 1 and Figure 1C). The age distributions differ slightly between sex, with males
trending towards an older age (Figure 1C). The PMBB cohort is diverse: 17% of enrolled
PMBB participants (and 25% of genotyped/sequenced participants) are identified as Afri-
can American, 4% as Asian, and 3.3% as Hispanic (Table 1; Figure 1D). With nearly 30,000
African-American patient participants currently enrolled, the PMBB has, to our
knowledge, the largest number of African-American participants of any single-institu-
tional medical biobank in the US. As shown in Figure 1E, most biobank participants reside
within the Philadelphia metropolitan area including New Jersey and Delaware (53.6%);
there is also a total of 3.3% of participants from other states across the US.
Table 1. Comparison of PMBB Participant Characteristics with UPHS Patient Characteristics.
PMBB
Participants
n (%)
Genotyped
Participants
n (%)
UPHS
Patients
n (%) †
Total 174,712 43,884 3,688,610
Gender
Female 97,674 (55.9%) 21,965 (50.1%) 2,042,868 (55.4%)
Male 77,055 (44.1%) 21,918 (49.9%) 1,604,210 (43.5%)
Other 17 (<1%) 1 (<1%) 107 (<1%)
Age Range (Years) 18–103 18–103 0–121
J. Pers. Med. 2022, 12, 1974 7 of 16
Age Groups
18–29 17,815 (10.2%) 4302 (9.8%) 392,532 (10.5%)
30–39 27,355 (15.7%) 5406 (12.3%) 518,679 (14.1%)
40–49 25,819 (14.8%) 5688 (13.0%) 430,489 (11.7%)
50–59 33,827 (19.4%) 9519 (21.7%) 458,952 (12.4%)
60–69 40,268 (23.0%) 10,839 (24.7%) 507,397 (13.8%)
70–79 28,582 (16.4%) 5941 (13.5%) 400,016 (10.8%)
80+ 8811 (5.0%) 2189 (5.0%) 333,781 (9.0%)
Self-reported Race
African American 29,372 (16.8%) 10,815 (24.6%) 672,461 (18.2%)
White 124,406 (71.2%) 29,329 (66.8%) 2,029,684 (55.0%)
Asian 7156 (4.1%) 979 (2.2%) 152,615 (4.1%)
Other 9386 (5.4%) 1372 (3.1%) 370,313 (10.0%)
Unknown 7499 (4.3%) 1761 (4.0%) 463,537 (12.6%)
Self-reported Ethnicity
Hispanic 5715 (3.3%) 1112 (2.5%) 174,179 (4.7%)
Non-Hispanic 165,713 (94.8%) 42,425 (96.7%) 3,290,018 (89.2%)
Unknown 3284 (1.9%) 347 (0.8%) 183,723 (5.0%)
Genetically-Inferred
Ancestry
African N/A 11,300 (25.7%) N/A
European N/A 30,360 (69.2%) N/A
East Asian N/A 680 (1.5%) N/A
South Asian N/A 573 (1.3%) N/A
Admixed American N/A 711 (1.6%) N/A
Other N/A 301 (0.7%) N/A
Median period of EHR
follow-up since enroll-
ment
7 years 5.7 years N/A
† Demographics based on EHR data on UPHS patients who have been seen at least once from 2008
to present. Data from following UPHS sites were included: Hospital of the University of Pennsyl-
vania, Penn Presbyterian Medical Center, Pennsylvania Hospital, Chester County Hospital, Prince-
ton Health.
3.2. Clinical Data Availability
There are over 66.7 million data points collected in the form of encounters, diagnosis
codes, procedure codes, and medication orders (Table 2). Across the PMBB cohort, over
46.7 million encounters, 10 million diagnosis codes, 3.6 million procedure codes, and 6.3
million medication orders have been recorded, averaging 268 encounters, 57 diagnosis
codes, 21 procedure codes, and 36 medication orders per individual (Table 2). The most
common diagnoses codes include hypertension, hypercholesterolemia, obesity, and in-
somnia (Figure 2).
J. Pers. Med. 2022, 12, 1974 8 of 16
Table 2. Clinical Data Availability for PMBB Participants.
Event Type Total Number of Events Mean Number of Events
(SD) *
Encounter 46,738,773 268 (321)
Diagnosis Code (number of
condition-related visits) 10,023,922 57.4 (58.7)
Procedure Code 3,621,056 20.7 (26.7)
Medication Order 27,914,486 159.8 (271.8)
* Average number of events each patient has for each event type.
Figure 2. Prevalence of diagnoses code among PMBB participants grouped by broader disease do-
main.
3.3. Phenome-Wide Association Study (PheWAS) of BMI
To evaluate the validity of the PMBB clinical data, a phenome-wide association study
(PheWAS) was conducted using mean body mass index (BMI) as the exposure and 1856
disease traits derived from grouping encounter diagnoses using phecodes as the outcome.
All the models were adjusted with age, sex, and self-reported race in the EHR. A total of
662 associations of BMI with at least one disease trait across the 18 disease categories
J. Pers. Med. 2022, 12, 1974 9 of 16
passed Bonferroni correction for multiple hypothesis testing (p < 2.6 × 10−5). The strongest
associations with BMI were with type 2 diabetes, hyperlipidemia, overweight, obesity,
sleep apnea, and osteoarthritis (all p < 1 × 10−331, Figure 3, Supplementary Table S1). The
associations with mean BMI show evidence of its effect on all organ systems, covering
associations with the spectrum of disease categories (Figure 3). Additional associations
include hypertension, heart failure, endometrial hyperplasia, bone fracture, depression,
pregnancy complications, and respiratory failure (Figure 3).
Figure 3. A phenome-wide association between mean body mass index and 1856 EHR-derived phe-
notypes.
J. Pers. Med. 2022, 12, 1974 10 of 16
3.4. Laboratory-Wide Association Study (LabWAS) of BMI
We extracted 24 clinical laboratory measurements from the EHR for all the PMBB
participants. These lab measurements were selected based on a common lab test in the
health system and contained at least 1000 individuals within each lab which was meas-
ured. We computed median values for each individual within each lab and, as a proof-of-
concept, evaluated their association with BMI. Linear regression was performed to test for
association and all the models were adjusted with age, sex and self-reported race. We rep-
licated many known associations between BMI and lab values (Figure 4, Table S2). For
example, blood glucose measures were significantly associated with increased BMI. Tri-
glyceride levels were significantly positively associated and high-density lipoprotein cho-
lesterol (HDL-C) levels were significantly inversely associated with BMI, as expected.
Markers of inflammation were also significantly associated with BMI. In this proof-of-
concept analysis of the lab measurements, the associations with BMI support the associa-
tion with the disease outcomes reported in the PheWAS.
Figure 4. A laboratory-wide association between mean body mass index and 24 laboratory meas-
urements derived from the electronic health records.
3.5. Genome-Wide Association (GWAS) with BMI
As a proof-of-concept, we conducted a GWAS for median BMI within five genetically
inferred ancestry groups. These included 30,360 EUR, 11,300 AFR, 711 AMR, 680 EAS, and
573 SAS individuals in the PMBB (Table 1). The analysis tested the association of ~7.6 mil-
lion SNPs with MAF > 1%, imputation R2 > 0.3, using a linear mixed model implemented
in REGNIE. All the models were adjusted with age, sex, and the first six ancestry specific
principal components to account for population stratification. We then conducted cross-
ancestry meta-analysis by integrating GWAS summary statistics from each ancestry
J. Pers. Med. 2022, 12, 1974 11 of 16
group using PLINK. Our meta-analysis identified 201 genome-wide significant SNP asso-
ciations with BMI (p < 5 × 10
−08
, Figure 5), replicating several previously reported associa-
tions in published GWAS of BMI. The strongest association in our PMBB analysis was
with FTO variant rs55872725 (p = 4.7 × 10
−28
, beta = 0.271), which has been previously re-
ported. Other known associations included rs7559547 (p = 3.92 × 10
−14
, beta = 0.41,
TMEM18) and rs539515 (p = 9.02 × 10
−11
, beta = 0.36, SEC16B), among others. Summary
statistics of results with p < 1 × 10
−04
are available in Supplementary Table S3.
Figure 5. Manhattan plot showing association between common genetic variants (MAF > 1%) and
BMI.
4. Discussion
The Penn Medicine BioBank was created to harness clinical data and biospecimens
for biomedical research within Penn Medicine, a large academic healthcare system.
Within a decade, it has emerged as a critical resource for translational medicine that has
fueled discovery science and facilitated precision medicine, empowering a genomics-en-
abled learning healthcare system. As of September 2022, the PMBB had enrolled over
174,000 participants, obtained biospecimens on >70,000 participants, and generated ge-
nomic data on ~44,000 of its participants. The PMBB intends to enroll >1 million partici-
pants, obtain biospecimens on >90% of participants, and generate genomic data on all par-
ticipants for whom biospecimens have been obtained.
The 2015 Institute of Medicine (now National Academy of Medicine) report on Trans-
lating Genomic-Based Research for Health [1] advocated for the development of ‘ge-
nomics-enabled learning healthcare systems’ based on the systematic summarized collec-
tion and use of genomic data, integrated with phenotypic data, to make discoveries and
enhance healthcare in large healthcare systems. More recently, in its strategic vision for
genomics research and application of genomics to clinical care, the National Human Ge-
nome Research Institute (NHGRI) emphasized the design and implementation of ge-
nomics-enabled learning healthcare systems to include infrastructure, resources, and tech-
nology development for genomics; the inclusion of underrepresented and minority
groups to make genomic research more equitable; the development of multi-omics studies
to get a comprehensive view of disease biology and the progression of diseases; and build-
ing tools to implement the knowledge back into the EHR to improve healthcare [9]. Large
disease-agnostic and diverse medical biobanks at academic medical centers, such as the
PMBB, are a critical component of fulfilling this vision.
Despite recapitulating health and disease traits from structured diagnosis codes, and
the successful integration of this with genomic data [10–16], it must be acknowledged that
diagnosis codes are crude approximations of underlying biological traits. As such, the fu-
ture of EHR-empowered genomics research lies in ‘advanced phenotyping’ beyond diag-
nosis codes. These approaches include laboratory data, medication data, and other forms
of structured data, all of which are relatively straightforward to access and use for re-
search. Laboratory data, procedure codes, family history, and billing codes are all being
mapped to concepts from various vocabularies (MONDO [17], SNOMED) to develop phe-
notypic algorithms that characterize the outcome of interest. Even more exciting and in-
formative is the extraction of meaningful quantitative information from unstructured data
(e.g., clinical notes, procedure reports, imaging reports, and pathology reports) using nat-
J. Pers. Med. 2022, 12, 1974 12 of 16
ural language processing (NLP) methods. To this end, the PMBB has deployed NLP soft-
ware (Linguamatics, Cambridge, UK) to extract phenotypes from clinical notes and other
unstructured data in PMBB participants. Furthermore, there is an immense amount of
clinical imaging performed in medical centers and these images are another potential
source of phenotypic data, sometimes referred to as ‘imaging-derived phenotypes’ (IDPs).
In several ongoing PMBB efforts, deep learning and machine learning techniques are be-
ing used to translate imaging data, such as CT, MRI, and fMRI, into quantitative IDPs to
fuel new discovery.
Another approach to collecting additional phenotype and exposure data that are ab-
sent in the EHR is using participant questionnaires. During the COVID-19 pandemic, an
initial COVID-19 survey [18] was deployed to PMBB participants to collect information
on symptoms, co-morbidities, and outcomes related to COVID-19. As the pandemic has
progressed, we now administer an active longitudinal survey to follow participants for
up to 18 months from their first COVID-19 diagnosis, yielding insights into long COVID.
Combining survey results with biospecimen and EHR-derived phenotypes can shed light
on factors that predict the onset of disease, refine preventative care, and optimize the clin-
ical trial design. Current efforts are focused on extending active data acquisition through
integrating mobile devices for both real-time data collection from survey questions and
biometric activity data.
A major focus of biobank research is the use of genomics to understand the genetic
architecture of health and disease and its implications for clinical care by linking phe-
nomic efforts with genomic data obtained from biospecimens. Leveraging these ap-
proaches, the PMBB has developed a robust and diverse genomics research enterprise.
Studies using PMBB data have highlighted the utility of the ‘genome-first’ approach’s util-
ity in studying rare variants at scale and identifying new associations between genes and
disease [10,11], as well as refined the range of the phenotypic presentation of individuals
carrying rare impactful variants in known disease genes [12,19].
To support equitable genomic research, a commitment to participant diversity has
been a hallmark feature of the PMBB since its inception. Seventeen percent of PMBB par-
ticipants (and 24% of those for whom biospecimens are currently available) are African
Americans or immigrants of African ancestry. This diversity has led to novel genomic and
biological insights that directly impact the health of underrepresented groups. For exam-
ple, recent work in the PMBB highlighted that hereditary amyloid transthyretin cardio-
myopathy was a common yet markedly underdiagnosed cause of heart failure among Af-
rican-American individuals [20], with many cases of the disease remaining undiagnosed
even at a tertiary medical center such as Penn Medicine. Given the availability of targeted
therapy, this finding advocates for the aggressive utilization of genomic and precision
medicine to diagnose transthyretin cardiomyopathy in this population. This ‘genome-
first’ approach is revealing an under-diagnosis of other genetic conditions. A ‘return of
actionable results’ program is underway, and the implications for the greater utilization
of genetic testing in clinical practice are clear.
The integration of genomic data into clinical practice is essential for the next genera-
tion of healthcare. Penn Medicine is at the forefront of developing techniques to provide
high-quality patient care based on real-world evidence and genomic discoveries [21,22].
An analysis of pharmacogenetic (PGx) variants in the PMBB concluded that anticipatory
genotyping can efficiently lead to the effective communication of PGx results to patients
[23]. Polygenic risk scores (PRS) have been posited as a novel approach to leverage com-
mon-variant genetics for clinical care to predict the genetic risk for complex diseases, alt-
hough the clinical utility of this approach has yet to be fully determined [24,25]. Research-
ers utilizing PMBB data reported that PRS for psychiatric disorders [26] and substance use
disorders [27] has shown cross-trait associations beyond traditional diagnostic bounda-
ries, suggesting broad effects of genetic liability for these disorders. Furthermore, PRS in
J. Pers. Med. 2022, 12, 1974 13 of 16
PMBB participants significantly increased the ability to identify the cancer status of Euro-
pean individuals but not African Americans, underscoring the need for large-scale ge-
nomic studies on non-white populations [13].
A critical feature of the PMBB is the availability of stored plasma and serum for bi-
omarker analyses and integration with clinical and genomic data. Although the effort and
expense of generating and storing plasma/serum aliquots are substantial, the benefit of
this approach is rapidly becoming apparent. Multiple investigators have utilized the ge-
nomic data to identify PMBB participants with genotypes of interest to then use stored
samples in cases and controls to measure biomarkers of interest. During the COVID-19
pandemic, access to stored samples from PMBB participants enrolled pre-pandemic who
subsequently developed COVID-19 permitted investigators to address a number of im-
portant research questions [28–31]. Now, with over seven years of median follow-up data
since the recruitment of PMBB patient participants, the stored samples are increasingly
precious for their use in assaying biomarkers that may be predictive of incident disease.
As methods for large-scale proteomics and metabolomics improve and the costs come
down, the opportunity to generate plasma-based large-scale omics and integrate with ge-
nomics and clinical data is increasingly feasible and promises to further enhance biologi-
cal discovery and precision medicine.
All PMBB participants consent to be recontacted, a critical feature of the protocol that
is useful for several purposes. Patient-reported surveys represent an important addition
to the EHR data for certain phenotypes as well as lifestyle and exposure data. Permission
to recontact facilitates ‘recall-by-genotype’ deep phenotyping studies, which represents a
tremendous opportunity for ‘genome-first’ discovery. Several investigators are actively
performing studies in which the genomic data are used to identify individuals that carry
rare variants in genes of interest or have a high polygenic risk for certain conditions, and
participants are contacted to consider participation in hypothesis-driven deep phenotyp-
ing studies. Deep phenotyping can include targeted imaging, immunological profiling,
provocative testing (e.g., oral glucose or fat tolerance test), creation of induced pluripotent
stem cells (iPSCs), or any number of other clinical phenotyping approaches driven by the
specific scientific question. Finally, the era of precision medicine will certainly include
many clinical trials that are targeted to individuals of a specific inherited genotype; large
medical biobanks with pre-existing genomic data, such as the PMBB, offer a fertile oppor-
tunity for the recruitment of individuals for such genotype-directed clinical trials.
5. Conclusions
The PMBB is a disease-agnostic institutional biobank under a single umbrella proto-
col based at a large academic health system with the purpose of promoting a genomics-
enabled learning healthcare system to fuel scientific discovery, translational science, and
precision medicine. A comprehensive biobank of DNA, plasma, and serum on all partici-
pants with selected other specimens and tissues on a subset of participants is linked to
rich EHR clinical data, imaging, and survey data. The clinical database is standardized to
OMOP and contains demographic, diagnoses (e.g., ICD-9/ICD-10 codes), procedures (e.g.,
Current Procedure Terminology—CPT codes), laboratory data, medication data, encoun-
ter types, socioeconomic factors, and survey data. The initiation of e-consenting has led to
a substantial increase in the rate of enrollment. As of September 2022, genome-wide ge-
nomic data have been generated on ~44,000 participants and plasma multi-omics data on
several thousand participants. The substantial representation of African-American patient
participants in the PMBB addresses the urgent need to increase diversity in human genetic
studies. Researchers with approved IRB protocols can request access to biobank samples
and data through a data access portal. Publications supported by PMBB data and speci-
mens can be found here: https://pmbb.med.upenn.edu/pmbb/publications.html (accessed
on 20 November 2022). The PMBB is one of several large medical biobanks at academic
medical centers in the US and is strongly supportive of the creation of a ‘medical biobank
consortium’ to facilitate replication, increase power for rare phenotypes and variants, and
J. Pers. Med. 2022, 12, 1974 14 of 16
promote harmonized collaboration around genotype-directed deep phenotyping and re-
cruitment into clinical trials.
Supplementary Materials: The following supporting information can be downloaded at:
https://www.mdpi.com/article/10.3390/jpm12121974/s1, Table S1: Phenome-wide association study
between body mass index (mean values) and EHR-derived phecodes. Table S2: Laboratory-wide
association study between body mass index (mean values) and clinical labs from the EHR. Table S3:
Summary statistics for genome-wide association study for BMI. Supplemental document: A full list
of PMBB author contributions.
Author Contributions: Conceptualization, D.J.R., M.D.R., A.V., N.N., S.M.D. (Scott M. Damrauer),
M.F. and K.L.N.; methodology, A.V., N.N., D.J.R., J.W., M.F., K.L.N. and M.D.R.; validation: A.V.,
C.M.K. and L.G.; software A.V. and S.M.D. (Scott M. Dudek); formal analysis, A.V., N.N., C.M.K.
and L.G.; data curation, N.N., C.M.K., L.G. and A.L.; writing—original draft preparation, J.W., A.V.,
N.N., D.J.R. and M.D.R.; writing—review and editing, S.M.D. (Scott M. Damrauer), S.S.V., G.S.,
R.L.K., T.G.D., S.M.D. (Scott M. Dudek), A.L., Y.B., R.J., E.M., K.L.N. and M.F.; visualization, N.N.,
C.M.K. and L.G.; supervision, D.J.R. and M.D.R.; project administration, J.W.; funding acquisition,
D.J.R. All authors have read and agreed to the published version of the manuscript.
Funding: The PMBB is supported by Perelman School of Medicine at University of Pennsylvania, a
gift from the Smilow family, and the National Center for Advancing Translational Sciences of the
National Institutes of Health under CTSA award number UL1TR001878. KLN is supported by the
Basser Center for BRCA.
Institutional Review Board Statement: The study was conducted in accordance with the Declara-
tion of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of Univer-
sity of Pennsylvania (protocol codes: 808346 approved 07/01/2008, 813913 approved 4/3/2013, and
817977 approved 6/6/2013” for studies involving humans.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the
study.
Data Availability Statement: All the data used to generate the figures were made available in sup-
plementary.
Acknowledgments: We thank the patient-participants of Penn Medicine who consented to partici-
pate in this research program. We acknowledge the efforts of the PMBB staff (a full list of contribu-
tors is provided in the supplement). We thank the outstanding Penn Medicine Corporate IS team
(Jessica Chen, Christine Vanzandbergen, Jeffrey Landgraf, Colin Wollack, Ned Haubein) for its ma-
jor efforts to implement e-consenting in the EHR as well as biospecimen acquisition and tracking.
We thank the Regeneron Genetics Center for partnership in generating genetic variant data and for
scientific interactions. We thank the Smilow family for their generous gift that made the launch of
the PMBB possible. Finally, we would like to thank Kevin Mahoney, Jon Epstein, Larry Jameson
(Penn Medicine leadership), Michael Parmacek (Department of Medicine), Garret FitzGerald (Insti-
tute for Translational Medicine and Therapeutics), and David Roth (Penn Center for Precision Med-
icine) for their vision and support.
Conflicts of Interest: S.M.D. receives research funding from RenalytixAI, in-kind research support
from Novo Nordisk, and personal consulting fees from Calico Labs. D.J.R. serves on scientific advi-
sory boards for Alnylam, Novartis, Pfizer, and Verve. Regeneron has generated genomic data in
PMBB participants. These entities had no role in the design of the study; in the collection, analyses,
or interpretation of data; in the writing of this manuscript; or in the decision to publish these results.
References
1. Medicine I. of Genomics-Enabled Learning Health Care Systems: Gathering and Using Genomic Information to Improve Patient Care and
Research: Workshop Summary; National Academies Press (US): Washington, DC, USA, 2015; ISBN 978-0-309-37112-4.
2. Loh, P.-R.; Danecek, P.; Palamara, P.F.; Fuchsberger, C.; AReshef, Y.; Finucane, H.K.; Schoenherr, S.; Forer, L.; McCarthy, S.;
Abecasis, G.R.; et al. Reference-based phasing using the Haplotype Reference Consortium panel. Nat. Genet. 2016, 48, 1443–1448.
https://doi.org/10.1038/ng.3679
3. Das, S.; Forer, L.; Schönherr, S.; Sidore, C.; Locke, A.E.; Kwong, A.; Vrieze, S.I.; Chew, E.Y.; Levy, S.; McGue, M.; et al. Next-
generation genotype imputation service and methods. Nat Genet. 2016, 48, 1284–1287. https://doi.org/10.1038/ng.3656
4. Price, A.L.; Patterson, N.J.; Plenge, R.M.; Weinblatt, M.E.; Shadick, N.A.; Reich, D. Principal components analysis corrects for
stratification in genome-wide association studies. Nat. Genet. 2006, 38, 904–909. https://doi.org/10.1038/ng1847
J. Pers. Med. 2022, 12, 1974 15 of 16
5. Klann, J.G.; Joss, M.A.H.; Embree, K.; Murphy, S.N. Data model harmonization for the All Of Us Research Program:
Transforming i2b2 data into the OMOP common data model. PLoS ONE 2019, 14, e0212463.
https://doi.org/10.1371/journal.pone.0212463.
6. McDonald, C.J.; Huff, S.M.; Suico, J.G.; Hill, G.; Leavelle, D.; Aller, R.; Forrey, A.; Mercer, K.; DeMoor, G.; Hook, J.; et al. LOINC,
a Universal Standard for Identifying Laboratory Observations: A 5-Year Update. Clin. Chem. 2003, 49, 624–633.
https://doi.org/10.1373/49.4.624.
7. Nelson, S.J.; Zeng, K.; Kilbourne, J.; Powell, T.; Moore, R. Normalized names for clinical drugs: RxNorm at 6 years. J. Am. Med.
Inform. Assoc. 2011, 18, 441–448. https://doi.org/10.1136/amiajnl-2011-000116.
8. Wu, P.; Gifford, A.; Meng, X.; Li, X.; Campbell, H.; Varley, T.; Zhao, J.; Carroll, R.; Bastarache, L.; Denny, J.C.; et al. Mapping
ICD-10 and ICD-10-CM Codes to Phecodes: Workflow Development and Initial Evaluation. JMIR Med. Inform. 2019, 7, e14325.
https://doi.org/10.2196/14325.
9. Green, E.D.; Gunter, C.; Biesecker, L.G.; Di Francesco, V.; Easter, C.L.; Feingold, E.A.; Felsenfeld, A.L.; Kaufman, D.J.; Ostrander,
E.A.; Pavan, W.J.; et al. Strategic vision for improving human health at The Forefront of Genomics. Nature 2020, 586, 683–692.
https://doi.org/10.1038/s41586-020-2817-4.
10. Park, J.; Levin, M.G.; Haggerty, C.M.; Hartzel, D.N.; Judy, R.; Kember, R.L.; Reza, N.; Ritchie, M.D.; Owens, A.T.; Damrauer,
S.M.; et al. A genome-first approach to aggregating rare genetic variants in LMNA for association with electronic health record
phenotypes. Genet. Med. 2020, 22, 102–111. https://doi.org/10.1038/s41436-019-0625-8.
11. Park, J.; Packard, E.A.; Levin, M.G.; Judy, R.L.; Regeneron Genetics Center; Damrauer, S.M.; Day, S.M.; Ritchie, M.D.; Rader,
D.J. A genome-first approach to rare variants in hypertrophic cardiomyopathy genes MYBPC3 and MYH7 in a medical biobank.
Hum. Mol. Genet. 2022, 31, 827–837. https://doi.org/10.1093/hmg/ddab249.
12. Damrauer, S.M.; Hardie, K.; Kember, R.L.; Judy, R.; Birtwell, D.; Williams, H.; Rader, D.J.; Pyeritz, R.E. FBN1 Coding Variants
and Nonsyndromic Aortic Disease. Circ. Genom. Precis. Med. 2019, 12, e002454. https://doi.org/10.1161/CIRCGEN.119.002454.
13. Wang, L.; Desai, H.; Verma, S.S.; Le, A.; Hausler, R.; Verma, A.; Judy, R.; Doucette, A.; Gabriel, P.E.; Nathanson, K.L.; et al.
Performance of polygenic risk scores for cancer prediction in a racially diverse academic biobank. Genet. Med. 2022, 24, 601–609.
https://doi.org/10.1016/j.gim.2021.10.015.
14. Kember, R.L.; Levin, M.G.; Cousminer, D.L.; Tsao, N.; Judy, R.; Schur, G.M.; Lubitz, S.A.; Ellinor, P.T.; McCormack, S.E.; Grant,
S.F.A.; et al. Genetically Determined Birthweight Associates With Atrial Fibrillation: A Mendelian Randomization Study. Circ.
Genom. Precis. Med. 2020, 13, e002553. https://doi.org/10.1161/CIRCGEN.119.002553.
15. Zhang, C.; Verma, A.; Feng, Y.; Melo, M.C.R.; McQuillan, M.; Hansen, M.; Lucas, A.; Park, J.; Ranciaro, A.; Thompson, S.; et al.
Impact of natural selection on global patterns of genetic variation and association with clinical phenotypes at genes involved in
SARS-CoV-2 infection. Proc. Natl. Acad. Sci. USA 2022, 119, e2123000119. https://doi.org/10.1073/pnas.2123000119.
16. Bajaj, A.; Ihegword, A.; Qiu, C.; Small, A.M.; Wei, W.-Q.; Bastarache, L.; Feng, Q.; Kember, R.L.; Risman, M.; Bloom, R.D.; et al.
Phenome-wide association analysis suggests the APOL1 linked disease spectrum primarily drives kidney-specific pathways.
Kidney Int. 2020, 97, 1032–1041. https://doi.org/10.1016/j.kint.2020.01.027.
17. Shefchek, K.A.; Harris, N.L.; Gargano, M.; Matentzoglu, N.; Unni, D.; Brush, M.; Keith, D.; Conlin, T.; Vasilevsky, N.; Zhang,
X.A.; et al. The Monarch Initiative in 2019: An integrative data and analytic platform connecting phenotypes to genotypes across
species. Nucleic Acids Res. 2020, 48, D704–D715. https://doi.org/10.1093/nar/gkz997.
18. Verma, S.S.; Chung, W.K.; Dudek, S.; Williamson, J.L.; Verma, A.; Robinson, S.; Rader, D.J.; Reilly, M.P.; Sengupta, S.; FitzGerald,
G.A.; et al. Research on COVID-19 through patient-reported data: A survey for observational studies in the COVID-19 pandemic.
J. Clin. Transl. Sci. 2021, 5, e17. https://doi.org/10.1017/cts.2020.509.
19. Drivas, T.G.; Lucas, A.; Zhang, X.; Ritchie, M.D. Mendelian pathway analysis of laboratory traits reveals distinct roles for ciliary
subcompartments in common disease pathogenesis. Am. J. Hum. Genet. 2021, 108, 482–501.
https://doi.org/10.1016/j.ajhg.2021.02.008.
20. Damrauer, S.M.; Chaudhary, K.; Cho, J.H.; Liang, L.W.; Argulian, E.; Chan, L.; Dobbyn, A.; Guerraty, M.A.; Judy, R.; Kay, J.; et
al. Association of the V122I Hereditary Transthyretin Amyloidosis Genetic Variant With Heart Failure Among Individuals of
African or Hispanic/Latino Ancestry. JAMA 2019, 322, 2191. https://doi.org/10.1001/jama.2019.17935.
21. Lau-Min, K.S.; Asher, S.B.; Chen, J.; Domchek, S.M.; Feldman, M.; Joffe, S.; Landgraf, J.; Speare, V.; Varughese, L.A.; Tuteja, S.;
et al. Real-world integration of genomic data into the electronic health record: The PennChart Genomics Initiative. Genet. Med.
Off. J. Am. Coll. Med. Genet. 2021, 23, 603–605. https://doi.org/10.1038/s41436-020-01056-y.
22. Lau-Min, K.S.; McKenna, D.; Asher, S.B.; Bardakjian, T.; Wollack, C.; Bleznuck, J.; Biros, D.; Anantharajah, A.; Clark, D.F.; Condit,
C.; et al. Impact of integrating genomic data into the electronic health record on genetics care delivery. Genet. Med. 2022, 24,
2338–2350. https://doi.org/10.1016/j.gim.2022.08.009.
23. Verma, S.S.; Keat, K.; Li, B.; Hoffecker, G.; Risman, M.; Regeneron Genetics Center; Sangkuhl, K.; Whirl-Carrillo, M.; Dudek, S.;
Verma, A.; et al. Evaluating the frequency and the impact of pharmacogenetic alleles in an ancestrally diverse Biobank
population. medRxiv 2022.08.26.22279261; doi: https://doi.org/10.1101/2022.08.26.22279261
24. Martin, A.R.; Kanai, M.; Kamatani, Y.; Okada, Y.; Neale, B.M.; Daly, M.J. Clinical use of current polygenic risk scores may
exacerbate health disparities. Nat. Genet. 2019, 51, 584–591. https://doi.org/10.1038/s41588-019-0379-x.
25. Sirugo, G.; Williams, S.M.; Tishkoff, S.A. The Missing Diversity in Human Genetic Studies. Cell 2019, 177, 26–31.
https://doi.org/10.1016/j.cell.2019.02.048.
J. Pers. Med. 2022, 12, 1974 16 of 16
26. Kember, R.L.; Merikangas, A.K.; Verma, S.S.; Verma, A.; Judy, R.; Regeneron Genetics Center; Damrauer, S.M.; Ritchie, M.D.;
Rader, D.J.; Bućan, M. Polygenic Risk of Psychiatric Disorders Exhibits Cross-trait Associations in Electronic Health Record
Data From European Ancestry Individuals. Biol. Psychiatry 2021, 89, 236–245. https://doi.org/10.1016/j.biopsych.2020.06.026.
27. Hartwell, E.E.; Merikangas, A.K.; Verma, S.S.; Ritchie, M.D.; Regeneron Genetics Center; Kranzler, H.R.; Kember, R.L. Genetic
liability for substance use associated with medical comorbidities in electronic health records of African- and European-ancestry
individuals. Addict. Biol. 2022, 27, e13099. https://doi.org/10.1111/adb.13099.
28. Alanio, C.; Verma, A.; Mathew, D.; Gouma, S.; Liang, G.; Dunn, T.; Oldridge, D.A.; Weaver, J.; Kuri-Cervantes, L.; Pampena,
M.B.; et al. Cytomegalovirus Latent Infection is Associated with an Increased Risk of COVID-19-Related Hospitalization. J. Infect.
Dis. 2022, 226, 463–473. https://doi.org/10.1093/infdis/jiac020.
29. Banday, A.R.; Stanifer, M.L.; Florez-Vargas, O.; Onabajo, O.O.; Papenberg, B.W.; Zahoor, M.A.; Mirabello, L.; Ring, T.J.; Lee, C.-
H.; Albert, P.S.; et al. Genetic regulation of OAS1 nonsense-mediated decay underlies association with COVID-19
hospitalization in patients of European and African ancestries. Nat. Genet. 2022, 54, 1103–1116. https://doi.org/10.1038/s41588-
022-01113-z.
30. Anderson, E.M.; Goodwin, E.C.; Verma, A.; Arevalo, C.P.; Bolton, M.J.; Weirick, M.E.; Gouma, S.; McAllister, C.M.; Christensen,
S.R.; Weaver, J.; et al. Seasonal human coronavirus antibodies are boosted upon SARS-CoV-2 infection but not associated with
protection. Cell 2021, 184, 1858–1864.e10. https://doi.org/10.1016/j.cell.2021.02.010.
31. Flannery, D.D.; Gouma, S.; Dhudasia, M.B.; Mukhopadhyay, S.; Pfeifer, M.R.; Woodford, E.C.; Gerber, J.S.; Arevalo, C.P.; Bolton,
M.J.; Weirick, M.E.; et al. SARS-CoV-2 seroprevalence among parturient women in Philadelphia. Sci. Immunol. 2020, 5, eabd5709.
https://doi.org/10.1126/sciimmunol.abd5709.
