The Components of GP2’s Sixth Data Release
Author(s)
-
Co-lead, Complex Disease Data Analysis
Hampton Leonard, MS
Data Tecnica International / National Institutes of Health | USA
Hampton has a background in data science and machine learning, which she applies to large multi-omic datasets in the neurodegenerative disease space. She is passionate about investigating differences on both clinical and omic levels and how these differences can affect clinical trial outcomes.
-
Lead, Complex Disease Data Analysis
Mike A. Nalls, PhD
Data Tecnica International | USA
Mike founded Data Tecnica in early 2017 after over a decade of experience in large dataset analytics and methods research in healthcare and other scientific fields. 400+ peer-reviewed publications in the field of applied statistics in large datasets, brain diseases, and genomics. He is a strong advocate of open science, collaboration, and transparency in science.
-
Working Group Participant
Dan Vitale, MS
Data Tecnica International | USA
Dan is a data science consultant for Data Tecnica, consulting primarily for the Laboratory of Neurogenetics and CARD at the National Institute on Aging of the National Institutes of Health. His work is focused on open science, automation and development of genetic analytic pipelines and software, and machine learning.
-
Working Group Participant
Mathew Koretsky, BSc
National Institutes of Health | USA
Mat is a post-baccalaureate student at the National Institutes of Health. He is passionate about pipeline development and meaningful applications of computer science in the biomedical research space.
-
Working Group Participant
Kristin Levine, MS
Data Tecnica International | USA
Kristin works with the Data Tecnica and National Institute on Aging (NIA) teams on data and code sharing plus real-world data analysis of biobanks and healthcare systems. She is also an accomplished writer, now applying her communication skills to scientific domains.
-
Co-lead, Data and Code Dissemination
Mary B. Makarious, BSc
National Institutes of Health | USA
Mary is a graduate student participating in the NIH graduate partnership program in collaboration with the University College London. She is a rising star in biomedical data science, with a background in genomics, machine learning, and open science platforms. She is also passionate about increasing representation in research and empowering scientists to analyze their own data.
-
Lietsel Jones
DataTecnica/National Institutes of Health | USA
Lietsel is an analyst with Data Tecnica with a keen interest in the intersection between epidemiology and genetics. She is also a clinical data manager with GP2 working to collect and harmonize large clinical datasets from worldwide contributors.
-
Lead, Monogenic Data Analysis
Zih-Hua Fang, PhD
German Center for Neurodegenerative Diseases | Germany
The lead of the monogenic data analysis efforts in GP2, they are making significant contributions to GP2’s efforts to study monogenic and familial Parkinson’s disease.
-
Co-lead, Operations and Compliance
J Solle, MBA
The Michael J. Fox Foundation | USA
J is the implementation Program Lead for GP2, co-lead for the Operations & Compliance Working Group, and a member of the Operations Committee. J joined the Michael J. Fox Foundation in March 2021 and is the Director of Clinical Research, leading the implementation of GP2.
Overview
In December 2023, GP2 announced the sixth data release on the Terra and the Verily® Workbench platforms in collaboration with AMP® PD. This release includes >20,000 additional participants, adding to the previous releases from the Complex and Monogenic Networks.
- The complex disease data (genotypes), including locally-restricted samples, now consists of a total of 44,831 genotyped participants (24,709 PD cases, 17,246 Controls, and 2,876 ‘Other’ phenotypes)
- When removing the locally-restricted samples, these now consist of 33,436 (17,129 PD cases, 13,872 Controls, and 2,435 ‘Other’ phenotypes)
- The monogenic disease data (whole genome sequences) now consists of a total of 2,324 sequenced participants (1,854 PD cases, 314 Controls, and 156 ‘Other’ phenotypes)
- When removing the locally-restricted samples, these now consist of 2,083 (1,650 PD cases, 309 Controls, and 124 ‘Other’ phenotypes)
- 12,585 individuals who have deep clinical phenotyping information also have matching genetic information
What’s New In This Release?
- Additional complex disease (genotyped) and monogenic disease (whole genome) samples
- Introducing locally-restricted GDPR samples via the Verily Viewpoint Workbench
- Introducing clinical data for ~12,000 individuals
- Introducing a new ancestry group → Complex Admixture History (CAH)
- Updates in quality control measures for released genotyping data
- Updates in variant calling, now with DeepVariant, for released whole genome data
Updates in Quality Control
In summary, GenoTools (v1.0.0) performs the following quality control steps: sex mismatches, call rate pruning, checking for duplicates, checking and reporting related individuals, and a heterozygosity rate check.
In contrast to previous releases, we no longer perform the following variant-level filtering: by minor allele frequency (MAF), Hardy-Weinberg (HWE), or minor allele count (MAC). If you would like to filter how we have done for previous releases, we recommend consulting the corresponding README for detailed information and suggested thresholds.
Complex Admixture History (CAH)
CAH, or Complex Admixture History, is a new ancestry group introduced to GP2 for release 6. It was created in response to a large number of samples with South African and other highly admixed individuals being incorrectly predicted as CAS (Central Asian) ancestry in release 5. For release 6, the CAH ancestry group mainly contains samples from Stellenbosch University (Cape Town, South Africa), The Coriell Institute (Camden, New Jersey, United States), and the Parkinson’s Foundation (Miami, Florida, United States). We consider any samples labeled as CAH to be too highly admixed to be included in analyses with other GP2 ancestry groups.
Locality-restricted GDPR samples via the Verily Viewpoint Workbench
We are excited to announce that some users will be able to access locally-restricted samples, otherwise known as samples governed by the General Data Protection Regulation (GDPR) policy, through our collaboration with the Verily Viewpoint Workbench. Workbench is a secure environment for governing and analyzing biomedical data, aimed at enhancing research collaboration and data reproducibility via cloud integration. It supports workspace sharing, including Python and R code, and offers a suite of cloud-native services for data management and analysis. Workbench supports secure data use integration and custom authentication, making it the ideal, secure, and scalable research environment for GP2 to host its locally restricted samples.
At this time, as GP2 continues to roll out data sharing solutions for GDPR-protected data, release 6 will be available to only GP2 consortium members and partners. As testing and implementation continues in early 2024, these solutions will be available to the broader research community. All release 6 samples can be found on Workbench, meanwhile all release 6 samples not governed by GDPR requirements can be found on the community workbench on Terra (like all previous releases). To gain access to the full release on Verily Viewpoint Workbench you must:
- Have approved GP2 Tier 2 access
- Be a GP2 consortium member (contributing cohort, GP2 partner, or project analyses team member)
- Fill out the GDPR-governed sample request form. Upon completing the form, you’ll receive follow up instructions for accessing the Workbench.
Clinical Data
We are also really excited to announce comprehensive deep clinical phenotyping data for 12,585 individuals in this release. This information consists of:
- Age at diagnosis and onset
- Primary, current, and latest diagnoses
- Cognitive exams such as the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA)
- Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS)
- Detailed “other” phenotypes, such as Lewy Body Dementia (LBD)
In this release, each of the 12,585 individuals who have clinical information also have matching genetic information.
Individual-Level Data
We now capture the data from a total of 74 cohorts, 46 are new to this release. Please refer to the GP2 Cohort Dashboard for more information on the cohorts that have been shared.
Genetically-determined ancestry of complex disease GP2 participants is broken into 11 ancestry groups. The table below details the genetically-determined ancestry of complex disease participants in this release that have passed quality control and been imputed. These numbers include samples from previous releases that have been reclustered using the new cluster file and gone through quality control along with the newly genotyped and shared samples unique to this current release.
Variant Calling Whole Genome Sequences using DeepVariant-GLnexus
In this release, in contrast to previous releases, we now use Google’s DeepVariant pipeline coupled with GLnexus for cohort-level variant calling. DeepVariant is a deep learning-based variant caller that outperforms existing state-of-the-art tools by accurately calling individual-level genetic variants. It also simplifies the process, enhancing accuracy and reliability.
Future data releases will continue to grow the diversity of participants available. You can check out our dashboard to see our progress. For users with tier 2 access already, you can explore the data further on our cohort browser, expanded on in a previous blog post.
As always, please refer to the README that accompanies each GP2 release for further details regarding pipelines, data, and analyses!
