Article | Open | Published:

Self-report data as a tool for subtype identification in genetically-defined Parkinson’s Disease

Scientific Reportsvolume 8, Article number: 12992 (2018) | Download Citation

Abstract

Through a targeted recruitment 23andMe has collected DNA and patient-reported symptoms from more than 10,000 subjects reporting a physician-verified diagnosis of PD. This study evaluated the potential of self-report, web-based questionnaires to rapidly assess disease natural history and symptomology in genetically-defined PD populations. While average age-at-diagnosis was significantly lower in GBA mutation carriers compared to idiopathic PD, or iPD (idiopathic PD, defined as no GBA mutations and no LRRK2 G2019S mutation), there were no significant differences in symptoms. Conversely, LRRK2 G2019S carrier status significantly associated with reporting of milder daily symptoms of lightheadedness and several differences were observed at a false discovery rate < 0.1, including increased reporting of changes in walking as an initial symptom of disease, decreased reporting of lightheadedness upon standing, and milder symptoms related to daily functioning. The subclinical differences in symptoms reported by LRRK2 G2019S carriers suggest differences in underlying pathophysiology and/or disease progression in LRRK2 carriers compared to iPD. Importantly, we confirm previous findings in PD genetic subsets where disease characteristics were ascertained through clinical exam. Overall, these data support the effective use of self-report and genetic data to rapidly analyze information from a large disease population or difficult to identify genetic subgroups.

Introduction

While recent GWAS have identified more than 40 loci associated with Parkinson’s disease (PD) susceptibility1,2, most of our current biological understanding of the disease stems from research on rare mutations associated with PD. Two such genes, LRRK2 (leucine rich repeat kinase 2) and GBA (glucosidase beta acid, glucocerebrosidase), are recognized risk loci and known causes of autosomal dominant PD. Identified through genome-wide association studies (GWAS) as well as linkage studies of familial PD, mutations in these genes are not fully penetrant and increase risk of PD in an age-dependent manner2,3,4,5,6,7,8. The use of GBA and LRRK2 mutations to model pathobiology in research and drug discovery has enabled researchers to gain increased understanding of cellular processes compromised by these mutations. The hope is that these same cellular processes are the basis of disease for all forms of Parkinson’s disease. Despite this progress, the utility of rare, single gene mutations to model complex disorders has been called into question. Previous studies of GBA and LRRK2 mutation carriers suggest that the manifestation of PD in these populations may differ from those with no known genetic cause, or idiopathic PD (iPD). Earlier age-at-onset is reported for both groups, while differences in cognition and depression rates are common in the GBA mutation carriers with PD, and gait dysfunction and postural instability, as well as, a more mild disease progression may be more common in the LRRK2 mutation carriers than that seen in iPD6,9,10,11,12. The data from these mutaitons suggest distinct differences in the underlying disease pathobiology, calling into question their ability to accurately model more common forms of PD.

GBA mutations are implicated in roughly 3–20% (including estimates from general disease population, ethnically-enriched, and familial studies) of iPD populations through a series of coding mutations believed to cause a loss of protein function7,12,13,14. The presence of biallelic GBA mutations results in the lysosomal storage disorder Gaucher’s disease, while the heterozygous state significantly increases risk for developing PD. LRRK2 mutations are far less common, ~1% in the sporadic PD population, and thought to cause disease through a gain-of-function resulting in increased kinase activity15. G2019S is the most commonly PD- associated LRRK2 mutation. Although many pathogenic mutations have been identified in both genes, the mutations have incomplete penetrance as evidenced by healthy mutation carriers and the identification of these two genes as risk factors for PD in studies of idiopathic PD. Incomplete penetrance of GBA and LRRK2 mutations can obscure the heritable nature of these mutations resulting in their contribution to sporadic cases of PD. Gaining a better clinical and genetic understanding of carriers of LRRK2 and GBA mutations will help elucidate similarities in pathogenic events in PD as well as identify disease subtypes that may have distinct biology or distinct clinical manifestation.

Gathering data to adequately power genetic association studies from disease subpopulations or genetically-defined groups can be difficult due to geographical limitations and the cost and accessibility of clinical assessments. In this study we present novel findings from a study of patient-reported symptoms related to Parkinson’s disease in the 23andMe PD research database. We utilize these data to test for association of specific genetic subgroups of PD, namely GBA and LRRK2 pathogenic mutation carriers, with specific disease symptomology, age-at-onset, and differences in medication use. Our findings support the use of patient-reported data and large genetic databases to rapidly test for novel genotype-phenotype associations and are supported by previous studies of clinically-ascertained phenotypes in GBA and LRRK2 carriers. Ultimately, these findings can be used to inform clinical understanding as well as clinical study design, as these endpoints may be more sensitive to the targets of new compounds in development for PD. Also, our results further support the importance and potential power of Patient Centered Outcomes.

Results

Through a coordinated recruitment, 23andMe (in collaboration with the Michael J. Fox Foundation, the National Parkinson Foundation, and The Parkinson’s Institute and Clinical Center) has recruited more than 10,000 PD participants into their research-focused PD Community cohort, which collects both genetic information and online questionnaire-based assessments. Subjects recruited through this process have reported a physician diagnosis of PD and completed questionnaires that range from normal daily functioning to disease symptoms and severity.

GBA mutation carriers have an earlier age of disease onset

To better understand if carriers of GBA and LRRK2 mutations with PD represent a specific disease subtype, we evaluated differences in disease natural history and manifestation in three genetically-defined groups: LRRK2 G2019S carriers, carriers of 17 mutations in GBA (Table 1), and iPD (defined in this instance as the PD population not carrying any of the before mentioned GBA or LRRK2 variants). From the PD Community, 6,894 individuals completed detailed surveys assessing disease onset and current manifestation and a total of 6,883 were included in our analysis, we did not include individuals carrying both GBA and LRRK2 mutations. We tested for differences in the association of each genetically-defined group with the following conditions: age-at-diagnosis, sex distribution, first symptoms of disease, general symptoms since onset and severity of symptoms, daily-living/functioning (similar to the Unified Parkinson’s Disease Rating Scale (UPDRS) II and I), medications, and the presence of comorbidities. The distribution for sex and current age is shown in Table 2. GBA mutation carriers and idiopathic PD groups had a higher percentage of males reporting a diagnosis of PD than females, while LRRK2 G2019S carriers showed no sex effect, consistent with a recent meta-analysis16. While the current age-at-questionnaire distribution is similar between the iPD and GBA groups, there are more GBA mutation carriers under 60 years old compared to the iPD and LRRK2 groups, consistent with an earlier age-of-disease-onset. There is no significant difference in the occurrence of type 2 diabetes, ulcerative colitis or Crohn’s disease, or any autoimmune diseases between groups (Supplementary Table 1).

Table 1 LRRK2 and GBA mutations used to define LRRK2 and GBA mutation carrier groups, respectively.
Table 2 Demographics of the PD Community.

Compared to the iPD group, average age-at-diagnosis and age-at-symptom-onset are significantly earlier for GBA mutation carriers (β = −3.53 (−4.91, −2.91) years, p = 5.10 × 10−7; and β = −3.58 (−5.09, −2.10) years, p = 2.2 × 10−6, respectively) but not significant for LRRK2 mutation carriers (FDR > 0.05 and FDR > 0.1, respectively). Prevalence of PD in G2019S carriers in the 23andMe population was 58.67% for subjects 50–60 years old and 42.86% for those 70–80 years comparing age-at-diagnosis in affected LRRK2 G2019S to age-at-survey for unaffected carriers in the PD Community and the broader 23andMe research database. The overall prevalence of PD for all GBA mutation carriers is lower, with an estimate of 17.84% in the 50–60-year-old age group and 6.98% for those between 70–80 years old. The low prevalence in this group may be affected by the inclusion of less pathogenic GBA mutations such as T369M and E326K variants12,13,17 or the previously uncited D140H or P454R, and the exclusion of L444P carriers. The L444P mutation cannot be accurately detected via standard SNP array due to homology of the GBA pseudogene (GBAP).

LRRK2 G2019S carriers present with significant differences in disease characteristics

To determine if there are differences in disease symptomology between the genetically-defined disease groups and iPD, we tested the association of each group with a broad spectrum of phenotypes relevant to motor functions and daily living in PD patients.

We tested for differences in self-report of symptom severity related to daily living for symptoms experienced during the week prior to the subject taking the survey. While the GBA mutation group showed no significant differences in symptomology compared to the iPD group, LRRK2 G2019S carriers report significantly milder symptoms of lightheadedness. In addition, several symptoms were reported as milder for LRRK2 G2019S carriers at FDR < 0.1: excess salivation, fatigue, problems eating, constipation, problems with handwriting, bladder control, and problems dressing (nominally significant) compared to the iPD group at a comparable time after symptom onset (Table 3).

Table 3 Comparison of differences in disease manifestation (characteristics) between genetically-defined disease groups and idiopathic disease.

GBA mutation carriers also showed no significant differences in first symptoms or current and past symptoms (general symptoms), while LRRK2 G2019S carrier status showed several significant differences at FDR < 0.1. These include increased likelihood of reporting changes in walking as a first symptom of disease, an increased reporting of falling as a general symptom, and a decreased likelihood to report changes in handwriting, tremor, or shaking as initial symptoms of disease. Full results with applied FDRs are shown in Table 3.

Medication usage by GBA and LRRK2 mutation carriers

As a potential indicator of disease manifestation or disease course, we tested for association of each disease group with differences in current and past medication use. No findings achieved FDR < 0.05. Of the results nominally significant (p < 0.05) without multiple test adjustment, LRRK2 mutation carrier status was associated with increased likelihood to report current use of Symmetrel (weak dopamine agonist), Sinemet CR (levodopa carbidopa controlled release), and Sinemet than the iPD population (Supplementary Table 2) and more likely to report past and/or current use of Symmetrel, Parlodel (dopamine agonist), Sinemet CR, Tasmar (COMT inhibitor), Sinemet, and Apokyn injections (dopamine agonist). Interestingly, the GBA mutation carrier status was associated with increased likelihood to currently take Sinemet controlled release (levodopa carbidopa), Requip (dopamine agonist), Cogentin (anticholinergic), and Azilect (monoamine oxidase inhibitor), but there were no differences in medication usage when looking at reporting of both past medication and current use. The differences seen in current medications used by GBA mutation carriers may be the result of an earlier disease onset and therefore a longer history of disease and medication use.

Discussion

This study evaluated the potential of web-based questionnaires to rapidly assess disease natural history and symptomology in difficult to access genetically-defined populations. By utilizing data collected from the 23andMe PD community, we were able to rapidly identify and evaluate individuals with specific DNA variants. Findings in these community-derived groups replicated previous studies from clinically-ascertained datasets.

The primary focus of this study was to evaluate the utility of a self-report dataset to identify disease subtypes of genetically-defined populations. While GBA mutation carriers had an earlier age-at-diagnosis on average, only the LRRK2 G2019S carriers presented with significant differences in motor phenotypes and daily functioning, GBA mutation carriers were indistinguishable from the iPD group. Overall, our findings suggest LRRK2 G2019S carriers represent a subtype of PD characterized by more mild general disability in daily living activities or a slower disease progression, with more changes in walking as an initial disease symptom than the general PD population. Similar findings have been reported in studies of LRRK2 G2019S carriers with Ashkenazi Jewish ancestry18, in early-onset PD11, and in iPD cohorts19, as well as reports of gait disturbances or gait variability in asymptomatic LRRK2 G2019S carriers20. Previous findings suggest differences seen in LRRK2 mutation carriers are a result of early-onset disease or are due to enrichment of unique Ashkenazi-specific genetics rather than pathogenic events specific to LRRK2. Studies done in a variety of different mutation carrier groups have controlled for these differences (early-onset disease, Ashkenazi Jewish PD, and even non-PD mutation carriers) providing further evidence that these differences are central to LRRK2 dysregulation11,18,18,20. These studies examined similar functional motor endpoints but were measured in a clinical setting and therefore are likely to be more accurate and sensitive than subjective responses collected by patient reporting. Importantly, the ability of this study to replicate these previous findings encourages the continued use and development of carefully designed patient-based questionnaires to quickly access large or genetically specified groups of subjects and demonstrates the power of online surveys in a large genotyped cohort. We also discovered a novel association of LRRK2 mutation carrier status with milder symptoms associated with lightheadedness compared to the sporadic PD population. An analysis of medication use did not produce results passing a strict threshold for multiple test correction (FDR < 0.05) but there were nominally significant results (pvalue < 0.05) that warrant future study. Importantly, future studies should expand data collection to include medication dosage as well as interventional treatment with deep brain stimulation.

A potential weakness of our study is the aggregation of GBA mutations into one group, compared to the single variant analysis conducted for LRRK2. The aggregation for GBA was necessary due to the relatively small population size for each individual mutation. This was in contrast to the G2019S LRRK2 mutation, which was sufficiently powered (n = 145 individuals with PD). While the assessment of mutations as a single group (GBA mutation group) rather than as individual mutations may introduce additional heterogeneity, we expect that if all mutations assessed are pathogenic they will likely contribute to the PD phenotype through similar molecular mechanisms (i.e. loss of function or common effect on key protein-protein interactions or pathways). Our findings are also in agreement with previous reports showing no significant difference in disease symptoms in GBA mutation carriers, except for early occurrence of dementia in this population10,21, which we did not test in our analysis. Future characterization of residual enzymatic activity associated with each mutation will allow for more refined correlation analyses between disease-relevant endpoints and enzyme activity. These analyses would further test the hypothesis that loss of glucocerebrosidase is the causative molecular event resulting from GBA pathogenic mutations in PD.

Overall, we identified phenotypic differences in the LRRK2 carriers that may reflect differences in the underlying pathophysiology of disease, suggesting that genetically-defined disease populations may be one method for identifying disease subtypes or endophenotypes within the broader spectrum of disease.

It is important for researchers to consider potential biases of human-based research in the application of these results to future research or study design. While biases were not studied in this dataset directly, research participants in human-based studies are often Caucasian, more affluent, and attend school longer than those in the general population. This can ultimately result in findings that are not applicable in a real-world setting. The participants in this study were recruited through a variety of channels including the 23andMe research database as well as a targeted email campaign in collaboration with the Michael J. Fox Foundation, The Parkinson’s Institute and Clinical Center, and numerous other PD patient groups and clinics. Given the facility of web-based surveys, future studies should strive to diversify participation and seek to conduct research that reflects the heterogeneity of the general population.

These differences have important implications for clinical study design. As the first phase of LRRK2 kinase inhibitors may soon be tested in early trials of safety and efficacy, a clear understanding of the patient population will provide better opportunities for success by assessing differences in the target population’s unique clinical manifestation and disease decline trajectory. It is logical to hypothesize that the clinical phenotypes most closely linked to the therapeutic target (i.e. phenotypes identified as early symptom presentations in genetically-defined population where that gene is targeted by investigational drugs) are the phenotypes that are most likely to respond to the therapy with early intervention. Importantly, the identification of disease subtypes does not imply that therapeutic success in this population for this target is limited to the subpopulation. The advantage of subpopulation studies is to provide an early indicator of efficacy in a more homogenous population. From this point, early biomarker studies or sub-group analyses could enable careful expansion of the therapy, from use in a genetically-defined population into non-mutation carrier groups.

Overall our data support the use of clinically-relevant patient questionnaires to quickly access genetically predefined subpopulations and evaluate disease relevant differences. Ideally, future clinical trials will test therapeutic efficacy of new treatments in prodromal PD populations and the ability of treatments to delay or prevent the onset of significant motor symptoms. As these populations grow large enough for sufficiently powered studies, evaluation of genetic markers and phenotypic traits paired with collection of biomarker data will enable early identification of prodromal patient populations for prevention studies.

Methods

Study Design and Population (23andMe PD Community)

Research participants were recruited by 23andMe, Inc., a direct-to-consumer genetics company. The 23andMe PD cohort was described in detail previously22. Briefly, participants with PD were recruited from a variety of channels, including from the existing 23andMe customer base and through external channels and media, such as a targeted email campaign in conjunction with the Michael J. Fox Foundation, the Parkinson’s Institute and Clinical Center, and numerous other PD patient groups and clinics. All research participants included in the analyses provided informed consent and answered surveys online according to the 23andMe human subjects protocol, which was reviewed and approved by Ethical & Independent Review Services, an AAHRPP-accredited, external institutional review board. Details of questionnaires, which included questions from parts I and II of the MDS UPDRS, are included in Supplemental Information.

Genotype- phenotype association analyses

We tested if genotype (carrier status of either LRRK2 G2019S or GBA mutations) associated with differences in age-at-diagnosis and age-at-symptom-onset by fitting a linear regression model and adjusting for sex and the first five principal components for genetic ancestry (see ancestry determination methods in23,24). The same adjustments were included for association of genotype with symptom severity (scale of 1–5) with the additional adjustment for age and age-at-symptom-onset, also using linear regression. Significance is determined across GBA and LRRK2 mutation carrier tests based on achieving p < 7.7 × 10−4, the Bonferroni threshold for 66 tests. False Discovery Rates (FDRs) were calculated using the Benjamini Hochberg method25 separately for the LRRK2 and GBA groups, and were assigned to each p-value in each list as the lowest FDR for which the p-value would pass from within its list. Sets of results with FDR < 0.1 are cited. Symptom questionnaires are described in detail in the Supplemental Information.

We also tested for association of genotype (carrier status of either LRRK2 G2019S or GBA mutations) with binary traits, including differences in current and former medications, first symptoms, general symptoms of disease, and comorbidities using logistic regression and adjusting for age, sex, and the first five principal components for genetic ancestry.

In Table 3, continuous traits assessed by linear regression are given effect sizes in the units of that trait, along with standard error of that effect size (e.g. Beta (SE) as denoted in the table). Binary traits assessed by logistic regression show effect sizes of Odds Ratio per allele, with an associated 95% confidence interval (denoted OR (CI) in the table).

Detailed questionnaires are included in the Supplemental Information.

Comorbidities

Three phenotypes were tested for association with the occurrence of comorbidities in three categories: any autoimmune disease, type 2 diabetes, and irritable bowel disease (including Crohn’s disease and ulcerative colitis).

Genotyping, Imputation, and Quality Control

Genotyping, imputation and quality control methods previously described26. DNA extraction and genotyping were performed on saliva samples by National Genetics Institute (NGI), a CLIA licensed clinical laboratory and subsidiary of Laboratory Corporation of America. Samples were genotyped on one of four genotyping platforms. The V1 and V2 platforms were variants of the Illumina HumanHap550 + BeadChip, including ~25,000 custom SNPs selected by 23andMe, with a total of ~560,000 SNPs. The V3 platform was based on the Illumina OmniExpress + BeadChip, with custom content to improve overlap with the V2 array, with a total of ~950,000 SNPs. The V4 platform (currently used) is a fully custom array, including a lower redundancy subset of V2 and V3 SNPs with additional coverage of lower-frequency coding variation, and ~570,000 SNPs. Samples that failed to reach 98.5% call rate were re-analyzed. Individuals whose analyses failed repeatedly were re-contacted by 23andMe customer service to provide additional samples, as is done for all 23andMe customers.

Additional information

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  1. 1.

    Nalls, M. A. et al. Large-scale meta-analysis of genome-wide association data identifies six new risk loci for Parkinson’s disease. Nature genetics 46, 989–993, https://doi.org/10.1038/ng.3043 (2014).

  2. 2.

    Chang, D. et al. A meta-analysis of genome-wide association studies identifies 17 new Parkinson’s disease risk loci. Nature genetics 49, 1511–1516, https://doi.org/10.1038/ng.3955 (2017).

  3. 3.

    Neudorfer, O. et al. Occurrence of Parkinson’s syndrome in type I Gaucher disease. QJM: monthly journal of the Association of Physicians 89, 691–694 (1996).

  4. 4.

    Tayebi, N. et al. Gaucher disease with parkinsonian manifestations: does glucocerebrosidase deficiency contribute to a vulnerability to parkinsonism? Molecular genetics and metabolism 79, 104–109 (2003).

  5. 5.

    Zimprich, A. et al. Mutations in LRRK2 cause autosomal-dominant parkinsonism with pleomorphic pathology. Neuron 44, 601–607, https://doi.org/10.1016/j.neuron.2004.11.005 (2004).

  6. 6.

    Healy, D. G. et al. Phenotype, genotype, and worldwide genetic penetrance of LRRK2-associated Parkinson’s disease: a case-control study. Lancet Neurol 7, 583–590, https://doi.org/10.1016/S1474-4422(08)70117-0 (2008).

  7. 7.

    Anheim, M. et al. Penetrance of Parkinson disease in glucocerebrosidase gene mutation carriers. Neurology 78, 417–420, https://doi.org/10.1212/WNL.0b013e318245f476 (2012).

  8. 8.

    Lee, A. J. et al. Penetrance estimate of LRRK2p.G2019S mutation in individuals of non-Ashkenazi Jewish ancestry. Movement disorders: official journal of the Movement Disorder Society 32, 1432–1438, https://doi.org/10.1002/mds.27059 (2017).

  9. 9.

    da Silva, C. P. et al. Clinical profiles associated with LRRK2 and GBA mutations in Brazilians with Parkinson’s disease. J Neurol Sci 381, 160–164, https://doi.org/10.1016/j.jns.2017.08.3249 (2017).

  10. 10.

    Alcalay, R. N. et al. Cognitive performance of GBA mutation carriers with early-onset PD: the CORE-PD study. Neurology 78, 1434–1440, https://doi.org/10.1212/WNL.0b013e318253d54b (2012).

  11. 11.

    Alcalay, R. N. et al. Motor phenotype of LRRK2 G2019S carriers in early-onset Parkinson disease. Archives of neurology 66, 1517–1522, https://doi.org/10.1001/archneurol.2009.267 (2009).

  12. 12.

    Clark, L. N. et al. Mutations in the glucocerebrosidase gene are associated with early-onset Parkinson disease. Neurology 69, 1270–1277, https://doi.org/10.1212/01.wnl.0000276989.17578.02 (2007).

  13. 13.

    Lwin, A., Orvisky, E., Goker-Alpan, O., LaMarca, M. E. & Sidransky, E. Glucocerebrosidase mutations in subjects with parkinsonism. Molecular genetics and metabolism 81, 70–73 (2004).

  14. 14.

    Sidransky, E. et al. Multicenter Analysis of Glucocerebrosidase Mutations in Parkinson’s Disease. New England Journal of Medicine 361(17), 1651–1661 (2009).

  15. 15.

    West, A. B. et al. Parkinson’s disease-associated mutations in leucine-rich repeat kinase 2 augment kinase activity. Proceedings of the National Academy of Sciences of the United States of America 102, 16842–16847, https://doi.org/10.1073/pnas.0507360102 (2005).

  16. 16.

    Gan-Or, Z. et al. LRRK2 mutations in Parkinson disease; a sex effect or lack thereof? A meta-analysis. Parkinsonism & related disorders 21, 778–782, https://doi.org/10.1016/j.parkreldis.2015.05.002 (2015).

  17. 17.

    Gan-Or, Z. et al. Differential effects of severe vs mild GBA mutations on Parkinson disease. Neurology 84, 880–887, https://doi.org/10.1212/WNL.0000000000001315 (2015).

  18. 18.

    Alcalay, R. N. et al. Parkinson disease phenotype in Ashkenazi Jews with and without LRRK2 G2019S mutations. Movement disorders: official journal of the Movement Disorder Society 28, 1966–1971, https://doi.org/10.1002/mds.25647 (2013).

  19. 19.

    Mirelman, A. et al. Fall risk and gait in Parkinson’s disease: the role of the LRRK2 G2019S mutation. Movement disorders: official journal of the Movement Disorder Society 28, 1683–1690, https://doi.org/10.1002/mds.25587 (2013).

  20. 20.

    Mirelman, A. et al. Gait alterations in healthy carriers of the LRRK2 G2019S mutation. Annals of neurology 69, 193–197, https://doi.org/10.1002/ana.22165 (2011).

  21. 21.

    Malec-Litwinowicz, M. et al. Cognitive impairment in carriers of glucocerebrosidase gene mutation in Parkinson disease patients. Neurologia i neurochirurgia polska 48, 258–261, https://doi.org/10.1016/j.pjnns.2014.07.005 (2014).

  22. 22.

    Do, C. B. et al. Web-based genome-wide association study identifies two novel loci and a substantial genetic component for Parkinson’s disease. PLoS genetics 7, e1002141, https://doi.org/10.1371/journal.pgen.1002141 (2011).

  23. 23.

    Eriksson, N. et al. Web-Based, Participant-Driven Studies Yield Novel Genetic Associations for Common Traits. PLoS Genetics 6(6), e1000993 (2010).

  24. 24.

    Eriksson, N. et al. Novel Associations for Hypothyroidism Include Known Autoimmune Risk Loci. PLoS ONE 7(4), e34442 (2012).

  25. 25.

    Benjamini, Y. & Hochberg, Y. Controlling the false discover rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological) 57, 289–300 (1995).

  26. 26.

    Jones, A. V. et al. GWAS of self-reported mosquito bite size, itch intensity and attractiveness to mosquitoes implicates immune-related predisposition loci. Hum Mol Genet 26, 1391–1406, https://doi.org/10.1093/hmg/ddx036 (2017).

Download references

Acknowledgements

We would like to thank the research participants and employees of 23andMe for making this work possible.

Author information

Author notes

    • Ashley R. Winslow

    Present address: Orphan Disease Center, University of Pennsylvania, 125 S. 31st St., Philadelphia, PA, 19104, USA

    • Warren D. Hirst

    Present address: Parkinson’s Disease and Movement Disorders Research, Biogen Research and Early Development, 115 Broadway, Cambridge, MA, 02142, USA

Affiliations

  1. Human Genetics and Computational Biomedicine, Pfizer Worldwide Research and Development, 1 Portland Street, Cambridge, MA, 02139, USA

    • Ashley R. Winslow
    •  & Jemma B. Wilk
  2. Early Clinical Development, Pfizer Worldwide Research and Development, 558 Eastern Point Rd., Groton, CT, 06340, USA

    • Craig L. Hyde
  3. Neuroscience, Pfizer Worldwide Research and Development, 1 Portland Street, Cambridge, MA, 02139, USA

    • Warren D. Hirst
  4. 23andMe Inc., 899W. Evelyn Avenue, Mountain View, CA, 94041, USA

    • Nicholas Eriksson
    •  & Paul Cannon
  5. Human Genetics and Computational Target Validation, Pfizer Worldwide Research and Development, 1 Portland Street, Cambirdge, MA, 02139, USA

    • Melissa R. Miller

Authors

  1. Search for Ashley R. Winslow in:

  2. Search for Craig L. Hyde in:

  3. Search for Jemma B. Wilk in:

  4. Search for Nicholas Eriksson in:

  5. Search for Paul Cannon in:

  6. Search for Melissa R. Miller in:

  7. Search for Warren D. Hirst in:

Contributions

J.B.W., A.R.W., C.H., N.E. designed analyses. N.E., C.H., A.R.W. performed analyses. A.R.W., C.H., W.D.H., J.B.W., M.R.M. interpreted findings. A.R.W., C.H., M.R.M., W.D.H., J.B.W., P.C. wrote and reviewed manuscript.

Competing Interests

This research was funded by Pfizer, Inc. C.H. and M.R.M. are current employees of Pfizer and may own stock or stock options in Pfizer, Inc. A.R.W., J.B.W., and W.D.H. were employees of Pfizer when the study was initiated and may own stock in Pfizer, Inc. W.D.H. is currently an employee of Biogen and owns stock in Biogen. J.B.W. is a consultant working with companies in Cambridge, M.A. A.R.W. is a current employee of the Orphan Disease Center at the Perelman School of Medicine at the University of Pennsylvania. A.R.W. has received honoraria from the Children’s Hospital of Pennsylvania and Drexel University; has consulted for the health sciences sector; is on the board of Philadelphia Science Action. N.E. and P.C. are current employees of 23andMe and may own stock or stock options in 23andMe, Inc. PC is a member of the International Movement Disorders Society.

Corresponding author

Correspondence to Ashley R. Winslow.

Electronic supplementary material

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/s41598-018-30843-6

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.