Cohort Analytics Core

National Human Genome Research Institute

Investigators

Linked publications, trials & patents

Abstract

The Cohort Analytics Core has worked on more than 20 collaborative projects this year across 10 investigators. We have also held training sessions for groups and mentored trainees throughout NHGRI to help get scientists acquainted with EHR-based analyses. Some highlights from our recent work include the following: 1. Uric Acid  This project is a collaboration with the Purine and Pyrimidine Metabolism Unit at NHGRI, headed by Dr. Oleg Shchelochkov. We explored the question of whether high uric acid is a causal risk factor for cardiovascular and related diseases in a diverse, trans-ancestry All of Us cohort. To control for confounding effects of environmental variables, we used a genetic risk score for uric acid as a proxy for lab-based measurements. In the GRS-PheWAS, only gout was significant after correction for multiple testing, though hypertension reached the nominal significance level. Follow-up self-controlled case series analysis showed a slight decrease in systolic blood pressure after treatment with allopurinol, a uric acid-lowering drug. This manuscript is under review. 2. Genotype-first evaluation of rare Mendelian disorders In this second collaboration with the Shchelochkov group, we implemented a genotype-first approach in All of Us to assess the prevalence and phenotypic spectrum of rare disorders. This involved developing a stepwise pipeline for filtering variants based on ClinVar classifications and in silico predictors. We applied this pipeline to generate genotype-first prevalence estimates and explore genotype-phenotype associations for autosomal dominant polycystic kidney disease (ADPKD). Our findings indicate a gap between genotype- and phenotype-first prevalence estimates for ADPKD and provide evidence that evolving variant classification, incomplete penetrance, and underutilization of diagnostic imaging contribute to this gap. This analysis pipeline is designed to be generalizable to any gene of interest. In an ongoing project, we are applying this approach to study disorders of purine and pyrimidine metabolism. 3. Genetic variants and clinical diagnoses in Marfan Syndrome Marfan syndrome is a rare autosomal dominant connective tissue disorder with variable clinical phenotypes affecting the cardiovascular, musculoskeletal, and sensory systems. It is largely caused by autosomal dominant variants in fibrillin 1 (FBN1); 25% of cases are de novo and <10% of cases lack FBN1 mutations or exhibit recessive inheritance. Similarities with other disorders, such as Loeys-Dietz syndrome, complicate an accurate diagnosis. To alleviate this, the 2010 Revised Ghent Nosology criteria emphasize certain clinical symptoms, such as aortic root dilation and ectopia lentis as key criteria for Marfan diagnosis. We leveraged All of Us data to explore discrepancies between participants with FBN1 genetic variants and clinical Marfan diagnoses. The Marfan-diagnosed cohort showed associations with aortic dissection (OR=47.0, p=49e-5), mitral valve prolapse (OR=17.5, p=3.14e-15), pneumothorax (OR=7.54, p=1.49e-5), and joint hypermobility (OR=27.9, p=2.18e-15). These were absent in the FBN1 1* PLP cohort, which instead showed a notable association with congenital lens dislocation (OR=23.2, p=1.56e-5). 4. Phenotype association of Methylenetetrahydrofolate reductase (MTHFR) variants This project is a collaboration with Laboratory Genetics and Genomics. MTHFR is crucial in folate and homocysteine metabolism, with the p.(Ala222Val) variant linked to decreased enzyme activity and elevated homocysteine. Despite previous studies associating this variant with various conditions, meta-analyses have not supported these links, and testing is not recommended by expert groups like ACMG and NSGC. To evaluate in a population cohort, we conducted PheWAS in All of Us and UK Biobank, finding significant associations only with disturbances in amino-acid metabolism (a phenotype that can arise when homocysteinemia has been noted) in the All of Us cohort. Previous MTHFR associated conditions, including thrombosis, major depressive disorder, and breast cancer were not statistically significant in either cohort. Our findings support current guidelines advising against routine MTHFR variant testing. 5. Systemic Lupus Erythematosus (SLE) Cohort Building In collaboration with Dr. Lindsey Criswell's GARDS lab, two projects aim to enhance our understanding of SLE using All of Us data. The first involved identifying genetic predisposition factors for SLE through EHR phenotyping and GWAS, revealing 59 significant SNPs with expected notable associations in HLA-B. An African ancestry-specific GWAS highlights associations in STAT4, BANK1, PHRF1, and ITGAM, emphasizing the importance of studying varied populations. The second project focused on creating a robust, computable phenotyping algorithm for accurate SLE case identification from EHR data, incorporating multiple established classification systems to balance sensitivity and specificity. This framework facilitates reproducible downstream analyses, including GWAS and clinical outcome research. 6. Identification of Novel Genetic-Risk Factors for Type 2 Diabetes Mellitus and Alzheimer's Diseases In collaboration with Dr. Jianhua Zhang at NIDDK, a gene function–centered method was used to identify novel genes associated with T2DM and AD. This approach complements traditional GWAS and has proven effective in uncovering disease-associated genes often missed by GWAS. Analyzing ancestry- and gender-specific exome datasets, the project focused on high-risk and pathogenic variants, identifying candidate genes for T2DM or AD association. These candidates will undergo further molecular and pathway analyses to assess their disease relevance, potentially advancing our understanding of the genetic architecture of T2DM and AD and informing personalized disease management. 7. Putative Pleiotropic Functions of the 5q33 Red Cell Immunity Risk Locus Collaborating with Dr. Jacqueline Piekos at CCDGS, we are exploring the genetic variation at chromosome 5q33.3, which is linked to various traits, including longevity, cardiovascular health, autoimmune disorders, and resistance to tuberculosis infection. Our research has identified associations between 5q33 variants and RBC alloimmunization in SCD patients, with significant risks found in individuals of African ancestry. Further PheWAS analyses using the All of Us dataset revealed a protective effect of these variants against secondary anemia in non-SCD contexts, suggesting immunological pleiotropy. Ongoing efforts aim to validate these findings in other cohorts and elucidate the underlying immune mechanisms, particularly the role of IL12B expression. 8. Genetics of HIV Disease Progression in Children from Africa This collaboration with fellow NHGRI investigator Dr. Neil Hanchard investigates the genetic factors influencing HIV disease progression in African children, identifying key loci associated with rate of progression. Utilizing All of Us, PheWAS analyses were conducted on specific variants to uncover their broader impact on health and disease beyond HIV progression. These analyses aim to identify potential pleiotropic effects, offering deeper insights into the genetic underpinnings of HIV disease progression and their implications for clinical outcomes in diverse populations. 9. Hematological Studies We collaborated with Drs. Katherine Calvo and Kalpana Upadhyaya at DLM Hematology section, as well as Dr. Danica Novacic at NHGRI Undiagnosed Diseases Program (UDP) to evaluate hematological and immunological phenotypes with All of Us for variants of unknown significance (VUS) detected during routine patient care at the Clinical Center. We also collaborated with Dr. Kalpana Manthiram at NIAID to investigate hematological and inflammatory phenotypes that are associated with trisomy 8 in the general population with All of Us and UK Biobank datasets.