Estimating variance components in population scale family trees

Shor, Tal; Geiger, Dan; Erlich, Yaniv; Weissbrod, Omer

doi:10.1101/256396

Cited by 7 publications

(9 citation statements)

References 102 publications

(128 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Examples of such databases include the Utah Population database and a number of reliable Icelandic databases 51,52 . The availability of population-scale family trees 53 would further allow the determination of the parental origin of alleles 54 and generally enable large scale human population studies on epidemiological history 55 . Overall, the reliability of data and its impact on T1D and RA diagnoses must be discussed.…”

Section: Discussionmentioning

confidence: 99%

Genomic imprinting analyses identify maternal effects as a cause of phenotypic variability in type 1 diabetes and rheumatoid arthritis

Blunk

Thomsen

Reinsch

et al. 2020

Sci Rep

View full text Add to dashboard Cite

Imprinted genes, giving rise to parent-of-origin effects (POEs), have been hypothesised to affect type 1 diabetes (T1D) and rheumatoid arthritis (RA). However, maternal effects may also play a role. By using a mixed model that is able to simultaneously consider all kinds of POEs, the importance of POEs for the development of T1D and RA was investigated in a variance components analysis. The analysis was based on Swedish population-scale pedigree data. With P = 0.18 (T1D) and P = 0.26 (RA) imprinting variances were not significant. Explaining up to 19.00% (± 2.00%) and 15.00% (± 6.00%) of the phenotypic variance, the maternal environmental variance was significant for T1D (P = 1.60 × 10 −24) and for RA (P = 0.02). For the first time, the existence of maternal genetic effects on RA was indicated, contributing up to 16.00% (± 3.00%) of the total variance. Environmental factors such as the social economic index, the number of offspring, birth year as well as their interactions with sex showed large effects. The failure of the immune system to distinguish self from non-self antigens is the basis for autoimmune disorders (AIs) 1. Type I diabetes (T1D) is an AI that causes chronic destruction of pancreatic islet ß-cells and hyperglycemia due to reduced insulin production 2. With the incidence said to be increasing by 3-4% yearly, more than 20 million individuals are estimated to have T1D worldwide 3. Rheumatoid arthritis (RA) is associated with autoantigen presentation with antigen specific T and B cell activation and aberrant inflammatory cytokine production. Consequences thereof include synovitis, proliferation of synovia and cartilage, and subchondral bone destruction 4. The occurrence of RA is relatively constant and ranges between 0.5 and 1.0% in European and North-American populations 5. The exact etiology of T1D and RA remains largely unknown 1 , however, a complex interplay of genetic, environmental, and epigenetic factors is assumed 4,6,7. With regard to genetic factors, the strongest effects have been found within the major histocompatibility complex or human leukocyte antigen system. T1D and RA show genetic overlap in terms of associations within HLA, PTPN22, CTLA4, and TAGAP 8. Causal loci explain over 80% of T1D heritability which reportedly ranges between 40 and 92% 3. For RA, associated variants outside and inside of the major histocompatibility complex region explain about 5% and 60% of the heritability 9,10. Heritability estimates range between 12 and 60% 11-13. As disorder concordance rates in monozygotic twins have been observed to be less than 100%, AIs are assumed to be subject to epigenetic modifications 4,14-16. Perhaps the best-known example for all epigenetic phenomena is imprinting, in which the expression of genes is either maternally or paternally inactivated. Inactivation

show abstract

Section: Discussionmentioning

confidence: 99%

Genomic imprinting analyses identify maternal effects as a cause of phenotypic variability in type 1 diabetes and rheumatoid arthritis

Blunk

Thomsen

Reinsch

et al. 2020

Sci Rep

View full text Add to dashboard Cite

show abstract

“…We computed heritability with the Sci-LMM Python package, which constructs and works with large scale relationships matrices and fits them to the corresponding LMM within several hours. Our Identity By Descent (IBD) matrix (an identity-by-descent relationships based matrix) was the ARM computed from the entire pedigree ( 27 ) . We used Haseman-Elston regression to compute the heritability measure , and we estimate the h 2 standard error via the average information restricted maximum likelihood (AI-REML) procedure ( 44 , 45 ) .…”

Section: Heritability Estimationmentioning

confidence: 99%

“…Today, heritability estimation is typically performed using Linear Mixed Models (LMMs) ( 26 ) and requires a kinship matrix, commonly computed from genetic information. Using the Sparse Cholesky Factorization (Sci-LMM) package ( 27 ) , a statistical modeling framework for analyzing population-size pedigrees, we constructed a relationship matrix and fitted the corresponding LMM solely from EHR information, without costly genetic testing. We estimated the heritability of absolute HbA1c% reduction in response to metformin to be 15.9% (95% CI 1.2%-30.5%) in males and 20.9% (95% CI 7.5%-34.3%) in females of the total explained variability.…”

Section: Introductionmentioning

confidence: 99%

Estimating Heritability of Glycaemic Response to Metformin using Nationwide Electronic Health Records and Population-Sized Pedigree

Kalka

Gavrieli

Shilo

et al. 2021

Preprint

View full text Add to dashboard Cite

Variability of response to medication is a well known phenomenon, determined by both environmental and genetic factors. Understanding the heritable component of the response to medication is of great interest but challenging due to several reasons, including small study cohorts and computational limitations. Here, we studied the heritability of variation in the glycaemic response to metformin, first-line therapeutic agent for type 2 diabetes (T2D), by leveraging 17 years of electronic health records (EHR) data from Israel’s largest healthcare service provider, consisting of over five million patients of diverse ethnicities and socio-economic background. Our cohort consisted of 74,871 T2D patients treated with metformin, with an accumulated number of 1,358,776 HbA1C measurements and 323,260 metformin prescriptions. We estimated the explained variance of glycated hemoglobin (HbA1c%) reduction due to heritability by constructing a six-generation population-size pedigree from national registries and linking it to medical health records. Using Linear Mixed Model-based framework, a common-practice method for heritability estimation, we calculated a heritability measure of h2 = 15.9% (95% CI, 1.2% − 30.5%) for absolute reduction of HbA1c% after metformin treatment in males and h2 = 20.9% (95% CI, 7.5% − 34.3%) in females. Results remained unchanged after adjusting for pre-treatment HbA1c% in females, and for both genders in proportional reduction of HbA1c%. To the best of our knowledge, our work is the first to estimate heritability of drug response using EHR data. We demonstrated that while response to metformin treatment has a heritable component, most of the variation is likely due to other factors, further motivating non-genetic analyses aimed at unraveling metformin’s mechanism of action.

show abstract

“…To reduce the computational complexity of genetic similarity matrix operations, several research programs have been conducted to store and manipulate sparse representations of the genetic similarity matrix (Shor et al, 2019 ). In these representations, researchers set a threshold and then zero out elements of the genetic similarity matrix with absolute value less than this threshold.…”

Section: Introductionmentioning

confidence: 99%

Kinship Solutions for Partially Observed Multiphenotype Data

Elliott

2020

Journal of Computational Biology

View full text Add to dashboard Cite

Current work for multivariate analysis of phenotypes in genome-wide association studies often requires that genetic similarity matrices be inverted or decomposed. This can be a computational bottleneck when many phenotypes are presented, each with a different missingness pattern. A usual method in this case is to perform decompositions on subsets of the kinship matrix for each phenotype, with each subset corresponding to the set of observed samples for that phenotype. We provide a new method for decomposing these kinship matrices that can reduce the computational complexity by an order of magnitude by propagating low-rank modifications along a tree spanning the phenotypes. We demonstrate that our method provides speed improvements of around 40% under reasonable conditions.

show abstract

Estimating variance components in population scale family trees

Cited by 7 publications

References 102 publications

Genomic imprinting analyses identify maternal effects as a cause of phenotypic variability in type 1 diabetes and rheumatoid arthritis

Genomic imprinting analyses identify maternal effects as a cause of phenotypic variability in type 1 diabetes and rheumatoid arthritis

Estimating Heritability of Glycaemic Response to Metformin using Nationwide Electronic Health Records and Population-Sized Pedigree

Kinship Solutions for Partially Observed Multiphenotype Data

Contact Info

Product

Resources

About