Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 42 of the 42 references it cites were found in a registry or an index; 5 of the 28 sources its claims cite were checked for support.

Integrated Model of De Novo and Inherited Genetic Variants Yields Greater Power to Identify Risk Genes

Open access

Abstract

De novo mutations affect risk for many diseases and disorders, especially those with early-onset. An example is autism spectrum disorders (ASD). Four recent whole-exome sequencing (WES) studies of ASD families revealed a handful of novel risk genes, based on independent de novo loss-of-function (LoF) mutations falling in the same gene, and found that de novo LoF mutations occurred at a twofold higher rate than expected by chance. However successful these studies were, they used only a small fraction of the data, excluding other types of de novo mutations and inherited rare variants. Moreover, such analyses cannot readily incorporate data from case-control studies. An important research challenge in gene discovery, therefore, is to develop statistical methods that accommodate a broader class of rare variation. We develop methods that can incorporate WES data regarding de novo mutations, inherited variants present, and variants identified within cases and controls. TADA, for Transmission And De novo Association, integrates these data by a gene-based likelihood model involving parameters for allele frequencies and gene-specific penetrances. Inference is based on a Hierarchical Bayes strategy that borrows information across all genes to infer parameters that would be difficult to estimate for individual genes. In addition to theoretical development we validated TADA using realistic simulations mimicking rare, large-effect mutations affecting risk for ASD and show it has dramatically better power than other common methods of analysis. Thus TADA's integration of various kinds of WES data can be a highly effective means of identifying novel risk genes. Indeed, application of TADA to WES data from subjects with ASD and their families, as well as from a study of ASD subjects and controls, revealed several novel and promising ASD candidate genes with strong statistical support.

Overview

CoChat Summary

TL;DR

TADA is a hierarchical Bayesian gene-based method that integrates de novo, inherited transmitted/nontransmitted, and case-control rare variant data to identify disease risk genes. Applied to autism spectrum disorder exome data, it estimates about 1,000 ASD risk genes with average loss-of-function relative risk about 20 and identifies novel candidate genes with greater power than de novo-only or meta-analytic tests.

Key findings

  • In simulations, TADA had superior power: for a gene with typical parameter values at type I error 0.001, its power was about fivefold larger than the De Novo and Meta tests; genome-wide at FDR below 10%, TADA identified more than 200 ASD risk genes, versus about 50 for the De Novo test and about 70 for the Meta test.
  • Using the model, the authors estimated approximately 1,000 ASD risk genes with an average relative risk of approximately 20 due to LoF mutations; an earlier multi-hit-count analysis gave a plausible range of 550 to 1000 risk genes.
  • De novo LoF mutations occurred at about a twofold higher rate in ASD probands than unaffected siblings; de novo probably damaging missense (Mis3) mutations were also higher in probands but less strikingly, while other mutation types showed essentially no difference.
  • Applying TADA to combined ASD data, 15 genes met an FDR less than 20%, including all five genes with two de novo LoF mutations and novel candidates such as KATNAL2, CHD8, S100G, DYRK1A, PPM1D, SCN2A, CUL3, DEAF1, BANK1, POGZ, WDR55, FAM91A1, and COL25A1.

Methods

  • TADA uses a gene-based likelihood model for rare deleterious mutations acting dominantly, with parameters for mutation rate, allele frequency, and gene-specific relative risk/penetrance; it tests relative risk equal to 1 versus greater than 1 using a Hierarchical Bayes/Empirical Bayes strategy that borrows information across genes and can combine multiple mutation types.
  • Data included de novo mutations from 932 ASD families across four whole-exome sequencing studies, transmitted/nontransmitted variants from 641 of those families, and case-control data from the ARRA Autism Sequencing Consortium consisting of 935 ASD subjects and 870 controls; analyses focused on LoF and PolyPhen2 “probably damaging” (Mis3) mutations.
  • The authors performed simulations under null and alternative models using estimated ASD genetic parameters, applied TADA to ASD data with Benjamini-Hochberg FDR control, and conducted sensitivity analyses varying mutation rates and the prior average relative risk.

Limitations

  • The paper states that genotyping/sequencing errors can bias family and combined data analyses, though it argues the variant calls were carefully validated and harmonized; modeling measurement errors is left for future research. It also notes TADA is a one-sided test powered to detect over-transmitted risk variants, so under-transmission bias is not a major concern.
  • The paper cautions that the S100G signal is based on a splice variant common in African populations and should be viewed with caution, and that a C1orf95 signal is likely due to population substructure.
  • Evident in the text: the method relies on PolyPhen2 classification and downweighting of missense mutations, and the top-gene list uses an FDR threshold of less than 20%, which is relatively permissive.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1371/journal.pgen.1003671

Cites

42 of 42 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Properties of the Multiplicity Test. (A) The probability a risk gene has two or more de novo LoF mutations in families (i.e., the power) depends on the mutation rate . Power per gene of the Multiplicity Test as a function of is shown for 4 mutation rates, which were chosen based on percentiles (25'th, 50'th, 75'th, 90'th) of the distribution of obtained from the full gene set. (B) The expected number of risk genes discovered by the Multiplicity Test at (red, solid) or 3 (blue, dashed) as a function of the sample size . The barplot shows the FDR at . The simulation assumes 1000 diseases genes out of 18,000, each with relative risk ; these parameters were estimated in the section on Genetic Architecture of ASD.
Figure 1 Properties of the Multiplicity Test. (A) The probability a risk gene has two or more de novo LoF mutations in families (i.e., the power) depends on the mutation rate . Power per gene of the Multiplicity Test as a function of is shown for 4 mutation rates, which were chosen based on percentiles (25'th, 50'th, 75'th, 90'th) of the distribution of obtained from the full gene set. (B) The expected number of risk genes discovered by the Multiplicity Test at (red, solid) or 3 (blue, dashed) as a function of the sample size . The barplot shows the FDR at . The simulation assumes 1000 diseases genes out of 18,000, each with relative risk ; these parameters were estimated in the section on Genetic Architecture of ASD.CC BY · PubMed Central
Figure 2 A probabilistic model for a family trio with an affected child. Genotype probabilities are computed as the marginal probability of parental genotypes times the conditional probability of the child, given the parents. The parameters and represent the mutation rate, and the population frequency of the genotype, respectively. Phenotype probabilities for the child, given genotype, are a function of , the penetrance of the genotype, and the relative risk of the mutation . Rate is the (approximate) rate of observing counts , and from the latter 3 types of trios, respectively.
Figure 2 A probabilistic model for a family trio with an affected child. Genotype probabilities are computed as the marginal probability of parental genotypes times the conditional probability of the child, given the parents. The parameters and represent the mutation rate, and the population frequency of the genotype, respectively. Phenotype probabilities for the child, given genotype, are a function of , the penetrance of the genotype, and the relative risk of the mutation . Rate is the (approximate) rate of observing counts , and from the latter 3 types of trios, respectively.CC BY · PubMed Central
Figure 3 The genetic parameters of ASD. (A) The relationship between the number of ASD risk genes () and the average relative risk (). stands for the total number of genes in the human genome, and for the fold enrichment of the de novo LoF mutations in probands vs. siblings (about 2 in our data). (B) The expected number of multi-hit genes () in families, as a function of the number of ASD risk genes (). The observed is 5, and we define the plausible range of as the values corresponding to to 6. The model assumes the relative risks of ASD risk genes follow a gamma distribution with the scale parameter . The variance of the relative risk () across genes equals ( is the average of of all ASD risk genes), which limits the range of plausible values for the model. The estimated value of the average is approximately 20. (C) For each gene, we compute the empirical allele frequency () of LoFs as the number of LoF variants divided by the sample size. The histogram of the LoF frequencies of all genes is shown. Also shown are the estimated distributions of under the null (red, solid line) and the alternative (blue, dashed line) models, respectively.
Figure 3 The genetic parameters of ASD. (A) The relationship between the number of ASD risk genes () and the average relative risk (). stands for the total number of genes in the human genome, and for the fold enrichment of the de novo LoF mutations in probands vs. siblings (about 2 in our data). (B) The expected number of multi-hit genes () in families, as a function of the number of ASD risk genes (). The observed is 5, and we define the plausible range of as the values corresponding to to 6. The model assumes the relative risks of ASD risk genes follow a gamma distribution with the scale parameter . The variance of the relative risk () across genes equals ( is the average of of all ASD risk genes), which limits the range of plausible values for the model. The estimated value of the average is approximately 20. (C) For each gene, we compute the empirical allele frequency () of LoFs as the number of LoF variants divided by the sample size. The histogram of the LoF frequencies of all genes is shown. Also shown are the estimated distributions of under the null (red, solid line) and the alternative (blue, dashed line) models, respectively.CC BY · PubMed Central
Figure 4 The power per gene of competing tests. The results of three tests are shown: novo (red), meta (blue), and TADA (purple). Results are shown for various values of , and with type I error fixed at 0.001. Parameter values are chosen to cover plausible parameter values according to our model estimation: (A) ; (B) ; and (C) .
Figure 4 The power per gene of competing tests. The results of three tests are shown: novo (red), meta (blue), and TADA (purple). Results are shown for various values of , and with type I error fixed at 0.001. Parameter values are chosen to cover plausible parameter values according to our model estimation: (A) ; (B) ; and (C) .CC BY · PubMed Central
Figure 5 Application of TADA to the genetic data of ASD. (A) De novo LoF and “probably damaging” missense mutations are enriched in ASD probands (red) compared with unaffected siblings (blue), based on a comparison including all trio and quad families. The other types of missense mutations are not enriched. To make the numbers comparable, the number of mutations in siblings is scaled by a constant multiplier (214/124) so that the numbers of silent mutations is equal in probands and in siblings. The annotations of missense mutations are based on PolyPhen. (B) Q-Q plot (log. scale) of the values for all genes in the ASD dataset based on a combined analysis of LoF and severe missense mutations.
Figure 5 Application of TADA to the genetic data of ASD. (A) De novo LoF and “probably damaging” missense mutations are enriched in ASD probands (red) compared with unaffected siblings (blue), based on a comparison including all trio and quad families. The other types of missense mutations are not enriched. To make the numbers comparable, the number of mutations in siblings is scaled by a constant multiplier (214/124) so that the numbers of silent mutations is equal in probands and in siblings. The annotations of missense mutations are based on PolyPhen. (B) Q-Q plot (log. scale) of the values for all genes in the ASD dataset based on a combined analysis of LoF and severe missense mutations.CC BY · PubMed Central
Figure 6 Bayesian hierarchical model of TADA. A fraction of the genes are associated with the phenotype under investigation and follow model , and the remainder follow model . The prior distribution of gene-specific parameters, relative risk () and allele frequency (), can vary under the competing models, or . Priors are specified by the hyperparameters, and , respectively, which are estimated from the data. Counts of events for the i-th gene follow a Poisson distribution, parameterized by and under , and under .
Figure 6 Bayesian hierarchical model of TADA. A fraction of the genes are associated with the phenotype under investigation and follow model , and the remainder follow model . The prior distribution of gene-specific parameters, relative risk () and allele frequency (), can vary under the competing models, or . Priors are specified by the hyperparameters, and , respectively, which are estimated from the data. Counts of events for the i-th gene follow a Poisson distribution, parameterized by and under , and under .CC BY · PubMed Central

Verification

Claims the paper makes

4 Valid · 7 Not valid · 19 Not checked · 80 Own results and interpretations

52 more claims

Create a free account to see all 110 claims and to verify the 23 references still open.

See all claims — free account
  • It may state two separate claims as one. 34
  • Probably not a substantial claim. 22
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Show all 42 references

Cited by

See all 318

Metadata

Authors
He, X., Sanders, S. J., Liu, L., De Rubeis, S., Lim, E. T., Sutcliffe, J. S., Schellenberg, G. D., Gibbs, R. A., Daly, M. J., Buxbaum, J. D., State, M. W., Devlin, B., & Roeder, K.
Affiliations
  • Carnegie Mellon University (US)
  • Yale University (US)
  • Broad Institute (US)
  • Vanderbilt University (US)
  • University of Pennsylvania (US)
  • Baylor College of Medicine (US)
  • Icahn School of Medicine at Mount Sinai (US)
  • University of Pittsburgh (US)
Corresponding author
Roeder, Kathryn
Venue
PLoS Genetics
Published
2013-08-15
Volume
9
Issue
8
Pages
e1003671
ISSN
  • 1553-7390 (ISSN-L)
  • 1553-7404
OpenAlex
W2125226755
Cited by
304
References
42
Citations by year
2026: 18 · 2025: 23 · 2024: 29 · 2023: 18 · 2022: 23 · 2021: 31 · 2020: 25 · 2019: 24 · 2018: 26 · 2017: 23 · 2016: 27 · 2015: 21 · 2014: 15 · 2013: 1
FWCI
14.14
Type
journal article
Language
en
Open access
Locations
  • PLoS Genetics (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • D-Scholarship@Pitt (University of Pittsburgh) (acceptedVersion)
  • Digital Access to Scholarship at Harvard (DASH) (Harvard University) (submittedVersion, cc-by)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • eScholarship (California Digital Library) (submittedVersion)
  • eScholarship (California Digital Library) (submittedVersion, cc-by)
  • Figshare (submittedVersion, cc-by)
Funders
  • Autism Speaks
  • Simons Foundation
  • Simons Foundation Autism Research Initiative
  • National Institutes of Health (MH057881)
  • National Institute of Mental Health (U01MH100209, R01MH061009, R01MH089025, U01MH100229, MH089004, MH089025, R01MH089175, R01MH089004, R37MH057881, R01MH089208, U01MH100239, R01MH057881, MH089482, MH057881, R01MH089482, MH061009, U01MH100233)
  • National Human Genome Research Institute
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genomics and Rare Diseases
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Autism Spectrum Disorder Research
Keywords
de novo mutations, autism spectrum disorder, TADA, Bayesian hierarchical model, penetrance, allele frequency, exome sequencing
MeSH terms
Bayes Theorem, Child Development Disorders, Pervasive, Humans, Models, Theoretical, Mutation, Risk Factors, Genetic Variation, Likelihood Functions, Case-Control Studies, Sequence Analysis, DNA, Genetic Predisposition to Disease, Exome

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Integrated Model of De Novo and Inherited Genetic Variants Yields Greater Power to Identify Risk Genes. https://app.cochat.ai/papers/10.1371/journal.pgen.1003671
The paper
He, X., Sanders, S. J., Liu, L., De Rubeis, S., Lim, E. T., Sutcliffe, J. S., Schellenberg, G. D., Gibbs, R. A., Daly, M. J., Buxbaum, J. D., State, M. W., Devlin, B., & Roeder, K. (2013). Integrated Model of De Novo and Inherited Genetic Variants Yields Greater Power to Identify Risk Genes. PLoS Genetics. https://doi.org/10.1371/journal.pgen.1003671

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.