The False Dawn of Polygenic Risk Scores for Human Disease Prediction.

Polygenic risk scores (PRSs) are being constructed for many diseases and are presented today as a promising avenue in the field of human genetics. These scores aim at predicting the risk of developing a disease by leveraging the many genome-wide association studies (GWAS) conducted during the two last decades. Important investments are being made to improve score estimates by increasing GWAS sample sizes, by developing more sophisticated methods, and by proposing different corrections for potential biases. PRSs have entered the market with direct-to-consumer companies proposing to compute them from saliva samples and even recently to help parents select the healthiest embryos. In this paper, we recall how PRSs arose and question the credit they are given by revisiting underlying assumptions in light of the history of human genetics and by comparing them with estimated breeding values (EBVs) used for selection in livestock.

Heritability: What's the point? What is it not for? A human genetics perspective.

In this paper, we explain the concept of heritability and describe the different methods and the genotype-phenotype correspondences used to estimate heritability in the specific field of human genetics. Heritability studies are conducted on extremely diverse human traits: quantitative traits (physical, biological, but also cognitive and behavioral measurements) and binary traits (as is the case of most human diseases). Instead of variables such as education and socio-economic status as covariates in genetic studies, they are now the direct object of genetic analysis. We make a review of the different assumptions underlying heritability estimates and dispute the validity of most of them. Moreover, and maybe more importantly, we show that they are very often misinterpreted. These erroneous interpretations lead to a vision of a genetic determinism of human traits. This vision is currently being widely disseminated not only by the mass media and the mainstream press, but also by the scientific press. We caution against the dangerous implication it has both medically and socially. Contrarily to the field of animal and plant genetics for which the polygenic model and the concept of heritability revolutionized selection methods, we explain why it does not provide answer in human genetics.

Mutations in KIAA0586 Cause Lethal Ciliopathies Ranging from a Hydrolethalus Phenotype to Short-Rib Polydactyly Syndrome.

KIAA0586, the human ortholog of chicken TALPID3, is a centrosomal protein that is essential for primary ciliogenesis. Its disruption in animal models causes defects attributed to abnormal hedgehog signaling; these defects include polydactyly and abnormal dorsoventral patterning of the neural tube. Here, we report homozygous mutations of KIAA0586 in four families affected by lethal ciliopathies ranging from a hydrolethalus phenotype to short-rib polydactyly. We show defective ciliogenesis, as well as abnormal response to SHH-signaling activation in cells derived from affected individuals, consistent with a role of KIAA0586 in primary cilia biogenesis. Whereas centriolar maturation seemed unaffected in mutant cells, we observed an abnormal extended pattern of CEP290, a centriolar satellite protein previously associated with ciliopathies. Our data show the crucial role of KIAA0586 in human primary ciliogenesis and subsequent abnormal hedgehog signaling through abnormal GLI3 processing. Our results thus establish that KIAA0586 mutations cause lethal ciliopathies.

Functional variants of POC5 identified in patients with idiopathic scoliosis.

Idiopathic scoliosis (IS) is a spine deformity that affects approximately 3% of the population. The underlying causes of IS are not well understood, although there is clear evidence that there is a genetic component to the disease. Genetic mapping studies suggest high genetic heterogeneity, but no IS disease-causing gene has yet been identified. Here, genetic linkage analyses combined with exome sequencing identified a rare missense variant (p.A446T) in the centriolar protein gene POC5 that cosegregated with the disease in a large family with multiple members affected with IS. Subsequently, the p.A446T variant was found in an additional set of families with IS and in an additional 3 cases of IS. Moreover, POC5 variant p.A455P was present and linked to IS in one family and another rare POC5 variant (p.A429V) was identified in an additional 5 cases of IS. In a zebrafish model, expression of any of the 3 human IS-associated POC5 variant mRNAs resulted in spine deformity, without affecting other skeletal structures. Together, these findings indicate that mutations in the POC5 gene contribute to the occurrence of IS.

Revisiting the Polygenic Additive Liability Model through the Example of Diabetes Mellitus.

Most studies on multifactorial diseases are performed under the assumption of a polygenic additive liability. In particular, missing heritability and individual risk scores are estimated under this model. In this paper, we use the example of diabetes to highlight the pitfalls of relying on such a model, when there are reasons to suspect etiological heterogeneity and/or departure from the hypotheses on the environmental factor effects.

Chromosome 21 scan in Down syndrome reveals DSCAM as a predisposing locus in Hirschsprung disease.

Hirschsprung disease (HSCR) genetics is a paradigm for the study and understanding of multigenic disorders. Association between Down syndrome and HSCR suggests that genetic factors that predispose to HSCR map to chromosome 21. To identify these additional factors, we performed a dose-dependent association study on chromosome 21 in Down syndrome patients with HSCR. Assessing 10,895 SNPs in 26 Caucasian cases and their parents led to identify two associated SNPs (rs2837770 and rs8134673) at chromosome-wide level. Those SNPs, which were located in intron 3 of the DSCAM gene within a 19 kb-linkage disequilibrium block region were in complete association and are consistent with DSCAM expression during enteric nervous system development. We replicated the association of HSCR with this region in an independent sample of 220 non-syndromic HSCR Caucasian patients and their parents. At last, we provide the functional rationale to the involvement of DSCAM by network analysis and assessment of SOX10 regulation. Our results reveal the involvement of DSCAM as a HSCR susceptibility locus, both in Down syndrome and HSCR isolated cases. This study further ascertains the chromosome-scan dose-dependent methodology used herein as a mean to map the genetic bases of other sub-phenotypes both in Down syndrome and other aneuploidies.

Male and female differential reproductive rate could explain parental transmission asymmetry of mutation origin in Hirschsprung disease.

Hirschsprung disease (HSCR, aganglionic megacolon) is a complex and heterogeneous disease with an incidence of 1 in 5000 live births. Despite the multifactorial determination of HSCR in the vast majority of cases, there is a monogenic subgroup for which private rare RET coding sequence mutations with high penetrance are found (45% of HSCR familial cases). An asymmetrical parental origin is observed for RET coding sequence mutations with a higher maternal inheritance. A parent-of-origin effect is usually assumed. Here we show that a differential reproductive rate for males and females also leads to an asymmetrical parental origin, which was never considered as a possible explanation till now. In the case of HSCR, we show a positive association between penetrance of the mutation and parental transmission asymmetry: no parental transmission asymmetry is observed in sporadic RET CDS mutation carrier cases for which penetrance of the mutation is low, whereas a parental transmission asymmetry is observed in affected sib-pairs for which penetrance of the mutation is higher. This allows us to conclude that the explanation for this parental asymmetry is that more severe mutations have resulted in a differential reproductive rate between male and female carriers.

Using affected sib-pairs to uncover rare disease variants.

OBJECTIVE: We propose a new test for rare variant mapping, based on an affected sib-pair sample and a control sample. In each sib-pair, only the index case needs to be sequenced, and the number of alleles shared identical-by-descent between the sibs is used as complementary information. The test makes use of both association and linkage information. We compare this test to the Armitage test on case-control data, with cases either from the general population of cases or from a sample of cases having an affected sib. METHODS: A score test based on the likelihood in a multiplicative risk model is proposed. Its power is estimated by simulations and compared to Armitage test's power. RESULTS: The affected sib-pairs design allows a tremendous gain of power over the case-control design (from 1 to 99% for a moderate sample size and relative risk values around 3, at an α level of 10(-11)). When cases are ascertained in a sample of cases having an affected sib, the use of linkage information in our test allows a gain of power of more than 20% in certain situations. CONCLUSION: We demonstrate the interest in using familial data and both association and linkage information for rare variant mapping.

Determination of the real effect of genes identified in GWAS: the example of IL2RA in multiple sclerosis.

Genome-wide association studies (GWAS), although efficient to detect genes involved in complex diseases, are not designed to measure the real effect of the genes. This is illustrated here by the example of IL2RA in multiple sclerosis (MS). Association between IL2RA and MS is clearly established, although the functional variation is still unknown: the effect of IL2RA might be better described by several SNPs than by a single one. This study investigates whether a pair of SNPs better explains the observed linkage and association data than a single SNP. In total, 522 trio families and 244 affected sib-pairs were typed for 26 IL2RA SNPs. For each SNP and pairs of SNPs, the phased genotypes of patients and controls were compared to determine the SNP set offering the best risk discrimination. Consistency between the genotype risks provided by the retained set and the identical by descent allele sharing in affected sib-pairs was assessed. After controlling for multiple testing, the set of SNPs rs2256774 and rs3118470, provides the best discrimination between the case and control genotype distributions (P-corrected=0.009). The relative risk between the least and most at-risk genotypes is 3.54 with a 95% confidence interval of [2.14-5.94]. Furthermore, the linkage information provided by the allele sharing between affected sibs is consistent with the retained set (P=0.80) but rejects the SNP reported in the literature (P=0.006). Establishing a valid modeling of a disease gene is essential to test its potential interaction with other genes and to reconstruct the pathophysiological pathways.

Association of TALS developmental disorder with defect in minor splicing component U4atac snRNA.

The spliceosome, a ribonucleoprotein complex that includes proteins and small nuclear RNAs (snRNAs), catalyzes RNA splicing through intron excision and exon ligation to produce mature messenger RNAs, which, in turn serve as templates for protein translation. We identified four point mutations in the U4atac snRNA component of the minor spliceosome in patients with brain and bone malformations and unexplained postnatal death [microcephalic osteodysplastic primordial dwarfism type 1 (MOPD 1) or Taybi-Linder syndrome (TALS); Mendelian Inheritance in Man ID no. 210710]. Expression of a subgroup of genes, possibly linked to the disease phenotype, and minor intron splicing were affected in cell lines derived from TALS patients. Our findings demonstrate a crucial role of the minor spliceosome component U4atac snRNA in early human development and postnatal survival.

New disease gene location and high genetic heterogeneity in idiopathic scoliosis.

Idiopathic scoliosis (IS) is a spine disorder of unknown origin with 1.5-3% prevalence in the general population. Besides the large multifactorial-form sample of IS, there is a good evidence for the existence of a monogenic subgroup in which the disease is inherited in a dominant manner. However, results from literature suggest a strong heterogeneity in the locations of the mutated genes. Using a high-resolution genome-wide scan, we performed linkage analyses in three large multigenerational IS families compatible with dominant inheritance including 9-12 affected members or obligate carriers. In two of these families, our results suggested intra-familial genetic heterogeneity, whereas, in the other, we observed a perfect marker disease co-segregation in two regions at 3q12.1 and 5q13.3. We can state that one of these two locations is a novel IS disease gene locus, as the probability of having this perfect co-segregation twice by chance in the genome is very low (P=0.001). Lastly, in all three families studied, linkage to the previously mapped dominant IS loci on chromosomes 19p13.3, 17p11.2, 9q34, 17q25 and 18q is unlikely, confirming that there is a high genetic heterogeneity within the subgroup of dominant forms of IS.

Modeling the effect of susceptibility factors (HLA and PTPN22) in rheumatoid arthritis.

Numerous genome-wide analyses on common multifactorial diseases have been recently published in providing, for each associated Single Nucleotide Polymorphism (SNP), an Odds Ratio (OR), either for one of the susceptibility variant allele versus none, or for two copies of it versus one copy. Besides the poor information attached to these measures, it is a simplistic idea to reduce the effect of a gene to the one of an allele or of an haplotype. It is a far cry from detecting a signal indicating the presence of a causative factor in a genomic region to its identification and the important task of estimating the disease risk due to it. The contrast between cases and controls may be used for the estimation of the genotype relative risks. However, the same population distribution of a marker can be coupled with different modes of inheritance of the trait, and hence different risk estimates. Other sources of information, in particular at familial level must be used and can be crucial in discriminating the genotypes according to the disease risk. Illustration is given on two susceptibility factors to Rheumatoid Arthritis: HLA and PTPN22. In both cases, thanks to the sharing of parental alleles in affected sibs, a refining of the modeling was obtained. Tezenas du Montcel et al. (Arthritis Rheum 52:1063-1068, 2005) show that six HLA genotypes can be distinguished with different RA risks. One HLA genotype confers a risk 6.6-fold higher than another HLA genotype. For PTPN22, Bourgey et al. (BMC Proc 1 (Suppl 1):S37, 2007) show that observed data is not explained by a single variant as initially reported and that using the information on 3 SNPs discriminates the genotypic relative risks (GRRs) from 1 to 4.7.

The ordered transmission disequilibrium test: detection of modifier genes.

We consider the problem of detection of modifier genes that lead to variations in a disease-related continuous variable (DRCV), such as the age of onset or a measure of disease severity, in a strategy of candidate genes. We propose a novel method, the ordered transmission disequilibrium test (OTDT), to test for a relation between the clinical heterogeneity expressed by a DRCV and marker genotypes of a candidate gene. The OTDT applies to trio families with one patients and his parents, all three genotyped at a bi-allelic marker M. The OTDT aims to find a critical value of the DRCV which separates the sample of families in two subsamples in which the transmission rates are significantly different. We investigate the power of the method by simulations under various genetic models and covariate distributions and compare it with a linear regression analysis.

IFIH1-GCA-KCNH7 locus is not associated with genetic susceptibility to multiple sclerosis in French patients.

A recent investigation reported, for the first time, an association between variants in the IFIH1-GCA-KCNH7 locus and multiple sclerosis (MS). We sought to replicate this genetic association in MS with a new independent MS cohort composed of French Caucasian MS trio families. The two most significant IFIH1 single nucleotide polymorphisms, rs1990760 and rs2068330, reported as involved in MS susceptibility, were genotyped in 591 French Caucasian MS trio families, and analyzed using the transmission/disequilibrium test. No association with MS was found (rs1990760, P=0.45 and rs2068330, P=0.27). Similarly, no significant association was detected after stratification for HLA-DRB1*1501 carriers. Reasons that may explain this discrepancy between the original report and our study are discussed.

Identifying modifier genes of monogenic disease: strategies and difficulties.

Substantial clinical variability is observed in many Mendelian diseases, so that patients with the same mutation may develop a very severe form of disease, a mild form or show no symptoms at all. Among the factors that may explain these differences in disease expression are modifier genes. In this paper, we review the different strategies that can be used to identify modifier genes and explain their advantages and limitations. We focus mainly on the statistical aspects but illustrate our points with a variety of examples from the literature.

Using linkage and association to identify and model genetic effects: summary of GAW15 Group 4.

Group 4 at Genetic Analysis Workshop 15 focused on methods that exploited both linkage and association information to map disease loci. All contributions considered the dichotomous trait of rheumatoid arthritis, using either affected sibpairs and/or unrelated controls. While one contribution investigated linkage and association approaches separately in genome-wide analyses, the remaining others focused on joint linkage and association methods in specific genomic regions. The latter contributions proposed new methods and/or examined existing methods that addressed whether one or more polymorphisms partially or fully explained a linkage signal, particularly the methods proposed by Li et al. that are implemented in the computer program Linkage and Association Modeling in Pedigrees (LAMP). Using simulated SNP data under linkage peaks, several contributions found that existing family-based association approaches such as those of Martin et al. and Lake et al. had power similar to LAMP and to several methods proposed by the contributors for testing that a single nucleotide polymorphism partially explains a linkage peak. In evaluating methods for identifying if a polymorphism or a set of polymorphisms fully accounted for a linkage signal, several contributions found that it was important to understand that these methods may be subject to low power in some situations and thus, a non-significant result was not necessarily indicative of the polymorphism(s) being fully responsible for the linkage signal. Finally, modeling the disease using association evidence conditional on linkage may improve understanding of the etiology of disease.

Discussing gene-gene interaction: warning--translating equations to English may result in jabberwocky.

Interest in mapping susceptibility alleles for complex diseases, which do not follow a classic single-gene segregation pattern, has driven interest in methods that account for, or use information from one locus when mapping another. Our discussion group examined methods related to epistasis or gene x gene interaction. The goal of modeling gene x gene interaction varied across groups; some papers tried to detect gene x gene interaction while others tried to exploit it to map genes. Most of the 10 papers summarized here applied newly created or newly modified statistical methods related to gene x gene interaction, while two groups primarily examined computational issues. As is often the case, comparisons are complicated by little overlap in the data used across the papers, and further complicated by the fact that the available data may not have been ideal for some gene x gene interaction methods. However, the main difficulty in comparing and contrasting methods across the papers is the lack of a consistent statistical definition of gene x gene interaction. But despite these issues, two clear trends emerged across the analyses: First, the methods for quantitative trait gene x gene interaction appeared to perform very well, even in families initially ascertained as affected sib pairs; and second, dichotomous trait gene x gene interaction methods failed to produce consistent results. The difficulty of using (primarily) affected sib pair data in a gene x gene interaction analysis is explored.