Gene-set analysis (GSA) evaluates the overall evidence of association between a phenotype and all genotyped single nucleotide polymorphisms (SNPs) in a set of genes, as opposed to testing for association between a phenotype and each SNP individually. We propose using the Gamma Method (GM) to combine gene-level P-values for assessing the significance of GS association. We performed simulations to compare the GM with several other self-contained GSA strategies, including both one-step and two-step GSA approaches, in a variety of scenarios. We denote a 'one-step' GSA approach to be one in which all SNPs in a GS are used to derive a test of GS association without consideration of gene-level effects, and a 'two-step' approach to be one in which all genotyped SNPs in a gene are first used to evaluate association of the phenotype with all measured variation in the gene and then the gene-level tests of association are aggregated to assess the GS association with the phenotype. The simulations suggest that, overall, two-step methods provide higher power than one-step approaches and that combining gene-level P-values using the GM with a soft truncation threshold between 0.05 and 0.20 is a powerful approach for conducting GSA, relative to the competing approaches assessed. We also applied all of the considered GSA methods to data from a pharmacogenomic study of cisplatin, and obtained evidence suggesting that the glutathione metabolism GS is associated with cisplatin drug response.

Height is a heritable and highly heterogeneous trait. Short stature affects 3% of the population and in most cases is genetic in origin. After excluding known causes, 67% of affected individuals remain without diagnosis. To identify novel candidate genes for short stature, we performed exome sequencing in 254 unrelated families with short stature of unknown cause and identified variants in 63 candidate genes in 92 (36%) independent families. Based on systematic characterization of variants and functional analysis including expression in chondrocytes, we classified 13 genes as strong candidates. Whereas variants in at least two families were detected for all 13 candidates, two genes had variants in 6 (UBR4) and 8 (LAMA5) families, respectively. To facilitate their characterization, we established a clustered network of 1025 known growth and short stature genes, which yielded 29 significantly enriched clusters, including skeletal system development, appendage development, metabolic processes, and ciliopathy. Eleven of the candidate genes mapped to 21 of these clusters, including CPZ, EDEM3, FBRS, IFT81, KCND1, PLXNA3, RASA3, SLC7A8, UBR4, USP45, and ZFHX3. Fifty additional growth-related candidates we identified await confirmation in other affected families. Our study identifies Mendelian forms of growth retardation as an important component of idiopathic short stature.

High throughput sequencing has greatly advanced disease gene identification, especially in heterogeneous entities. Despite falling costs this is still an expensive and laborious technique, particularly when studying large cohorts. To address this problem we applied Exome Pool-Seq as an economic and fast screening technology in neurodevelopmental disorders (NDDs). Sequencing of 96 individuals can be performed in eight pools of 12 samples on less than one Illumina sequencer lane. In a pilot study with 96 cases we identified 27 variants, likely or possibly affecting function. Twenty five of these were identified in 923 established NDD genes (based on SysID database, status November 2016) (ACTB, AHDC1, ANKRD11, ATP6V1B2, ATRX, CASK, CHD8, GNAS, IFIH1, KCNQ2, KMT2A, KRAS, MAOA, MED12, MED13L, RIT1, SETD5, SIN3A, TCF4, TRAPPC11, TUBA1A, WAC, ZBTB18, ZMYND11), two in 543 (SysID) candidate genes (ZNF292, BPTF), and additionally a de novo loss-of-function variant in LRRC7, not previously implicated in NDDs. Most of them were confirmed to be de novo, but we also identified X-linked or autosomal-dominantly or autosomal-recessively inherited variants. With a detection rate of 28%, Exome Pool-Seq achieves comparable results to individual exome analyses but reduces costs by >85%. Compared with other large scale approaches using Molecular Inversion Probes (MIP) or gene panels, it allows flexible re-analysis of data. Exome Pool-Seq is thus well suited for large-scale, cost-efficient and flexible screening in characterized but heterogeneous entities like NDDs.

Acrocallosal syndrome (ACLS) is an autosomal recessive neurodevelopmental disorder caused by KIF7 defects and belongs to the heterogeneous group of ciliopathies related to Joubert syndrome (JBTS). While ACLS is characterized by macrocephaly, prominent forehead, depressed nasal bridge, and hypertelorism, facial dysmorphism has not been emphasized in JBTS cohorts with molecular diagnosis. To evaluate the specificity and etiology of ACLS craniofacial features, we performed whole exome or targeted Sanger sequencing in patients with the aforementioned overlapping craniofacial appearance but variable additional ciliopathy features followed by functional studies. We found (likely) pathogenic variants of KIF7 in 5 out of 9 families, including the original ACLS patients, and delineated 1000 to 4000-year-old Swiss founder alleles. Three of the remaining families had (likely) pathogenic variants in the JBTS gene C5orf42, and one patient had a novel de novo frameshift variant in SHH known to cause autosomal dominant holoprosencephaly. In accordance with the patients' craniofacial anomalies, we showed facial midline widening after silencing of C5orf42 in chicken embryos. We further supported the link between KIF7, SHH, and C5orf42 by demonstrating abnormal primary cilia and diminished response to a SHH agonist in fibroblasts of C5orf42-mutated patients, as well as axonal pathfinding errors in C5orf42-silenced chicken embryos similar to those observed after perturbation of Shh signaling. Our findings, therefore, suggest that beside the neurodevelopmental features, macrocephaly and facial widening are likely more general signs of disturbed SHH signaling. Nevertheless, long-term follow-up revealed that C5orf42-mutated patients showed catch-up development and fainting of facial features contrary to KIF7-mutated patients.

Although single-locus approaches have been widely applied to identify disease-associated single-nucleotide polymorphisms (SNPs), complex diseases are thought to be the product of multiple interactions between loci. This has led to the recent development of statistical methods for detecting statistical interactions between two loci. Canonical correlation analysis (CCA) has previously been proposed to detect gene-gene coassociation. However, this approach is limited to detecting linear relations and can only be applied when the number of observations exceeds the number of SNPs in a gene. This limitation is particularly important for next-generation sequencing, which could yield a large number of novel variants on a limited number of subjects. To overcome these limitations, we propose an approach to detect gene-gene interactions on the basis of a kernelized version of CCA (KCCA). Our simulation studies showed that KCCA controls the Type-I error, and is more powerful than leading gene-based approaches under a disease model with negligible marginal effects. To demonstrate the utility of our approach, we also applied KCCA to assess interactions between 200 genes in the NF-κB pathway in relation to ovarian cancer risk in 3869 cases and 3276 controls. We identified 13 significant gene pairs relevant to ovarian cancer risk (local false discovery rate <0.05). Finally, we discuss the advantages of KCCA in gene-gene interaction analysis and its future role in genetic association studies.