Introduction:
Congenital Anomalies of Kidney and Urinary Tract (CAKUT) account for a substantial number of pediatric malformations frequently observed in clinic, and it is a major cause of kidney failure in children and young adults. CAKUT results from aberrant embryonic development through genetic or environmental drivers. Over a hundred genes have been implicated in CAKUT, many of which cause other genetic syndromes with multiple congenital anomalies. Although CAKUT is usually diagnosed prenatally, many cases remain undiagnosed until later in life.

Methods:
Given the high genetic and phenotypic heterogeneity of CAKUT, this study sought to retrospectively evaluate the diagnostic efficacy of clinical exome sequencing (cES) in affected individuals, most with syndromic CAKUT, by annotating 556 cases with clear indications of CAKUT with Human Phenotype Ontology (HPO) terms. We reviewed the diagnosis of each case based on updated pathogenicity classification by a board-certified clinical geneticist, the number of findings, and the expected mode of inheritance. Finally, we employed a machine learning-based approach to integrate the annotations of multiple genomic knowledge sources to discover novel phenotypic expansions involving CAKUT.

Results:
Retrospectively, we analyzed phenotypic and genetic findings of 556 individuals with CAKUT referred to a diagnostic laboratory for cES. After variant pathogenicity reanalysis, 29.3% (163/556) of cases had a definitive/probable molecular finding in a gene that explained most/all their presenting phenotypes, but only 64.4% (105/163) of these were findings in an established CAKUT gene. Although most cases, 96.9% (539/556) had abnormalities in other organ systems, we found no statistically significant difference (p-value >0.05) in cES efficacy among age, sex, specific CAKUT phenotypes, or presenting organ system abnormalities (including syndromic vs isolated). Based on these results, we recommend that cES be considered for all individuals with CAKUT. Understanding the molecular basis of individuals with CAKUT may inform further testing and clinical management in this patient population.



Next, we leveraged our database of ~17,000 clinical exome reports and publicly available genomic knowledge to identify disease genes, previously unrecognized to cause CAKUT. We hypothesized that genes involved in the same phenotypes will share biological pathways, expression, and interactions. Therefore, they will be similarly annotated across genomic knowledge sources. First, we trained a published machine learning algorithm using a curated list of known CAKUT genes to compare the annotations between our training set and all RefSeq genes. Scores were cross-validated genome-wide using a leave-one-out approach and represent the ranked similarity for our phenotype-specific training set. Using these scores, in conjunction with published case reports, available mouse models, and cES cases from our cohort with matching diagnoses and CAKUT phenotypes, there is sufficient evidence to suggest that deleterious variants in ADNP, CBL, EFTUD2, L1CAM, MED13L, PCNT, PHIP, SETD1A, SETD5, and WNT7A are associated with CAKUT.

Conclusion:
Our findings demonstrate the diagnostic utility of cES in patients with CAKUT across different organ system presentations and broaden the phenotypic features associated with genes already known to cause human genetic disorders.