Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Biotechnology Science and Engineering

Committee Chair

Joseph Ng

Committee Member

Matthew Niemiller

Committee Member

Paul Wolf

Committee Member

Greg Barsh


Exons (Genetics), Developmental neurobiology--Genetic aspects, Developmental disabilities--Genetic aspects


Genetic variants with poor annotation or unknown biological consequence are often systematically excluded in genomic analysis. Here I analyze variants near "poison exons” which are often absent from standard gene annotations and are alternative exons whose inclusion results in a premature termination codon. Neurodevelopmental disorders (NDDs) often result from rare genetic variation, but the genomic testing yield for NDDs remains around 50%, suggesting some clinically relevant rare variants may be missed by standard analyses. Variants that alter poison exons inclusion can lead to loss-of-function and may be highly penetrant contributors to disease. I analyzed variants found by genome sequencing in multiple NDD cohorts, and across 2,999 probands, I found six clinically relevant variants in poison exon regions that were previously overlooked. With only a minimal increase in variant analysis burden (most probands had zero or one candidate poison exon variants in a known NDD gene, with an average of 0.77 per proband), annotation of poison exons can improve diagnostic yield for NDDs and likely other congenital conditions. Introns containing poison exons were not broadly enriched for variation in the SFARI Autism Spectrum Disorder (ASD) cohort as compared to the unaffected siblings of children with ASD. This analysis did show that probands were more likely to have qualifying variants in poison exons in voltage-gated sodium and calcium channels (with and odds ratio of 1.719), but the difference was not statistically significant. These results show that ASD phenotype cannot be categorically attributed to variation in poison exons. Given the diversity of the ASD phenotype and the rarity of poison exon contribution in the NDD cohorts examined (~0.2%), it is possible that a small subset of ASD probands with NDD phenotype overlap may have causative poison exon variants, but this conclusion cannot be drawn from this analysis. This investigation also brought to light splicing variants that escape conventional detection because they do not meet pathogenic evidence criteria. I used SpliceAI, a deep neural network splice variant predictor, to analyze variants found in the KidsCanSeq pediatric cancer cohort. My investigation revealed one variant in DICER1, the gene responsible for pleuropulmonary blastoma, which inspired a similar evaluation of variants submitted to ClinVar. This revealed a shared variant in two unrelated probands with thyroid tumors, another phenotype of DICER1 loss-of-function. These analyses provide evidence for the addition of poison exons to conventional genome interpretation pipelines in NDD cohorts and the supplementation of intronic and synonymous variant curation with computational splice prediction.

Available for download on Sunday, April 20, 2025