[R01] Predicting, Interpreting, and Validating Noncoding Variants for Rare Disease Diagnosis
Ente: National Library of Medicine
Scadenza: 2031-08-31
Importo max: $716,515
Paese: US
Descrizione
Abstract:
Despite major advances in whole-exome and whole-genome sequencing, up to half of rare Mendelian disease
cases remain undiagnosed. A key limitation is the inability to interpret noncoding variants, particularly those
that alter RNA splicing or translation. Existing AI predictors often overestimate accuracy due to unrealistic
benchmarks, lack interpretability, and are not integrated with experimental or clinical validation. To address
these issues, this project will create clinically realistic benchmarks, develop and further optimize AI-based
variant interpretation methods, and systematically validate results through high-throughput assays, thereby
improving diagnostic yield and mechanistic understanding of disease. The optimized AI methods will be
applied to >50,000 unsolved patient genomes, with top candidates validated using a high-throughput assays
that can measure changes in RNA splicing, RNA G-quadruplexes (rG4s) formation, and 5' UTR effect on
protein level. All code, data, and benchmarks will be openly released as community resources. Overall, this
work will directly enhance the diagnostic utility of genome sequencing, provide interpretable AI frameworks for
noncoding variant assessment, and uncover new mechanisms of disease. By bridging computational modeling,
experimental validation, and clinical prioritization, this project will accelerate rare disease diagnosis and lay the
foundation for RNA-based therapeutic discovery.
Istituzione: UNIVERSITY OF PENNSYLVANIA
PI: Yoseph Barash
Progetto: 1R01LM015391-01
Settori: National Library of Medicine
Vai al bando originale
Registrati gratis su Bandolo per trovare bandi compatibili con la tua azienda.