ML Model for Rare Genetic Variants
Abstract
Implementation of a convolutional neural network trained on 50,000 genomes to identify variants of uncertain significance in non-coding regions. 94% accuracy in cross-validation.
Classification
Scientific Community
2 voicesDr. Yuki Tanaka
🇯🇵 Japón
94% accuracy on non-coding regions is remarkable. Our lab in Tokyo has been struggling with this exact problem. Would love to collaborate and test this model on our rare disease cohort.
Dra. Valentina Rossi
🇮🇹 Italia
Ottimo lavoro! La validazione su 50.000 genomi è un dataset impressionante. Come gestite il problema del class imbalance tra varianti patogeniche e benigne?
