Yaqi Su

predNMD improves clinical variant pathogenicity classification by predicting nonsense-mediated mRNA decay (NMD) more accurately than the 50nt rule, using a random forest model trained on diverse datasets.

This study develops GLM-Missense, a genomic language model fine-tuned for missense variant impact prediction, showing it captures nucleotide-level effects missed by existing predictors.

This study benchmarks eleven isoform detection methods for TGS data, finding StringTie2 and Bambu excel in sensitivity and precision, respectively.