Yi-Fei Huang

The paper introduces ASSESS, a method to estimate the distribution of fitness effects from GWAS summary statistics, demonstrating its accuracy through simulations and application to human trait data.

DosaCNV predicts gene haploinsufficiency and deletion pathogenicity using deep multiple-instance learning, outperforming existing methods by integrating gene-level features.

This study examines how non-B DNA motifs affect sequencing error rates across three technologies and proposes a method to estimate false-positive SNVs.

This paper introduces DeepLOF, a deep learning model that improves essential gene prediction by integrating population and functional genomic data, identifying 109 previously undetected short essential genes.

The paper introduces a method to estimate RNA half-lives using PRO-seq and RNA-seq data, finding correlations between RNA stability and splicing, miRNA binding, DNA methylation, and histone modifications.

This paper introduces UNEECON, a deep learning framework that jointly predicts deleterious missense variants and constrained genes by integrating variant-level and gene-level information, showing improved performance in prioritizing variants associated with autosomal dominant disorders.

The paper introduces LASSIE, a model to estimate allele-specific fitness effects across human protein-coding sequences, showing its relevance to evolution and disease.

Scikit-ribo provides accurate A-site prediction and translation efficiency estimation from Riboseq and RNAseq data, correcting biases and improving TE estimation for over 2000 genes in S. cerevisiae.

This study uses PRO-seq to analyze the rapid transcriptional response to celastrol in K562 cells, revealing two waves of transcriptional changes and identifying key transcription factors involved.