Title Information
Title
Genome-wide annotation of functional branchpoints in the human genome
Type of Resource (primo)
dissertations
Name: Personal
Name Part
Zhong, Yu
Role
Role Term: Text
creator
Name: Personal
Name Part
Singh, Ritambhara
Role
Role Term: Text
Reader
Name: Personal
Name Part
Bailey, Jeff
Role
Role Term: Text
Reader
Name: Personal
Name Part
Fairbrother, William
Role
Role Term: Text
Advisor
Name: Corporate
Name Part
Brown University. Center for Computational Molecular Biology
Role
Role Term: Text
sponsor
Origin Information
Copyright Date
2026
Physical Description
Extent
1, 35 p.
digitalOrigin
born digital
Note: thesis
Thesis (A. M.)--Brown University, 2026
Genre (aat)
theses
Abstract
Branchpoints (BPs) are pivotal control elements in RNA splicing, yet they remain the most elusive of the core splicing signals. Comprehensive annotation of the human branchpoint landscape has been hindered by the technical biases of enzymatic mapping and the scarcity of training data for the minor (U12-type) spliceosome. Here, we present DeepEnsemble, an ensemble-based deep learning framework that integrates sequence and genomic features to resolve the "splicing code" of both major and minor introns. By employing a transfer learning strategy, DeepEnsemble overcomes data sparsity to achieve unprecedented accuracy in predicting U12-type BPs, revealing strict evolutionary constraints distinct from the degenerate motifs of the major spliceosome. We rigorously validated our predictions using Massively Parallel Reporter Assays (MPRAs), systematically testing thousands of introns and variants. This experimental audit confirmed the existence of functional BPs in over 60% of introns lacking prior annotation and demonstrated a 70% concordance between our model and in vivo lariat formation. Furthermore, we developed DeepEnsemble-LR, an interpretable extension that integrates in silico mutagenesis with evolutionary conservation to prioritize pathogenic splicing variants, outperforming existing state-of-the-art methods. Our study provides a unified, experimentally validated atlas of human branchpoints and a robust tool for decoding the clinical impact of non-coding variation.
Subject (fast) (authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/01086257")
Topic
RNA splicing
Subject (fast) (authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/02032663")
Topic
Deep learning (Machine learning)
Language
Language Term (ISO639-2B)
English
Record Information
Record Content Source (marcorg)
RPB
Record Creation Date (encoding="iso8601")
20260427