Researchers have developed DeepVRegulome, a deep-learning framework designed to predict the functional impact of short genomic variants on the human regulome. This framework integrates 464 fine-tuned DNABERT models and employs quantitative scoring, attention-based motif analysis, and survival analysis to assess variant effects and link them to clinical outcomes. Benchmarked against experimental data and existing predictors, DeepVRegulome identified thousands of mutations affecting transcription factor binding and splice sites, with a subset linked to patient survival in glioblastoma samples. The framework's code, models, and a data portal are publicly available. AI
IMPACT This framework could improve the understanding and prioritization of non-coding genetic variants, potentially aiding in the diagnosis and treatment of diseases.
RANK_REASON The item describes a new computational framework and associated models published as a research paper on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
