PhosLoc-Transport
PhosLoc-Transport is a two-stage computational framework for prioritizing transcription factor phosphosites that may regulate nuclear transport.
GitHub: https://github.com/TaoYySC/phosloc-transport
Overview
The framework combines sequence-based and structure-aware machine learning to connect phosphorylation sites with subcellular localization changes and downstream transcriptional regulation.
| Module | Task |
|---|---|
| Localization-Regulatory Classifier | Identifies candidate localization-regulatory phosphosites |
| Localization Direction Classifier | Predicts nuclear accumulation vs. cytoplasmic redistribution |
| CPTAC analysis | Validates predicted nuclear accumulation-associated sites |
Key features
- Integrates local sequence, central phosphosite, and AlphaFold-derived structural features
- Two-stage prediction pipeline for regulatory potential and localization direction
- CPTAC-based validation against tumor multi-omics target-gene regulation
- Reproducible training, prediction, and analysis workflows with documented run settings
Highlights
- Built a curated benchmark of transcription factor phosphosites with localization-related evidence
- Achieved AUROC of 0.827 on a filtered held-out benchmark
- Integrated CPTAC phosphoproteomic, proteomic, and RNA-seq data for downstream validation
Associated work
Associated manuscript: A direction-aware framework links transcription factor phosphosites to localization and transcriptional output.
Processed data and model artifacts are available on Zenodo (DOI: 10.5281/zenodo.21064685).