A direction-aware framework links transcription factor phosphosites to localization and transcriptional output

This dataset contains the processed data bundles and model artifacts supporting the PhosLoc-Transport framework for phosphorylation-dependent transcription factor localization analysis. The release includes curated human transcription factor phosphosite datasets, precomputed ESM-2 sequence embeddings, AlphaFold-derived structural inputs, finalized model artifacts, prediction outputs, and CPTAC validation resources. The functional data bundle supports Stage 1 FuncTransport training and inference, which prioritizes phosphosites with potential localization-regulatory activity using local sequence, central phosphosite, and structural context features. The direction-classifier bundle supports Stage 2 prediction of nuclear accumulation versus cytoplasmic redistribution among candidate localization-regulatory sites. The CPTAC source bundle supports downstream validation across cancer proteogenomic cohorts, including analyses linking predicted nuclear accumulation-promoting phosphosites to TF target-gene expression. Together, these files enable reproduction of the finalized PhosLoc-Transport training, prediction, figure-generation, and CPTAC validation workflows described in the accompanying code repository.