Protein Subcellular-Localisation Prediction

Where in the cell does a protein work - and can its amino-acid sequence tell us? This project predicts the compartments of proteins (10 locations, several per protein) from 156 engineered sequence features with random forests, gradient boosting and stacking, and compares them with composition, Smith-Waterman alignment and an ESM-2 embedding on the DeepLoc 2.0 Swiss-Prot set. A random split flatters copying the closest relative (macro MCC 0.41, falling to 0.15 across homology partitions), while the stacking ensemble with ESM-2 keeps 0.53; on human proteins localised by imaging (Human Protein Atlas) every method is much weaker (best 0.19).

Category: Bioinformatics - Proteomics and Protein Bioinformatics

Proteomics & Protein Bioinformatics Project B22 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.