OutCyte 3.0 - reproduction package

Code and data to reproduce the analysis reported in the OutCyte 3.0 manuscript:prediction of unconventionally secreted proteins (UPS) from sequence with anensemble of four protein language models (ESM-2, Ankh, ESM-C 300M/600M).Everything is in the single archive below; nothing else needs downloading. To apply OutCyte 3.0 to your own sequences, use the maintained standalone tool: Tool: https://github.com/JaVanGri/OutCyte-3.0 Web: outcyte.proteome.hhu.de Getting started (Python 3.12-3.13) unzip outcyte3.zip cd outcyte3 # create a fresh environment -- venv: python3.13 -m venv .venv && source .venv/bin/activate # ...or conda: # conda create -n outcyte3 python=3.13 && conda activate outcyte3 pip install -r requirements.txt jupyter lab notebooks/3_training_evaluation.ipynb Install into a fresh environment, not a conda base env or an existing scientificstack - dependencies are pinned and will otherwise up-/downgrade your numpy andpandas. No configuration required - the code finds its data next to it. Thenotebooks run independently. See outcyte3/README.md in the archive for what eachnotebook does. Licence: CC BY-NC-SA 4.0 (the NonCommercial clause is inherited from Ankh; seeLICENSE in the archive). Corresponding author: Gereon Poschmann (ORCID: 0000-0003-2448-0611).