LLM-assisted PTM hypothesis generation from proteomics mass spectrometry data. Identifies delta masses in peptide fragmentation patterns and uses an OpenAI-compatible LLM to annotate and interpret them.
# requires uv — https://astral.sh/uv
uv syncFor the ADEPT browser UI and full agentic workflow:
uv sync --extra adeptUsing PTMdiscoverer's validate, run, and analyze commands on example test data:
export OPENAI_API_KEY=your_key
export OPENAI_MODEL=your_model_name
export OPENAI_BASE_URL=your_base_url
uv run ptmdiscoverer validate tests/fixtures/example_study
uv run ptmdiscoverer run tests/fixtures/example_study
uv run ptmdiscoverer analyze --run-dir tests/fixtures/example_study/output/output_<timestamp>note: the environment variables can also be set in a .env file instead of exported directly.
See the env.template
For the browser UI:
cp .env.template .env # fill in INTERNAL_LLM_* vars
./scripts/start-app.sh # opens http://localhost:8501→ Full usage guide — CLI reference, app workflow, data requirements, and development setup. Also see the API docs
This research was funded by the Generative AI for Science, Energy, and Security Science & Technology Investment under the Laboratory Directed Research and Development Program at Pacific Northwest National Laboratory (PNNL), a multiprogram national laboratory operated by Battelle for the U.S. Department of Energy.
Portions of this work were supported by the Center for AI at PNNL, the NW-BRaVE for Biopreparedness project funded by the U.S. Department of Energy (DOE), Office of Science, Biological and Environmental Research program, under FWP 81832, and performed at the Environmental Molecular Sciences Laboratory (EMSL), a DOE Office of Science User Facility sponsored by the Biological and Environmental Research program under Contract No. DE-AC05-76RL01830.
Generative AI tools were used to assist with software development.
aivett.bilbao@pnnl.gov · august.george@pnnl.gov
Software for version 1.0: https://doi.org/10.5281/zenodo.20030405