Docker Container Execution

docker-container-execution

Summary

Execute a containerized inference model (Mass2SMILES) on spectral data by mounting input/output directories and passing MGF files to a Docker container. This skill encapsulates deployment, parameter tuning, and batch inference for MS/MS-to-structure prediction without local dependency installation.

When to use

You have GNPS-style MGF spectral files as input and need to run Mass2SMILES MS/MS-to-structure inference without installing TensorFlow, CUDA, or Python dependencies locally. Use this when you want reproducible, containerized execution with tunable CPU thread allocation (via the cpu_threads parameter) on a TensorFlow-CPU or GPU build.

When NOT to use

• Your spectral data is not in MGF format (e.g., mzML, NetCDF, or proprietary vendor formats) without prior conversion.
• You require GPU-accelerated inference and have CUDA driver incompatibility issues that are unresolved (note: the README reports CUDA driver issues motivated the TensorFlow-CPU build).
• You need real-time or streaming inference rather than batch processing of complete MGF files.

Inputs

• MGF file (GNPS-style spectral data)
• input directory path (host machine)
• output directory path (host machine)
• cpu_threads parameter (integer, optional)

Outputs

• predicted SMILES structures (text output in output directory)
• inference log or report (optional)

How to apply

First, pull or load the delser292/mass2smiles:final Docker container (or build from the Dockerfile in Zenodo 14778327). Prepare an input directory containing your GNPS-style MGF files and designate an output directory for predicted SMILES structures. Mount both directories as volumes when running the container. If using the TensorFlow-CPU build, instantiate the InferenceModel with a configurable cpu_threads parameter (e.g., cpu_threads=128) to tune CPU core allocation for your hardware. Execute the container with the mounted paths and MGF filename, then collect the predicted SMILES output files from the mounted output directory.

Related tools

• Docker (container runtime for reproducible isolation and deployment of the Mass2SMILES inference engine with pre-configured dependencies) — https://www.docker.com
• delser292/mass2smiles:final (containerized Mass2SMILES model with TensorFlow-CPU or GPU inference backend) — https://zenodo.org/records/7883491
• TensorFlow (deep learning framework underlying the Mass2SMILES transformer model; TensorFlow-CPU build is used to avoid CUDA driver incompatibilities)
• Python (scripting language for the Mass2SMILES inference pipeline (mass2smiles_transformer.py)) — https://github.com/volvox292/mass2smiles

Examples

docker run -v c:/your_path/to_the_folder/mass2smiles/:/app mass2smiles:transformer_v1 conda run -n tf python app/mass2smiles_transformer.py your_mgf_file.mgf /app

Evaluation signals

• Output directory contains one or more SMILES structure files corresponding to each input MGF file (1:1 mapping).
• Predicted SMILES strings are valid (match SMILES grammar and can be parsed by chemistry libraries like RDKit).
• Actual CPU thread count observed during inference execution matches or is ≤ the configured cpu_threads parameter (verification via system monitoring or container logs).
• Inference completes without Docker runtime errors or OOM (out-of-memory) exceptions.
• Output file timestamps are later than container execution start time, confirming fresh inference rather than cached results.

Limitations

• CUDA driver incompatibilities motivated the TensorFlow-CPU build, resulting in slower inference compared to GPU execution; the article notes 'inference speed is highly improved' only when GPU is available and properly configured.
• The container requires valid directory mount paths; misconfigured host paths will cause runtime failures or produce empty output directories.
• No changelog is available; version pinning and reproducibility tracking across updates (e.g., between Zenodo 7883491 and 14778327) may be unclear.
• MGF files must follow GNPS-style format conventions; non-standard or malformed spectra may cause silent failures or degraded prediction quality.

Evidence

• [readme] Spectral data can be provided as MGF files (GNPS-syle) and model inference is most effciently performed via the provided docker container.: "Spectral data can be provided as MGF files (GNPS-syle) and model inference is most effciently performed via the provided docker container."
• [readme] the container is available as tarball in supplementary or via docker pull delser292/mass2smiles:final: "the container is available as tarball in supplementary or via docker pull delser292/mass2smiles:final"
• [intro] InferenceModel accepts a configurable cpu_threads parameter to adjust CPU cores used during inference: "The InferenceModel accepts a configurable cpu_threads parameter (e.g., cpu_threads=128) to adjust the number of CPU cores used during inference"
• [readme] You need to point to your input and output dir, now the mass2smiles model is built into the container.: "You need to point to your input and output dir, now the mass2smiles model is built into the container."
• [readme] the cddd does not seem to work on newer cuda drivers, therefore it is build using tensorflow cpu: "the cddd does not seem to work on newer cuda drivers, therefore it is build using tensorflow cpu"