matblum/tmbed repository overview

⁠TMbed

This Docker image runs TMbed, a tool for predicting transmembrane proteins using the ProtT5 protein language model.

⁠Downloading the model

The image does not include the model files by default. You must download them separately and provide the model directory path when running predictions.

The following command downloads the model to a t5 subdirectory in your current working directory:

docker run --rm -v $PWD/t5:/t5 matblum/tmbed:1.0.2 download -m /t5

⁠Predicting transmembrane proteins

First, download a few example protein sequences. This will create a file named sample.fasta in your current directory:

curl -o sample.fasta 'https://rest.uniprot.org/uniprotkb/accessions?accessions=P21796%2CP45880%2CP57764%2CQ96QA5%2CQ9Y277&format=fasta'

Next, run TMbed to predict transmembrane regions:

docker run --rm -v $PWD/t5:/t5 -v $PWD:/work matblum/tmbed:1.0.2 predict -f /work/sample.fasta -m /t5 -p /work/sample.pred

This writes the predictions to sample.pred in your working directory.

⁠More Information

For more information and available command-line options, visit the TMbed GitHub repository⁠.

⁠Citation

If you use TMbed in your work, cite the following publication: