Machine-learning prediction of residues driving homotypic transmembrane interactions.

bioinformatics, protein, transmembrane, residue, conservation, coevolution, covariance, evolutionary, couplings, polarity, hydrophobicity, randomforest, machinelearning, interface, LIPS, evolution
pip install thoipapy==1.2.0



The Transmembrane HOmodimer Interface Prediction Algorithm (THOIPA) is a machine learning method for the analysis of protein-protein-interactions.

THOIPA predicts transmembrane homodimer interface residues from evolutionary sequence information.

THOIPA helps predict potential homotypic transmembrane interface residues, which can then be verified experimentally. THOIPA also aids in the energy-based modelling of transmembrane homodimers.

Important links:

How does thoipapy work?

  • downloads protein homologues with BLAST
  • extracts residue properties (e.g. residue conservation and polarity)
  • trains a machine learning classifier
  • validates the prediction performance
  • creates heatmaps of residue properties and THOIPA prediction


pip install thoipapy

THOIPA has only been tested on Linux, due to reliance on external dependencies such as FreeContact, Phobius, CD-HIT and rate4site. For predictions only, a dockerised version is available that runs on Windows or MacOS. Please see the THOIPA webserver for the latest information.


We recommend the Anaconda python distribution, which contains all the required python modules (numpy, scipy, pandas,biopython and matplotlib). THOIPApy is currently tested for python 3.8.5. The requirements.txt contains a snapshot of compatible dependencies.

Development status

The code has been extensively updated and annotated for public release. However is released "as is" with some known issues, limitations and legacy code.

Usage as a standalone predictor

from thoipapy.thoipa import get_md5_checksum, run_THOIPA_prediction
from thoipapy.utils import make_sure_path_exists

protein_name = "ERBB3"
out_dir = "/path/to/your/desired/output/folder"
md5 = get_md5_checksum(TMD_seq, full_seq)
run_THOIPA_prediction(protein_name, md5, TMD_seq, full_seq, out_dir)

Example Output

  • the output includes a csv showing the THOIPA prediction for each residue, as well as a heatmap figure as a summary
  • below is a heatmap showing the THOIPA prediction, and underlying conservation, relative polarity, and coevolution

Create your own machine learning predictor

  • THOIPA can be retrained to any dataset of your choice
  • the original set of training sequences and other resources are available via the Open Science Foundation
  • the THOIPA feature extraction, feature selection, and training pipeline is fully automated
  • contact us for an introduction to the THOIPA software pipeline and settings
python path/to/thoipapy/ -s home/user/thoipa/THOIPA_settings.xlsx


THOIPApy is free software distributed under the permissive MIT License.


  • Contributors are welcome.
  • For feedback or troubleshooting, please email us directly and initiate an issue in Github.



Yao Xiao, Bo Zeng, Nicola Berner, Dmitrij Frishman, Dieter Langosch, and Mark George Teese (2020) Experimental determination and data-driven prediction of homotypic transmembrane domain interfaces, Computational and Structural Biotechnology Journal