Welcome to MEDUSA webserver

Cluster workload %

Multiclass flexibility prediction from sequences of amino acids

MEDUSA is a Deep Learning approach for prediction of protein flexibility from sequence. MEDUSA takes as input an amino acid sequence and returns a flexibility class of each residue in terms of the expected normalized B-factor value range. Prediction is simultaneously performed in two-, three- and five classes using a convolutional neural network trained on a dataset of non-redundant X-ray structures.

See an example of the webserver's output here.

Please use the following reference when citing the MEDUSA webserver:

Vander Meersche, Y., Cretin, G., de Brevern, A. G., Gelly, J. C., & Galochkina, T. (2021). MEDUSA: Prediction of protein flexibility from sequence. Journal of Molecular Biology, 166882. https://doi.org/10.1016/j.jmb.2021.166882

For now, the webserver works for single sequence prediction only. To launch batches of predictions please use the standalone program (Docker container) freely available on github. We provide a simple example of few bash lines to run MEDUSA on multiple sequences (multifasta).
Input sequence
Paste sequence example]:

Or, browse a sequence file:
Job options
The database used for HHBlits search is now: UniRef30_2021_03 (Previously UniRef30_2020_03)
The one used in the reference paper of MEDUSA was UniRef30_2016_09
This web server is for an academic non-commercial use only, for an other use please contact us.
Using MEDUSA web server means that you have read and you agree to comply with the following terms of use.
Total number of jobs submitted (since feb-2021):
Number of jobs submitted last 7 days:
Number of unique submitters: