PANDA-3D: protein function prediction based on AlphaFold models


Please provide an email address for receiving the prediction results.
Email address: 



Please paste your AlphaFold protein structures in PDB files (up to 1000 models per submission): 

File:

      

Example: Click here for an example input file. Click here for the corresponding output file.

PANDA-3D GitHub Repository


Click here for the local version of PANDA-3D.

Training, validation, and testing datasets


Datasets are available at http://dna.cs.miami.edu/PANDA-3D/download_files/alphafoldDB_pdb_1115/.

Proteins are randomly split into training (80%), validation (10%), and testing (10%). The UniProt IDs for training, validation, and testing are saved into train_0130_df.pkl, valid_0130_df.pkl, and test_0130_iden_pd3_dpfri.pkl, respectively.

The protein features, including UniProt ID, true GO annotations, 3D coordinates, GO term label, pLDDT scores, amino acid sequence, ESM features, and sequence length, are saved into a pickle file in the format of a Python dictionary. These features are further divided into subfolders based on the first two letters of the protein ID. For example, the features of protein A0A0A1C3I2 can be found in A0/A0A0A1C3I2.pkl.

Cite PANDA-3D


Zhao, C., Liu, T., and Wang, Z. (2023) PANDA-3D: protein function prediction based on AlphaFold models. Under Review.

Contact


If you have any question, please concact
Dr. Zheng Wang
Department of Computer Science
College of Arts and Sciences
University of Miami