Baidu PaddlePaddle's Wenxin Biological Computing Large Model Upgraded with Significant Improvement in Conformation Prediction Accuracy
-
Baidu PaddlePaddle's Propeller team announced upgrades to two biological computing large models: the protein-small molecule docking conformation prediction model HelixDock and the protein-protein complex structure prediction model HelixFold-Multimer.
The accuracy of these two models has been significantly improved, providing a more reliable foundation for structure-based drug design. HelixDock achieves an accuracy of up to 89% on the PDBBind core set and outperforms other methods, including DeepMind's AlphaFold, on more challenging datasets.
HelixFold-Multimer has made breakthrough progress in the prediction of antigen-antibody and peptide-protein complexes, with significantly higher accuracy compared to other methods.
The Baidu PaddlePaddle Helix team has also stated that they will make these models available to the public, hoping to promote technological innovation in the field of biological computing.
PaddlePaddle Helix access address: https://paddlehelix.baidu.com/