Healthcare Aug 30, 2021 10:24 AM (GMT+8) · EqualOcean
Recently, Xinhua bio Pharmaceutical (Guangzhou) Co., Ltd. (hereinafter referred to as "Xinhua bio") released the vibrantfold platform for antibody structure prediction, which is based on the Bumblebee deep learning framework independently developed by Xinhua bio. Xinhua biology plans to open the platform to global academic and industrial users for accurate antibody structure prediction. Vibrantfold conducted internal tests with two other international top protein prediction platforms alphafold2 and rosetafold before release. By comparing the differences between the predicted structure and the experimental structure, the three structure prediction platforms are scored. In the internal test, the prediction of most antibody structures can be completed within 1 minute. Vibrantfold's prediction accuracy has been evaluated and verified by structural biologists. Professor Zhang Linqi's team of Tsinghua University, Professor Wang Xinquan's team and researcher Zhou Qiang's team of West Lake University provided unpublished antibody structures for prediction. The results show that the three prediction platforms can achieve high-precision structure prediction. The antibody structure prediction accuracy of vibrantfold is comparable to alphafold2 and rosettafold, and has reached the global leading level. What's more exciting is that the prediction speed of vibrantfold is unparalleled, and the calculation efficiency is several orders of magnitude higher than the above two software. This feature makes the construction and application of large-scale virtual antibody structure library possible.