Baidu Wenxin large models was recognized by the international top magazine again! Heavyweight Biocomputing Achievements Published in the Nature Sub issue

Tech 2023-10-11 00:26:14 Source: Network

On the evening of October 9, Machine Intelligence, a subsidiary of the international top academic journal Nature, published another major achievement of the Wenxin big model of biological computing, Amethodfor, which was developed by Baidu PaddlePaddle propeller and BiotechMultiple sequence alignment free protein instruction prediction using a protein language model "proposed the world's first open source and online service, HelixFold Single, for protein structure prediction without the need for MSA input.This study is another breakthrough achievement in the protein field by Baidu, following two heavyweight efforts in the field of biological computing, HelixGEM and LinearDesign

On the evening of October 9, Machine Intelligence, a subsidiary of the international top academic journal Nature, published another major achievement of the Wenxin big model of biological computing, Amethodfor, which was developed by Baidu PaddlePaddle propeller and Biotech
Multiple sequence alignment free protein instruction prediction using a protein language model "proposed the world's first open source and online service, HelixFold Single, for protein structure prediction without the need for MSA input.

This study is another breakthrough achievement in the protein field by Baidu, following two heavyweight efforts in the field of biological computing, HelixGEM and LinearDesign. This work broke the speed bottleneck of mainstream MSA retrieval models such as AlphaFold2, and increased the average protein structure prediction speed by hundreds of times, achieving second level prediction. At the same time, this achievement also brings a protein structure prediction solution with lower usage barriers and wider applicability to various industries, universities, and research institutes, which is expected to promote further development in fields such as life sciences, biopharmaceuticals, and protein research in China.

In recent years, AI has been committed to breaking through the problem of protein structure prediction and has made significant progress in prediction accuracy. Especially AlphaFold2 has pushed protein prediction to a new frontier, but mainstream protein structure prediction methods represented by the AlphaFold2 model heavily rely on co evolutionary information extracted from multiple sequence alignments and templates.

This study broke the speed bottleneck of relying on MSA retrieval models. Compared to AlphaFold2, the HelixFold Single model has an average inference speed improvement of hundreds of times, achieving second level prediction. The efficient HelixFold Single model can not only better adapt to tasks such as protein design and large-scale virtual screening that require frequent prediction of protein structure, but also outperform AlphaFold2 in highly variable protein scenarios more related to macromolecular drug design, such as peptides, antibodies, and nanoantibodies.

HelixFold Single has been established at the National Supercomputing Chengdu Center, empowering scientific research institutions in the protein field in the Sichuan Chongqing region through the supercomputing platform. In the application scenario of macromolecular drugs, HelixFold Single has also been integrated into the Baitu Biotech AIGP platform, providing Baitu with more efficient protein analysis capabilities and assisting it in exploring innovative macromolecular drugs.

According to the research and development team, based on the experience accumulated during the development of HelixFold Single and HelixFold, the team has developed a more universal and robust complex structure prediction algorithm HelixFold Multimer for more challenging antigen-antibody and peptide protein interaction scenarios. Compared with similar methods in the industry, the accuracy has been improved several times. This work will also be launched on the paddlehelix.baidu.com platform in the near future, Provide services to users.

Currently, AI big model technology is driving the rapid development of the field of biological computing. The PaddlePaddle PaddleHelix platform built based on Wenxin's big model technology of biological computing will help researchers in the field of life sciences to more conveniently and efficiently apply the big model technology, better understand the composition and change laws of life bodies, so as to help researchers carry out more exploratory research, such as exploring treatment methods for specific cancers and viral infections, developing new antibiotics and targeted drugs, Or develop more efficient industrial enzymes, etc., to contribute continuous value to human health and industrial development.

Upstream News Yang Xinhua

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])

Mobile advertising space rental

Tag: the Baidu Wenxin large models was recognized by international