Biotech startup Chai Discovery has launched the AI model Chai-1, aimed at predicting molecular structures to aid drug development. This model surpasses Google's DeepMind AlphaFold in accuracy and has been supported by OpenAI and Thrive Capital. Chai-1's free web interface and open-source code encourage non-commercial research. The model excels in predicting various types of molecules, particularly demonstrating unexpected accuracy when data is incomplete, contributing significantly to drug development.

AlphaFold 3, a significant breakthrough at the intersection of biological sciences and computer science, was released by Google's DeepMind. However, the lack of code or model weights led to a wave of reproduction efforts within the scientific community. The startup Ligo became the first team to successfully reproduce AlphaFold 3, with its three founders being undergraduate students from Oxford University, achieving this feat in just four months and showcasing scientific potential. This achievement is seen as a milestone in the field of biological sciences, especially with great potential in protein structure prediction. However, DeepMind...

AlphaSeq, launched by A-Alpha Bio, has broken the limitations of protein-protein interaction (PPI) prediction, becoming the world's largest PPI dataset with over 750 million measurement results. It contrasts the success of AlphaFold in protein structure prediction and excels particularly in PPI prediction, simplifying the process through innovative methods. The experimental platform of AlphaSeq efficiently measures the binding affinities of millions of PPIs, accelerating the pace of protein research. With expertise in computational biology...

AlphaFold3（AF3）是DeepMind团队的最新增强工具，用于预测蛋白质、核酸和小分子的结构。凭借能够解析单个蛋白质序列及其复合结构的能力，AF3实现了"食谱到三维结构"的转变。其架构由三项核心模块组成：输入准备梳理序列、检索具备相似结构的分子；表示学习运用多层注意力机制更新表征；结构预测通过条件扩散模型实现预测。AF3创造性地将各类分子转化为独特数值张量，仿佛给每个分子一个独一无二的"身份证"，并运用注意力机制捕捉分子间的互相关系。在预测层面，条件扩散模型从随机噪声开始，逐步构建结构预测。训练中多种损失函数和信心评估共同作用，提高了AF3的预测准确性和可靠性，实现自我反省与提升。AF3的可视化指导和先进架构，展示了人工智能在生物学领域预测模型的革命性进步。

The latest version of AlphaFold can predict the structures of proteins, nucleic acids, and molecules containing PTM, achieving high-accuracy atomic-level predictions for ligands, proteins, and nucleic acids. It significantly surpasses earlier versions in antibody-binding-related issues and is capable of predicting protein-ligand structures more quickly and accurately, aiding in the understanding of complex biological mechanisms and accelerating medical applications.

Harvard and Columbia have released the open dataset OpenProteinSet, containing 16 million protein multiple sequence alignments and related data to support AI model training. AlphaFold 2 has achieved a milestone in the accuracy of protein structure predictions, but its proprietary data has limited progress for other researchers. OpenProteinSet includes MSAs for all proteins in the PDB, providing ample precomputed MSA resources for the protein machine learning community. This dataset can be used for various tasks in structural biology.