使用奖励引导的推理时搜索，以 RFD4-Proteina 设计蛋白质
Design Proteins with RFD4-Proteina Using Reward-Guided Inference-Time Search
NVIDIA 介绍如何用奖励引导的推理时搜索结合 RFD4-Proteina 进行蛋白质设计。蛋白质设计本质是搜索问题：8 个残基的序列就有超过 250 亿种组合，100 个残基的组合数超过可观测宇宙的原子数，因此需要搜索方法在巨大空间中找到可行序列。
Protein design is a search problem. Choose amino acids for an eight-residue protein sequence and more than 25 billion combinations are possible. Choose amino...
Protein design is a search problem. Choose amino acids for an eight-residue protein sequence and more than 25 billion combinations are possible. Choose amino acids for a 100-residue protein sequence and the number of possibilities exceeds the number of atoms in the observable universe. Somewhere in that space is a protein that opens a path to a new therapy. There are too many combinations to…
来源：NVIDIA Developer Blog · developer.nvidia.com