protein folding
- protein code
- structure prediction
- folding speed and mechanism
Anfinsen Experiment：蛋白质折叠所需的所有信息都包含在其一维序列中
Protein structure prediction using energy function based methods
AlphaFold：利用了共进化的信息，不仅仅局限于能量函数
- 可预测：
- single chain
- protein multimers
- multisubunit protein-protein interaction
- 难预测：
- Multiple conformations
- effect of point mutations
- antigen-antibody interactions
- 不能预测
- protein-DNA / RNA
- 核酸结构
- Ligand and ion binding
- 翻译后修饰
- membrane plane for transmembrane
ColabFold
AlphaFold3
蛋白质设计 Protein Design
Evolution has sampled a tiny fraction of what is possible to make with proteins
functioned protein只占蛋白质序列空间的一小部分，directed evolution相当于在这些小岛屿周围探索，de novo design相当于在未知的中寻找到全新的蛋白质。
2003, 第一个用计算设计的蛋白质Top 7, Kuhlman, Baker
传统方法：Backbone generation, sequence design, in silico selection
Parametric design 参数化生成 of alpha-helical bundles: crick coiled-coil equations, backbone generation
- Enzyme design
- Small molecular binders: near infrared fluorescent proteins
- protein binder
- transmembrane channels，设计较困难，数据少
- Vaccine scaffolds
- accurate de novo design of membrane-tranversing
- Peptide design
- Antibody design
protein structure prediction
New design methodologies: backbone generation，利用diffusion model生成
RoseTTAFold
sequence design:
- structure based, ProteinMPNN, LigandMPNN, CarbonDesign
- language based, ProGen, ESM3