Unverified50% confidenceFactExact time
AlphaFold2的核心技术将多序列比对（MSA）与注意力机制融合，通过Evoformer模块迭代精化并以端到端可微分方式预测三维原子坐标
1
Sources
50%
Confidence
Long-term
Relevance
7/5/2026
First Seen
Sources
Related Claims
UnverifiedAlphaFold2通过Evoformer模块将多序列比对信息转化为残基对空间约束，将预测精度提升至RMSD 1-2Å的原子级分辨率78% similarUnverified该攀岩分析工具核心由三个AI模型组合使用：VLM Run的Orion、ViTPose+ Huge、RT-DETR66% similarUnverifiedApple M4 Pro的Neural Engine达38TOPS，高通骁龙X Elite集成45TOPS NPU，英特尔Core Ultra也已集成NPU支持本地AI推理59% similarUnverifiedSmall model engineering capabilities include model quantization (INT8/INT4), ONNX format export and inference acceleration, and TensorRT deployment.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/112347API
curl https://kongchang.com/api/v1/knowledge/claims/112347MCP
get_claim(id=112347)