FDTBX——从原子坐标模拟X射线纤维衍射图案的计算工具
FDTBX - Computational Tools for Simulation of X-ray Fiber Diffraction Patterns from Atomic Coordinates
浏览论文内容
中文总结 AI 辅助
研究团队推出开源Python工具箱fdtbx，可从原子坐标模拟X射线纤维衍射图案，纳入多种物理展宽机制，用于纤维素等体系，助力可复现的模型分析，兼具研究与教学价值。
中文摘要 AI 辅助
纤维衍射是少数能够解析部分有序、非晶体系（如纤维蛋白、生物聚合物和合成聚合物）分子结构的实验技术之一。由于纤维衍射图案较为复杂，表现出次晶有序、堆积缺陷以及宽泛重叠的反射峰，其解释依赖于基于模型的精修，即从候选原子模型模拟得到的理论图案与实验图案进行迭代比较。我们推出fdtbx，这是一款现代、开源且可扩展的Python工具箱，可直接从原子坐标模拟X射线纤维衍射图案。该工具箱基于cctbx库构建，结构兼顾可读性与多核执行能力，能从PDB模型计算结构因子，并构建真实的倒易空间反射轮廓，其中纳入了纤维衍射的主要物理展宽机制：有限微晶尺寸、取向无序（含高斯、洛伦兹和沃伊特角峰形），以及Hosemann公式中的次晶（第二类）晶格无序。每反射壳层求积方案可在任意查询点而非固定网格上计算所需卷积，从而直接控制精度与速度间的权衡；解析埃瓦尔德投影例程则将每个采样的反射（附带适当的洛伦兹校正）映射到平面探测器，生成模拟图案以直接与实测图像对比。我们以纤维素I-α/I-β、α-几丁质和三乙酸纤维素为例说明该工具箱，并概述了实用的模拟工作流程。通过提供透明、文档完善且可并行化的纤维衍射专用算法实现，fdtbx降低了可复现的基于模型分析的门槛，兼具研究与教学资源的双重用途。
英文摘要
Fiber diffraction is one of the few experimental techniques capable of resolving molecular structure in partially ordered, non-crystalline systems such as fibrous proteins, biopolymers and synthetic polymers. Because fiber diffraction patterns are complex, exhibiting paracrystalline order, packing defects and broad, overlapping reflections, their interpretation relies on model-based refinement, in which theoretical patterns simulated from candidate atomic models are iteratively compared with experiment. We present fdtbx, a modern, open and extensible Python toolbox for simulating X-ray fiber diffraction patterns directly from atomic coordinates. Built on the crystallographic library cctbx and designed for readability and multicore execution, fdtbx computes structure factors from a PDB model and constructs realistic reciprocal-space reflection profiles that incorporate the principal physical broadening mechanisms of fiber diffraction: finite crystallite size, orientational disorder (with Gaussian, Lorentzian and Voigt angular peak shapes) and paracrystalline (second-kind) lattice disorder in Hosemann's formulation. A per-reflection shell-quadrature scheme evaluates the required convolutions at arbitrary query points rather than on a fixed grid, giving direct control over the trade-off between accuracy and speed, and an analytic Ewald-projection routine maps each sampled reflection, with an appropriate Lorentz correction, onto a flat detector to produce a simulated pattern for direct comparison with measured images. We illustrate the toolbox on $α$-chitin, cellulose~I$α$ and I$β$, cellulose triacetate and A-form DNA, and outline a practical simulation workflow. By providing a transparent, well-documented and parallelizable implementation of the specialized algorithms of fiber diffraction, fdtbx lowers the barrier to reproducible model-based analysis.
