Replication Data for: HOPS: high-performance library for (non-)uniform sampling of convex-constrained models

This collection contains showcased models and pre-processing results (determining an independent flux space and rounding) that are used as basis for benchmarking the uniform sampling performance of the CHHR implementations of the HOPS library [10.1093/bioinformatics/btaa872] (https://github.com/modsim/hops) and the COBRA toolbox [doi.org/10.1038/s41596-018-0098-2]. Here, the pre-processing was performed with the COBRA toolbox for all benchmarks to guarantee fair comparability. The models are split in two classes: 1) simplices with 64, 256, 512, 1024, and 2048 dimensions; 2) metabolic network models (e_coli_core, iAT_PLT_256, iJO1366, RECON1, Recon2, Recon3D_301) [http://bigg.ucsd.edu/], [doi.org/10.1038/nbt.2488] formulated in SBML format [doi.org/10.1093/bioinformatics/btg015]. For each of these models, the left hand side of the constraint system (called A), the right hand side (called b), a shift of the transformation from sampling space, meaning the null space, to parameter space, meaning the full space of the model (called p_shift), the linear transformation from sampling space to parameter space (called N) and the Chebyshev center (called start) are provided. These files come in rounded and unrounded types, indicating if the rounding algorithm has been applied. For convenience, sometimes the rounding transformation is also given (indicated by T). The rounding transformation is only given in the rounded form, because it is identity otherwise. Additionally, the dataset contains an E. coli WT model with carbon atom transitions and isotope labelling measurements [10.1038/nprot.2009.58] used in the Bayesian inference non-uniform sampling example. The E. coli WT model is contained in a single FluxML file [doi.org/10.3389/fmicb.2019.01022]. To reproduce the data generated with the E. coli WT model, the rounding was calculated using the HOPS library and independent fluxes were provided by the high-performance simulator 13CFLUX2 [doi.org/10.1093/bioinformatics/bts646].

本数据集收录了用于基准测试HOPS库[10.1093/bioinformatics/btaa872]（https://github.com/modsim/hops）与COBRA工具箱[doi.org/10.1038/s41596-018-0098-2]的CHHR实现的均匀采样性能的展示模型及预处理结果（含独立通量空间确定与舍入操作）。为保证公平可比，所有基准测试的预处理均通过COBRA工具箱完成。
数据集包含的模型分为两类：1）64、256、512、1024及2048维单纯形；2）以系统生物学标记语言（SBML，Systems Biology Markup Language）格式[doi.org/10.1093/bioinformatics/btg015]构建的代谢网络模型（e_coli_core、iAT_PLT_256、iJO1366、RECON1、Recon2、Recon3D_301）[http://bigg.ucsd.edu/]、[doi.org/10.1038/nbt.2488]。针对上述每类模型，均提供了约束系统的左侧项（记为A）、右侧项（记为b）、从采样空间（即零空间）到模型全空间对应参数空间的变换偏移量（记为p_shift）、从采样空间到参数空间的线性变换（记为N）以及切比雪夫中心（记为start）。
上述文件分为舍入与未舍入两种类型，以标识是否应用了舍入算法。为便于使用，部分场景下还会提供舍入变换（以T标识）。由于舍入变换在未舍入时为恒等变换，因此仅在舍入类型的文件中提供该变换。
此外，本数据集还包含用于贝叶斯推断非均匀采样示例的、带有碳原子跃迁与同位素标记测量数据的大肠杆菌野生型（E. coli WT）模型[10.1038/nprot.2009.58]。该大肠杆菌野生型模型存储于单个通量标记语言（FluxML）文件[doi.org/10.3389/fmicb.2019.01022]中。若需复现该大肠杆菌野生型模型生成的数据集，可通过HOPS库计算舍入操作，且独立通量由高性能模拟器13CFLUX2 [doi.org/10.1093/bioinformatics/bts646]提供。