Predicting Molecular Conformation via Dynamic Graph Score Matching

Authors: Shitong Luo, Chence Shi, Minkai Xu, Jian Tang

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments across multiple tasks show that the DGSM outperforms state-of-the-art baselines by a large margin, and it is capable of generating conformations for a broader range of systems such as proteins and multi-molecular complexes.
Researcher Affiliation Academia Shitong Luo*1, Chence Shi*2,3, Minkai Xu2,3, Jian Tang2,4,5 1Peking University 2Mila Québec AI Institute 3Université de Montréal 4HEC Montréal 5CIFAR AI Research Chair luost@pku.edu.cn , chence.shi@umontreal.ca minkai.xu@umontreal.ca , jian.tang@hec.ca
Pseudocode Yes Algorithm 1 Annealed Langevin dynamics
Open Source Code No The paper does not include an unambiguous statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes Following previous works [34, 43], we use the GEOM-QM9 and GEOM-Drugs [1] datasets for this task.
Dataset Splits No The paper mentions 'train-test split' but does not explicitly state the use of a distinct 'validation' dataset split for hyperparameter tuning or early stopping.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper does not provide specific version numbers for software dependencies or libraries used in the experiments.
Experiment Setup Yes The threshold δ of COV score is 0.5Å for GEOM-QM9 and 1.25Å for GEOM-Drugs following Xu et al. [43]. We here take the prior distribution as a standard Gaussian N(R0 | 0, I). Then, we update the conformation by running T steps of Langevin dynamic to get a sample from each noise conditional score network sθ(R, σi) sequentially with a special step size schedule αi = ε σ2 i /σ2 L.