Regularized Molecular Conformation Fields
Authors: Lihao Wang, Yi Zhou, Yiqun Wang, Xiaoqing Zheng, Xuanjing Huang, Hao Zhou
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our model constantly outperforms state-of-the-art models for the conformation generation task on the GEOM-Drugs dataset. We demonstrate the effectiveness of RMCF using the GEOM-QM9 and GEOM-Drugs dataset[Axelrod and Gomez-Bombarelli, 2022], where results show that RMCF significantly outperforms state-ofthe-art ML models. |
| Researcher Affiliation | Collaboration | Lihao Wang1 , Yi Zhou2, Yiqun Wang2, Xiaoqing Zheng1, Xuanjing Huang1,3, Hao Zhou4 1Fudan University, 2Byte Dance AI Lab 3Shanghai Collaborative Innovation Center of Intelligent Visual Computing 4Institute for AI Industry Research (AIR), Tsinghua University {wanglh19, zhengxq, xjhuang}@fudan.edu.cn {zhouyi.naive, yiqun.wang}@bytedance.com, zhouhao@air.tsinghua.edu.cn |
| Pseudocode | Yes | The detailed algorithm can be found in Appendix C. |
| Open Source Code | Yes | 2The code is available at https://github.com/leowang1217/RMCF. Did you include the code, data, and instructions needed to reproduce the main experi- mental results (either in the supplemental material or as a URL)? [Yes] |
| Open Datasets | Yes | We benchmark our model performance using the GEOM-QM9 and GEOM-Drugs dataset[Axelrod and Gomez-Bombarelli, 2022], which contains small and mid-sized organic molecules with high quality conformations. Open-source data |
| Dataset Splits | Yes | The final training/validation/test set contains 271,539/30,171/1,034 molecules, respectively. |
| Hardware Specification | Yes | Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [Yes] See Appendix |
| Software Dependencies | No | The paper mentions "We adopt the Message-Passing Neural Network (MPNN) [Gilmer et al., 2017] as the framework for implementing our graph neural network.", but it does not specify version numbers for MPNN or any other software libraries or tools. |
| Experiment Setup | Yes | As for the angular discretization of dihedral angles, the 360 degree interval is evenly divided into 72 bins. Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] See Appendix and Section 4 |