Equivariant Scalar Fields for Molecular Docking with Fast Fourier Transforms

Authors: Bowen Jing, Tommi S. Jaakkola, Bonnie Berger

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We benchmark our scoring functions on two simplified docking-related tasks: decoy pose scoring and rigid conformer docking. Our method attains similar but faster performance on crystal structures compared to the widely-used Vina and Gnina scoring functions, and is more robust on computationally predicted structures. Code is available at https://github.com/bjing2016/scalar-fields.
Researcher Affiliation Academia Bowen Jing1, Tommi Jaakkola1, Bonnie Berger1 2 1CSAIL, Massachusetts Institute of Technology 2Dept. of Mathematics, Massachusetts Institute of Technology
Pseudocode Yes Algorithm 1: TRANSLATIONAL FFT (...) Algorithm 2: ROTATIONAL FFT (...) Algorithm 3: TRANSLATIONAL SCORING (...) Algorithm 4: ROTATIONAL SCORING
Open Source Code Yes Code is available at https://github.com/bjing2016/scalar-fields.
Open Datasets Yes We train and test our model on the PDBBind dataset (Liu et al., 2017) with splits as defined by St ark et al. (2022).
Dataset Splits Yes As noted previously, we use train, validation, and test splits from St ark et al. (2022).
Hardware Specification Yes All runtime measurements were performed on a machine with 64 Intel Xeon Gold 6130 CPUs and 8 Nvidia Tesla V100 GPUs.
Software Dependencies No The paper mentions software like 'RDKit', 'Gnina', 'Vina', 'TANKBind', and 'Diff Dock', but it does not specify version numbers for these or any other ancillary software components, nor does it list specific library versions (e.g., Python, PyTorch versions).
Experiment Setup Yes All other model and inference-time hyperparameters are discussed in Appendix E. (...) These hyperparameters and their search spaces are detailed in Table 4 (...) The search space for these resolution hyperpameters is detailed in Table 5.