A 3D Generative Model for Structure-Based Drug Design

Authors: Shitong Luo, Jiaqi Guan, Jianzhu Ma, Jian Peng

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments to evaluate our approach. Quantitative and qualitative results show that: (1) our method is able to generate diverse drug-like molecules that have high binding affinity to specific targets based on 3D structures of protein binding sites; (2) our method is able to generate molecules with fairly high drug-likeness score (QED) [4] and synthetic accessibility score (SA) [6] even if the model is not specifically optimized for them; (3) in addition to molecule generation, the proposed method is also applicable to other relevant tasks such as linker design.
Researcher Affiliation Collaboration Shitong Luo Heli Xon Research luost@helixon.com luost26@gmail.com Jiaqi Guan University of Illinois Urbana-Champaign jiaqi@illinois.edu Jianzhu Ma Peking University majianzhu@pku.edu.cn Jian Peng University of Illinois Urbana-Champaign jianpeng@illinois.edu
Pseudocode No The paper describes algorithms in prose (e.g., Section 3.2 'Sampling') and includes illustrations (Figure 1), but it does not contain a structured pseudocode or algorithm block.
Open Source Code Yes Other details about model architectures and training parameters are provided in the supplementary material and the open source repository: https://github.com/ luost26/3D-Generative-SBDD.
Open Datasets Yes We use the Cross Docked dataset [7] following [20]. The dataset originally contains 22.5 million docked protein-ligand pairs at different levels of quality.
Dataset Splits No We use mmseqs2 [31] to cluster data at 30% sequence identity, and randomly draw 100,000 protein-ligand pairs for training and 100 proteins from remaining clusters for testing. No explicit validation set is mentioned.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. It only implies that models were trained and experiments conducted.
Software Dependencies No The paper mentions software components like 'Adam optimizer' and 'Open Babel [21, 20]' but does not provide specific version numbers for these or any other key software dependencies required for replication.
Experiment Setup Yes The number of message passing layers in context encoder L is 6, and the hidden dimension is 256. We train the model using the Adam optimizer at learning rate 0.0001.