Graph Diffusion Transformers for Multi-Conditional Molecular Generation

Authors: Gang Liu, Jiaxin Xu, Tengfei Luo, Meng Jiang

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

Reproducibility Variable Result LLM Response
Research Type Experimental We extensively validate Graph Di T for multi-conditional polymer and small molecule generation. Results demonstrate the superiority of Graph Di T across nine metrics from distribution learning to condition control for molecular properties. In experiments, we evaluate model performance on one polymer and three small molecule datasets.
Researcher Affiliation Academia Gang Liu, Jiaxin Xu, Tengfei Luo, Meng Jiang University of Notre Dame {gliu7, jxu24, tluo, mjiang2}@nd.edu
Pseudocode No Not found. The paper contains architectural diagrams and descriptions but no explicit pseudocode or algorithm blocks.
Open Source Code Yes Code is provided in the supplementary materials. Data and code will be on Github after publication.
Open Datasets Yes We have one polymer dataset [40] for materials, featuring three numerical gas permeability conditions: O2Perm, CO2Perm, and N2Perm. For drug design, we create three class-balanced datasets from Molecule Net [46]: HIV, BBBP, and BACE
Dataset Splits Yes We randomly split the dataset into training, validation, and testing (reference) sets in a 6:2:2 ratio.
Hardware Specification Yes All experiments can be run on a single A6000 GPU card.
Software Dependencies No Not found. The paper does not specify software dependencies with version numbers for reproducibility (e.g., Python, PyTorch, or other libraries with their specific versions).
Experiment Setup No Not found. The paper describes architectural and encoding choices but does not provide specific numerical hyperparameters (e.g., learning rate, batch size, number of epochs) or system-level training settings.