Floating Anchor Diffusion Model for Multi-motif Scaffolding

Authors: Ke Liu, Weian Mao, Shuaike Shen, Xiaoran Jiao, Zheng Sun, Hao Chen, Chunhua Shen

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5. Experiments, Dataset. Two datasets are utilized in this work..., Evaluation metrics., Ablation study is conducted to evaluate the effectiveness of TSP and noise scale., Finally, we compare our approaches with the conditional generation and inpainting methods...
Researcher Affiliation Collaboration 1Zhejiang University, China 2The University of Adelaide, Australia 3Swansea University, UK 4Ant Group.
Pseudocode No Not found.
Open Source Code Yes Code is available at https://github.com/aim-uofa/FADiff.
Open Datasets Yes Two datasets are utilized in this work, including the virtual motif dataset (VM dataset) from the PDB database (Berman et al., 2000) for training and the evaluation multi-motif scaffolding benchmark MS Benchmark that we collected from the PROSITE database(Sigrist et al., 2012).
Dataset Splits No Not found. The paper mentions “training” and “evaluation” sets but does not specify explicit numerical splits (percentages or counts) for train, validation, and test sets.
Hardware Specification Yes All our experiments are conducted on a computing cluster with 8 GPUs of NVIDIA Ge Force RTX 4090 24GB and CPUs of AMD EPYC 7763 64-Core of 3.52GHz. All the inferences are conducted on a single GPU of NVIDIA Ge Force RTX 4090 24GB.
Software Dependencies No Not found. The paper refers to other models/frameworks (Frame Diff, VFN-Diff) but does not list specific software dependencies (e.g., libraries, programming languages) with their version numbers.
Experiment Setup Yes We train FADiff for 90,000 steps with a coordinate scale of 0.1 and 0.02 based on the pre-trained VFN-Diff (Mao et al., 2024).Euler-Maruyama discretization with 500 steps implemented as a geodesic random walk is adopted in this work...