Reprogramming Pretrained Target-Specific Diffusion Models for Dual-Target Drug Design
Authors: Xiangxin Zhou, Jiaqi Guan, Yijia Zhang, Xingang Peng, Liang Wang, Jianzhu Ma
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate the effectiveness of our method compared with various baselines. |
| Researcher Affiliation | Academia | Xiangxin Zhou1,2 Jiaqi Guan3 Yijia Zhang4 Xingang Peng5 Liang Wang1,2 Jianzhu Ma4,6, 1School of Artificial Intelligence, University of Chinese Academy of Sciences 2New Laboratory of Pattern Recognition (NLPR), State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS), Institute of Automation, Chinese Academy of Sciences (CASIA) 3Department of Computer Science, University of Illinois Urbana-Champaign 4Department of Electronic Engineering, Tsinghua University 5Institute for Artificial Intelligence, Peking University 6Institute for AI Industry Research, Tsinghua University |
| Pseudocode | No | The paper describes the methods in prose and mathematical equations but does not include any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | We require authorization to release the code. Once we obtain the necessary approval, we will proceed with the code release. |
| Open Datasets | Yes | To collect drug combination pairs, we start from Drug Comb DB2 [35]. Drug Bank3 [30] and Therapeutic Target Database (TTD)4 [67]. ...if they are available in PDBBind [38], a repository of protein-ligand binding structures sourced from the Protein Data Bank (PDB) [2]. ...Alpha Fold Protein Structure Database (Alpha Fold DB) [59] |
| Dataset Splits | No | The paper states 'All 12,917 pairs of targets... are used for evaluation,' implying a single evaluation set rather than distinct train/validation/test splits for their proposed method. It mentions pretrained models but does not detail their specific split for the new task. |
| Hardware Specification | No | The experiments are done on several clusters. The overall accurate computer resources are not well recorded. |
| Software Dependencies | No | The paper mentions 'Auto Dock Vina [12]' but does not provide specific version numbers for it or any other software libraries or dependencies used in the experiments. |
| Experiment Setup | Yes | More specifically, we first dock a ligand molecule to pocket P1 and pocket P2 separately. ... where we additionally introduce a hyperparameter η here to control the strength of the drift. In theory, this is equivalent to making a more flexible assumption that p(x|P1, T P2) [pθ(x|P1)pθ(x|T P2)]η. In practice, we set η = 1/2 by default. ... we use each method to design 10 molecules for each pair of targets, denoted as P1 and P2. ... We employ Auto Dock Vina [12] to estimate the target binding affinity... We also evaluate the RMSD between docked poses towards dual targets. ... We perform different methods to align the pockets of dual targets for COMPDIFF and DUALDIFF. ... We conduct ablation on different strategies of identifying fragments for Diff Linker and Linker Net for dual-target drug design. Since we use Vina Dock to select fragments, we have tried different box sizes for docking, i.e., 5Å and 8Å. |