M²N: A Progressive Macro-to-Micro 3D Modeling Scheme for Unveiling Drug-Target Affinity
Authors: Tianxu Lv, Jie Zhu, Jinyi Liu, Shiyun Nie, Hongnian Tian, Yang Xiao, Yuan Liu, Llihua Li, Xiang Pan
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on two datasets indicate that M²N not only outperforms state-of-theart methods under various conditions, but also provides a new paradigm for target and drug unified modeling. ... Experiments Datasets We conduct a comprehensive set of experiments on two benchmark datasets, namely DAVIS(Davis et al. 2011) and KIBA(Tang et al. 2014). ... Tables 2-4 report the performances of the proposed method and recent state-of-the-art approaches on the DAVIS and KIBA datasets under three challenging scenarios. ... Ablation Study Effectiveness of designed components To investigate the impact of each designed component in the proposed approach, we conduct an extensive ablation analysis by evaluating different M2N variants as follows: ... Figure 2: Analysis of hyper-parameter sensitivity concerning the number of neighbors (a) and the balance coefficients (b). |
| Researcher Affiliation | Academia | 1 School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi 214122, China 2 Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen 518071, China 3 Institute of Biomedical Engineering and Instrumentation, Hangzhou Dianzi University, Hangzhou 310018, China 4 The PRC Ministry of Education Engineering Research Center of Intelligent Technology for Healthcare, Wuxi, Jiangsu 214122, China EMAIL, EMAIL, EMAIL |
| Pseudocode | No | The paper describes the methodology using mathematical equations and block diagrams (e.g., Figure 1b for the graph transformer block), but does not present a dedicated section or figure labeled 'Pseudocode' or 'Algorithm', nor a structured code-like block. |
| Open Source Code | No | The paper does not contain any explicit statement about providing source code for the described methodology, nor does it include a link to a code repository. |
| Open Datasets | Yes | We conduct a comprehensive set of experiments on two benchmark datasets, namely DAVIS(Davis et al. 2011) and KIBA(Tang et al. 2014). |
| Dataset Splits | Yes | All experiments use a five-fold cross-validation approach, and the mean scores of the results are presented as the final result. ... S1: The target protein within the drug-target pair has not been encountered during the training phase. S2: The drug molecule within the drug-target pair has not been encountered during the training phase. S3: The target protein and the drug molecule are both novel, representing the most rigorous evaluation. |
| Hardware Specification | Yes | The experiments in this study are executed on a platform comprising two NVIDIA Ge Force RTX 4090 GPUs to accelerate the training process. |
| Software Dependencies | No | The developed model is implemented utilizing Pytorch (Paszke et al. 2019) along with Pytorch Geometric (Fey and Lenssen 2019). While PyTorch and PyTorch Geometric are mentioned, specific version numbers are not provided. |
| Experiment Setup | Yes | All hyperparameters of M2N are listed in Table 1. ... Table 1: Hyperparameter settings of our M2N model. Hyperparameter Value(s) Epoch 200 Batch size 32 Optimizer Adam W Learning rate 0.0005 Weight decay 0.0001 Balancing coefficient λ1 0.1 Balancing coefficient λ2 0.2 Output dimension of graph embedding 256 Number of GT layers in each GTformer 3 Number of neighbours per node k1 in GDFG 5 Number of neighbours per node k2 in GT Res 25 |