A transformer-based molecular docking paradigm for large-scale virtual screening

In this study, a novel transformer-based architecture named Dockformer is proposed to overcome the above-mentioned issues of current DL-based docking methods.

Specifically, Dockformer uses two separate transformer encoders to leverage multimodal information to generate latent embeddings of proteins and ligands and can thus effectively capture molecular geometric details, including 2D graph topology and 3D structural knowledge. To strengthen the transformer’s spatial sensitivity, we augment atom representations with 3D point positional encodings, furnishing rich geometric context for subsequent structure prediction.

Next, a binding module is then employed to detect intermolecular relationships effectively on the basis of learned latent embeddings.

Finally, in the structure module, the established relationships are utilized to generate the complex conformations directly, and the coordinates of the ligand atoms are calculated in an end-to-end manner. In addition, the corresponding confidence measures of each generated conformation are utilized to distinguish binding strengths instead of traditional scoring functions.

In summary, distinct from conventional DL-based and optimization-based docking methods, the multimodal information fusion equips Dockformer with superior docking accuracy, and the end-to-end architecture enables it to simultaneously speed up the conformation generation process by orders of magnitude. Thus, this method can meet the rapid throughput requirements of LSVS tasks. Dockformer, as a robust and reliable proteinligand docking approach, may significantly reduce the development cycle and cost of drug design.

Papers

• Zhangfan Yang, Junkai Ji, Shan He, Jianqiang Li, Tiantian He, Ruibin Bai, Zexuan Zhu, Yew Soon Ong. “Dockformer: A transformer-based molecular docking paradigm for large-scale virtual screening.” arXiv preprint arXiv:2411.06740 (2024). View Paper

Context-Aware Fragment Assembly via LLM-MCTS Fusion: A New Paradigm for Structure-Based Drug Generation

The rapid evolution of large language models (LLMs), exemplified by architectures such as ChatGPT and DeepSeek, has profoundly reshaped the natural language processing (NLP) landscape through synergistic integration of two paradigm-shifting technologies: self-supervised learning on massive textual corpora and reinforcement learning (RL) optimization frameworks. Leveraging the generalization capabilities demonstrated by pretrained LLMs, we firstly propose Fragment-based GPT (FRAGPT), a fragment-based molecular language model that learns a generalized chemical distribution from millions of SMILES while autoregressively assembling context-aware fragments into complete molecules.

Our proposed method introduces a collaborative framework named Hodor, integrating FRAGPT with Monte Carlo Tree Search (MCTS) to explicitly guide, filter, and optimize molecular outputs, thus achieving a more interpretable generation process. In this setting, each node encodes an intermediate (partial) molecule, and FRAGPT operates as an ‘action generator’, proposing chemically relevant fragments by leveraging learned fragment-level contextual features and the broader chemical space distribution. MCTS then leverages upper-confidence exploration to balance exploitation of high-value fragments against exploration of novel chemotypes, while its pluggable reward function conveniently incorporates domain-specific criteria, such as affinity, pharmacological properties and SAR.

Our goal is to build an AI-powered retrosynthesis engine that can automatically trace a complex target molecule—or any user-defined product concept—back to a handful of inexpensive, readily purchasable starting materials, proposing step-wise disconnections, plausible reagents, and reaction conditions in natural-language form; the entire pipeline is driven by a large-language-model framework augmented with retrieval and reinforcement learning, enabling rapid, explainable route planning without hand-coded templates.

The Drug Design Agent consists of data layer, tools layer, and applications layer.

The agent automates and unifies the entire drug discovery pipeline, addressing the fragmented nature of existing AI tools in the field.

The agent democratizes access to powerful computational methods, making them more accessible to researchers without specialized programming skills.

Inspired by biological neurons and neural circuits, we model artificial neural network architectures with biological interpretability. The Dendritic Neural Model (DNM) is a novel neural model with plastic dendritic morphology that we have proposed. The DNM consists of synaptic layer, dendritic layer, membrane layer, and soma, with each layer performing corresponding neural functions through different activation functions. Through a neural pruning scheme, DNM can eliminate redundant synapses and dendritic branches to simplify its architecture and form unique neuronal morphologies for each specific task.

Through a logic approximation scheme, DNM can be converted into a Logic Circuit Classifier (LCC), which consists only of comparators and logic AND, OR, and NOT gates.

LCC is easy to implement for large-scale parallel computing on hardware, such as Field-Programmable Gate Arrays (FPGA) and Very Large Scale Integration (VLSI).

For model details, related papers, and source code of DNM, please visit the model website: https://jijunkai123.github.io/DNM/.

Papers

• Yajiao Tang, Zhenyu Song, Yulin Zhu, Maozhang Hou, Cheng Tang, and Junkai Ji. “Adopting a dendritic neural model for predicting stock price index movement.” Expert Systems with Applications (2022): 117637. View Paper

• Cheng Tang, Junkai Ji, Qiuzhen Lin, and Yan Zhou. “Evolutionary Neural Architecture Design of Liquid State Machine for Image Classification.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022. View Paper

• Junkai Ji, Jiajun Zhao, Qiuzhen Lin, and Kay Chen Tan. “Competitive Decomposition-Based Multiobjective Architecture Search for the Dendritic Neural Model.” IEEE Transactions on Cybernetics (2022). View Paper

• Junkai Ji, Cheng Tang, Jiajun Zhao, Zheng Tang, and Yuki Todo. “A Survey on Dendritic Neuron Model: Mechanisms, Algorithms and Practical Applications.” Neurocomputing (2022). View Paper

• Junkai Ji, Minhui Dong, Qiuzhen Lin, and Kay Chen Tan. “Noninvasive Cuffless Blood Pressure Estimation With Dendritic Neural Regression.” IEEE Transactions on Cybernetics (2022). View Paper

• Cheng Tang, Yuki Todo, Junkai Ji, and Zheng Tang. “A novel motion direction detection mechanism based on dendritic computation of direction-selective ganglion cells.” Knowledge-Based Systems (2022): 108205. View Paper

• Zhenyu Song, Cheng Tang, Junkai Ji, Yuki Todo, and Zheng Tang. “A Simple Dendritic Neural Network Model-Based Approach for Daily PM2. 5 Concentration Prediction.” Electronics 10, no. 4 (2021): 373. View Paper

• Minhui Dong, Cheng Tang, Junkai Ji, Qiuzhen Lin, and Ka-Chun Wong. “Transmission trend of the COVID-19 pandemic predicted by dendritic neural regression.” Applied Soft Computing 111 (2021): 107683. View Paper

• Shangce Gao, Mengchu Zhou, Ziqian Wang, Daiki Sugiyama, Jiujun Cheng, Jiahai Wang, and Yuki Todo. “Fully complex-valued dendritic neuron model.” IEEE Transactions on Neural Networks and Learning Systems (2021). View Paper

• Xiliang Zhang, Yuki Todo, Cheng Tang, and Zheng Tang. “The Mechanism of Orientation Detection Based on Dendritic Neuron.” 2021 IEEE 4th International Conference on Big Data and Artificial Intelligence (BDAI). IEEE, 2021. View Paper

• Jiajun Zhao, Qiuzhen Lin, and Junkai Ji. “Network Intrusion Detection by an Approximate Logic Neural Model.” 2021 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW). IEEE, 2021. View Paper

• Cheng Tang, Zhenyu Song, Yajiao Tang, Huimei Tang, Yuxi Wang, and Junkai Ji. “An Evolutionary Neuron Model with Dendritic Computation for Classification and Prediction.” International Conference on Intelligent Computing. Springer, Cham, 2021. View Paper

• Junkai Ji, Minhui Dong, Qiuzhen Lin, and Kay Chen Tan. “Forecasting Wind Speed Time Series Via Dendritic Neural Regression.” IEEE Computational Intelligence Magazine 16.3 (2021): 50-66. View Paper

• Zhenyu Song, Cheng Tang, Jin Qian, Bin Zhang, and Yuki Todo. “Air Quality Estimation Using Dendritic Neural Regression with Scale-Free Network-Based Differential Evolution.” Atmosphere 12.12 (2021): 1647. View Paper

• Shuangbao Song, Xingqian Chen, Shuangyu Song, and Yuki Todo. “A neuron model with dendrite morphology for classification.” Electronics 10.9 (2021): 1062. View Paper

• Junkai Ji, Yajiao Tang, Lijia Ma, Jianqiang Li, Qiuzhen Lin, Zheng Tang, and Yuki Todo. “Accuracy Versus Simplification in an Approximate Logic Neural Model.” IEEE Transactions on Neural Networks and Learning Systems (2020). View Paper

• Cheng Tang, Junkai Ji, Yajiao Tang, Shangce Gao, Zheng Tang, and Yuki Todo. “A novel machine learning technique for computer-aided diagnosis.” Engineering Applications of Artificial Intelligence 92 (2020): 103627. View Paper

• Zhenyu Song, Yajiao Tang, Junkai Ji, and Yuki Todo. “Evaluating a dendritic neuron model for wind speed forecasting.” Knowledge-Based Systems 201 (2020): 106052. View Paper

• Xiaoxiao Qian, Cheng Tang, Yuki Todo, Qiuzhen Lin, and Junkai Ji. “Evolutionary Dendritic Neural Model for Classification Problems.” Complexity 2020 (2020). View Paper

• Jiajun Zhao, Minhui Dong, Cheng Tang, Junkai Ji, and Ying He. “Improving Approximate Logic Neuron Model by Means of a Novel Learning Algorithm.” In International Conference on Intelligent Computing, pp. 484-496. Springer, Cham, 2020. View Paper

• Junkai Ji, Minhui Dong, Cheng Tang, Jiajun Zhao, and Shuangbao Song. “A Novel Plastic Neural Model with Dendritic Computation for Classification Problems.” In International Conference on Intelligent Computing, pp. 471-483. Springer, Cham, 2020. View Paper

• Zhenyu Song, Tianle Zhou, Xuemei Yan, Cheng Tang, and Junkai Ji. “Wind Speed Time Series Prediction Using a Single Dendritic Neuron Model.” In 2020 2nd International Conference on Machine Learning, Big Data and Business Intelligence (MLBDBI), pp. 140-144. IEEE, 2020. View Paper

• Junkai Ji, Shuangbao Song, Yajiao Tang, Shangce Gao, Zheng Tang, and Yuki Todo. “Approximate logic neuron model trained by states of matter search algorithm.” Knowledge-Based Systems 163 (2019): 120-130. View Paper

• Yuki Todo, Zheng Tang, Hiroyoshi Todo, Junkai Ji, and Kazuya Yamashita. “Neurons with multiplicative interactions of nonlinear synapses.” International journal of neural systems 29, no. 08 (2019): 1950012. View Paper

• Tang, Yajiao, Junkai Ji, Yulin Zhu, Shangce Gao, Zheng Tang, and Yuki Todo. “A differential evolution-oriented pruning neural network model for bankruptcy prediction.” Complexity 2019 (2019). View Paper

• Shuangyu Song, Xingqian Chen, Cheng Tang, Shuangbao Song, Zheng Tang, and Yuki Todo. “Training an approximate logic dendritic neuron model using social learning particle swarm optimization algorithm.” IEEE Access 7 (2019): 141947-141959. View Paper

• Jingyao Wu, Jiaxin He, and Yuki Todo. “The dendritic neuron model is a universal approximator.” In 2019 6th International Conference on Systems and Informatics (ICSAI), pp. 589-594. IEEE, 2019. View Paper

• Fei Teng, and Yuki Todo. “Dendritic Neuron Model and Its Capability of Approximation.” In 2019 6th International Conference on Systems and Informatics (ICSAI), pp. 542-546. IEEE, 2019. View Paper

• Jiaxin He, Jingyao Wu, Guangchi Yuan, and Yuki Todo. “Dendritic Branches of DNM Help to Improve Approximation accuracy.” In 2019 6th International Conference on Systems and Informatics (ICSAI), pp. 533-541. IEEE, 2019. View Paper

• Junkai Ji. “The improvement and hybridization of artificial neural networks and swarm intelligence.” Doctoral dissertation, University of Toyama, 2018. View Paper

• Tao Jiang, Shangce Gao, Dizhou Wang, Junkai Ji, Yuki Todo, and Zheng Tang. “A neuron model with synaptic nonlinearities in a dendritic tree for liver disorders.” IEEJ Transactions on Electrical and Electronic Engineering 12, no. 1 (2017): 105-115. View Paper

• Wei Chen, Jian Sun, Shangce Gao, Jiu-Jun Cheng, Jiahai Wang, and Yuki Todo. “Using a single dendritic neuron to forecast tourist arrivals to Japan.” IEICE Transactions on Information and Systems 100, no. 1 (2017): 190-202. View Paper

• Ying Yu, Yirui Wang, Shangce Gao, and Zheng Tang. “Statistical modeling and prediction for tourism economy using dendritic neural network.” Computational intelligence and neuroscience 2017 (2017). View Paper

• Junkai Ji, Shangce Gao, Jiujun Cheng, Zheng Tang, and Yuki Todo. “An approximate logic neuron model with a dendritic structure,” Neurocomputing 173 (2016): 1775-1783. View Paper

• Tianle Zhou, Shangce Gao, Jiahai Wang, Chaoyi Chu, Yuki Todo, and Zheng Tang. “Financial time series prediction using a dendritic neuron model.” Knowledge-Based Systems 105 (2016): 214-224. View Paper

• Junkai Ji, Zhenyu Song, Yajiao Tang, Tao Jiang, and Shangce Gao. “Training a dendritic neural model with genetic algorithm for classification problems.” In 2016 International Conference on Progress in Informatics and Computing (PIC), pp. 47-50. IEEE, 2016. View Paper

• Zijun Sha, Lin Hu, Yuki Todo, Junkai Ji, Shangce Gao, and Zheng Tang. “A breast cancer classifier using a neuron model with dendritic nonlinearity.” IEICE Transactions on Information and Systems 98, no. 7 (2015): 1365-1376. View Paper

• Tao Jiang, Dizhou Wang, Junkai Ji, Yuki Todo, and Shangce Gao. “Single dendritic neuron with nonlinear computation capacity: A case study on xor problem.” In 2015 IEEE International Conference on Progress in Informatics and Computing (PIC), pp. 20-24. IEEE, 2015. View Paper

• Yuki Todo, Hiroki Tamura, Kazuya Yamashita, and Zheng Tang. “Unsupervised learnable neuron model with nonlinear interaction on dendrites.” Neural Networks 60 (2014): 96-103. View Paper

• Zheng Tang, Hiroki Tamura, Makoto Kuratu, Okihiko Ishizuka, and Koichi Tanno. “A model of the neuron based on dendrite mechanisms.” Electronics and Communications in Japan (Part III: Fundamental Electronic Science) 84, no. 8 (2001): 11-24. View Paper