WSWanxiang Shen

Wanxiang Shen

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Wanxiang Shen has a total of 85 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Activity-cliff awareness enables robust graph learning for molecular property prediction

    by Chao Cui, Xiaorui Su, Zaixi Zhang, Alejandro Velez-Arce, Jianmin Wang, Xiangcheng Shi, Yanbin Zhang, Jie Wu, Yu Zong Chen, Marinka Zitnik, et al.

    Accurately predicting molecular activity is hindered by activity cliffs (ACs), which are sharp potency changes between highly similar compounds that distort the smoothness assumed by modern quantitative structure–activity relationship (SAR) models and graph neural networks (GNNs). Here we introduce AC awareness (ACA), an inductive bias that reshapes GNN latent spaces to account for these discontinuities. Implemented through an ACA loss combining regression with soft-margin triplet contrastive learning, the method dynamically mines high-value activity-cliff triplets during training and corrects inconsistent neighborhoods in latent space. This process yields progressively fewer cliff violations, more coherent activity gradients, and reduced label incoherence across diverse chemical spaces. Evaluated on 52 datasets spanning low-sample narrow-scaffold series, large mixed-scaffold benchmarks, matched-pair cliff classification, and absorption, distribution, metabolism, excretion and toxicity (ADMET) property prediction, ACA consistently improves predictive accuracy over strong extended-connectivity fingerprint (ECFP) and GNN baselines. The approach generalizes across multiple GNN backbones and remains robust under fixed hyperparameters. These results suggest that ACA provides a principled strategy for enhancing molecular property prediction by aligning latent representations with the nonadditive behavior underlying ACs.


  • Open Access

    Generalizable AI predicts immunotherapy outcomes across cancers and treatments

    by Wanxiang Shen, Intae Moon, Thinh H. Nguyen, Michelle M. Li, Yepeng Huang, Nitya Nair, Daniel Marbach, Marinka Zitnik

    Immune checkpoint inhibitors (ICIs) are a standard treatment across cancers, yet most patients do not respond, and existing biomarkers generalize poorly across tumor types and therapies. Here we present COMPASS, a pan-cancer foundation model that predicts immunotherapy response from bulk tumor transcriptomes using a concept bottleneck transformer. COMPASS encodes gene expression through 44 biologically grounded immune concepts representing immune cell states, tumor−microenvironment interaction and signaling pathways. Trained on 10,184 tumors across 33 cancer types, COMPASS achieves better average performance than 22 methods across 16 clinical cohorts spanning seven cancers and six ICIs, improving accuracy by 8.5% and area under the precision-recall curve by 15.7% on average across cohorts. COMPASS generalizes to cancer types and treatments not represented during fine-tuning and may inform indication selection and patient stratification. In survival analyses, patients classified by COMPASS as responders had longer overall survival (hazard ratio = 4.7, P < 0.0001). Personalized response maps connect gene expression to immune concepts, identifying programs associated with response and resistance; in immune-inflamed non-responders, COMPASS highlights programs including TGFβ signaling, endothelial exclusion, CD4+ T cell dysfunction and B cell deficiency. COMPASS predicts immunotherapy response and provides hypothesis-generating mechanistic insight for trial design and translational studies.


  • Open Access

    MusicMol: Bidirectional Molecule–Music Translation via Neighborhood-Aligned Generation and Transformer Decoding

    by Jiayi Tang, Ziyan Zhu, Chao Cui, Nan Ke, Chenxiao Song, Yong Li, Yuan Zhou, Chunting Yu, Pengfei Jing, Chaoyang Zhu, et al.

    What if a molecule could be heard as a piano melody, and a melody could be used to generate a molecule? We present MusicMol, a bidirectional framework for generative translation between small molecules and symbolic music. In the molecule-to-music direction, a label-free, rule-guided conditional flow-matching generator trained on two million ChEMBL compounds maps each heavy atom to four note parameters—pitch, inter-onset step, duration, and velocity. To make the generated music chemically meaningful, MusicMol aligns molecular similarity neighborhoods with differentiable piano-roll neighborhoods using a kernelized alignment loss that combines softmax KL-divergence with rank preservation, so that structurally similar molecules are mapped to acoustically similar scores. In the music-to-molecule direction, a REMI/BPE encoder–decoder Transformer autoregressively decodes symbolic music into valid SELFIES molecular strings, enabling user-composed or performed music to generate molecular candidates. To support this bidirectional loop, we construct, to our knowledge, the largest molecule–MIDI paired dataset to date, comprising ten million compound–score pairs. On held-out music–molecule pairs, the inverse model achieves high token-level generation accuracy and near-perfect molecular validity. Across 30 molecular activity prediction tasks, the music-derived piano-roll representation outperforms four established molecular encoders spanning SMILES, graph, image, and 3D modalities in frozen linear-probe Spearman correlation, R2, and RMSE, and achieves the lowest local label incoherence index (LII) from k=5 to k=100. Deployed as an interactive platform at the Future Learning Center of Zhejiang University, MusicMol lets users hear molecules in real time and translate music into molecular candidates, providing a perceptual and generative interface between chemistry and music.

  • Open Access

    Activity-cliff awareness enables robust graph learning for molecular property prediction

    by Chao Cui, Xiaorui Su, Zaixi Zhang, Alejandro Velez-Arce, Jianming Wang, Xiang Cheng Shi, Yan Bing Zhang, Jie Wu, Yu Zong Chen, Marinka Zitnik, et al.

    Accurately predicting molecular activity is hindered by activity cliffs (ACs), which are sharp potency changes between highly similar compounds that distort the smoothness assumed by modern quantitative structure–activity relationship (SAR) models and graph neural networks (GNNs). Here we introduce AC awareness (ACA), an inductive bias that reshapes GNN latent spaces to account for these discontinuities. Implemented through an ACA loss combining regression with softmargin triplet contrastive learning, the method dynamically mines high-value activity-cliff triplets during training and corrects inconsistent neighbourhoods in latent space. This process yields progressively fewer cliff violations, more coherent activity gradients, and reduced label incoherence across diverse chemical spaces. Evaluated on 52 datasets spanning low-sample narrow-scaffold series, large mixed-scaffold benchmarks, matched-pair cliff classification, and absorption, distribution, metabolism, excretion and toxicity (ADMET) property prediction, ACA consistently improves predictive accuracy over strong extended-connectivity fingerprint (ECFP) and GNN baselines. The approach generalizes across multiple GNN backbones and remains robust under fixed hyperparameters. These results suggest that ACA provides a principled strategy for enhancing molecular property prediction by aligning latent representations with the nonadditive behaviour underlying ACs.

  • Open Access

    Activity-cliff awareness enables robust graph learning formolecular property prediction

    by Chao Cui, Xiaorui Su, Zaixi Zhang, Alejandro Velez-Arce, Jianming Wang, Xiang Cheng Shi, Yan Bing Zhang, Jie Wu, Yu Zong Chen, Marinka Zitnik, et al.

    Accurately predicting molecular activity is hindered by activity cliffs (ACs), which are sharp potency changes between highly similar compounds that distort the smoothness assumed by modern quantitative structure–activity relationship (SAR) models and graph neural networks (GNNs). Here we introduce AC awareness (ACA), an inductive bias that reshapes GNN latent spaces to account for these discontinuities. Implemented through an ACA loss combining regression with softmargin triplet contrastive learning, the method dynamically mines high-value activity-cliff triplets during training and corrects inconsistent neighbourhoods in latent space. This process yields progressively fewer cliff violations, more coherent activity gradients, and reduced label incoherence across diverse chemical spaces. Evaluated on 52 datasets spanning low-sample narrow-scaffold series, large mixed-scaffold benchmarks, matched-pair cliff classification, and absorption, distribution, metabolism, excretion and toxicity (ADMET) property prediction, ACA consistently improves predictive accuracy over strong extended-connectivity fingerprint (ECFP) and GNN baselines. The approach generalizes across multiple GNN backbones and remains robust under fixed hyperparameters. These results suggest that ACA provides a principled strategy for enhancing molecular property prediction by aligning latent representations with the nonadditive behaviour underlying ACs.

  • Open Access

    Activity Cliff-Informed Contrastive Learning for Molecular Property Prediction

    by Chao Cui, Xiaorui Su, Zaixi Zhang, Alejandro Velez Arce, Jianming Wang, Xiangcheng Shi, Yanbing Zhang, Jie Wu, Yu Zong Chen, Marinka Zitnik, et al.

    Accurately predicting molecular activity is hindered by activity cliffs, which are sharp potency changes between highly similar compounds that distort the smoothness assumed by modern QSAR and graph neural networks (GNNs). Here we introduce activity cliff awareness (AC-awareness), an inductive bias that reshapes GNN latent spaces to account for these discontinuities. Implemented through an Activity Cliff Awareness (ACA) loss combining regression with soft-margin triplet contrastive learning, the method dynamically mines high-value cliff triplets during training and corrects inconsistent neighbourhoods in latent space. This yields progressively fewer cliff violations, more coherent activity gradients, and substantially reduced label incoherence across diverse chemical spaces. Evaluated on 52 datasets spanning low-sample narrow-scaffold series, large mixed-scaffold benchmarks, matched-pair cliff classification, and ADMET delta property prediction, AC-awareness consistently improves predictive accuracy and outperforms strong ECFP- and GNN-based baselines. The approach generalizes across multiple GNN backbones and remains effective under fixed hyperparameters, demonstrating that cliff-aware contrastive geometry provides a robust, architecture-independent mechanism for mitigating structure–activity discontinuities. These results establish AC-awareness as a principled strategy for enhancing molecular property prediction by aligning latent representations with the nonadditive behaviour underlying activity cliffs.

  • Top Cited Papers
    Open Access

    Medea: An omics AI agent for therapeutic discovery

    by Pengwei Sui, Michelle M. Li, Shanghua Gao, Wanxiang Shen, Valentina Giunchiglia, Andrew Shen, Yepeng Huang, Zhenglun Kong, Marinka Zitnik

    AI agents promise to empower biomedical discovery, but realizing this promise requires the ability to complete transparent, long-horizon analyses using tools. Agents must make intermediate decisions explicit, and validate each decision and output against data and tool constraints as the analysis unfolds. We present M edea , an AI agent that takes an omics objective and executes a transparent multi-step analysis using tools. M edea comprises four modules: research planning with context and integrity verification, code execution with pre- and post-run checks, literature reasoning with evidence-strength assessment, and a consensus stage that reconciles evidence across datasets, tools, and literature. M edea uses 20 tools spanning single-cell and bulk transcriptomic datasets, cancer vulnerability maps, pathway knowledge bases, and machine learning models. We evaluate M edea across 5,679 analyses in three open-ended domains: target identification across five diseases and cell type contexts (2,400 analyses), synthetic lethality reasoning in seven cell lines (2,385 analyses), and immunotherapy response prediction in bladder cancer (894 patient analyses). In evaluations that vary large language models, tool sets, omics objectives, and agentic modules, M edea improves the performance of existing approaches by up to 46% for target identification, 22% for synthetic lethality, and 24% for immunotherapy response prediction, while maintaining low failure rates and calibrated abstention. M edea shows that verification-aware AI agents improve performance by producing transparent analyses, not simply more efficient workflows.


  • Open Access

    Generalizable AI predicts immunotherapy outcomes across cancers and treatments

    by Wanxiang Shen, Intae Moon, Thinh H. Nguyen, Michelle M. Li, Yepeng Huang, Nitya Nair, Daniel Marbach, Marinka Zitnik

    Immune checkpoint inhibitors are standard across cancers, yet most patients do not respond and existing biomarkers generalize poorly across tumor types, drugs and clinical settings. We present C ompass , a pan-cancer foundation model that predicts immunotherapy response from bulk tumor transcriptomes using a concept-bottleneck transformer. C ompass encodes gene expression through 44 biologically grounded immune concepts representing immune cell states, tumor-microenvironment interactions, and signaling pathways. Trained on 10,184 tumors across 33 cancer types, C ompass outperforms 22 baseline methods in 16 independent clinical cohorts spanning seven cancers and six immune checkpoint inhibitors, increasing accuracy by 8.5% and area under the precision-recall curve by 15.7%, with minimal additional training. The model generalizes to unseen cancer types and treatments, supporting indication selection and patient stratification in early-phase clinical trials. In survival analyses, C ompass -stratified responders have longer overall survival (hazard ratio = 4.7, p < 0.0001). Personalized response maps connect gene expression to immune concepts, revealing mechanisms of response and resistance; in immune-inflamed non-responders, C ompass highlights programs including TGF- β signaling, endothelial exclusion, CD4+ T cell dysfunction, and B cell deficiency. By combining interpretability with transfer learning, C ompass enables robust prediction and mechanistic insight to inform trial design and translational studies.


  • Open Access

    Efficient generation of protein pockets with PocketGen

    by Zaixi Zhang, Wan Xiang Shen, Qi Liu, Marinka Zitnik

    Designing protein-binding proteins is critical for drug discovery. However, artificial-intelligence-based design of such proteins is challenging due to the complexity of protein–ligand interactions, the flexibility of ligand molecules and amino acid side chains, and sequence–structure dependencies. We introduce PocketGen, a deep generative model that produces residue sequence and atomic structure of the protein regions in which ligand interactions occur. PocketGen promotes consistency between protein sequence and structure by using a graph transformer for structural encoding and a sequence refinement module based on a protein language model. The graph transformer captures interactions at multiple scales, including atom, residue and ligand levels. For sequence refinement, PocketGen integrates a structural adapter into the protein language model, ensuring that structure-based predictions align with sequence-based predictions. PocketGen can generate high-fidelity protein pockets with enhanced binding affinity and structural validity. It operates ten times faster than physics-based methods and achieves a 97% success rate, defined as the percentage of generated pockets with higher binding affinity than reference pockets. Additionally, it attains an amino acid recovery rate exceeding 63%.


  • Open Access

    Activity Cliff-Informed Contrastive Learningfor Molecular Property Prediction

    by Wanxiang Shen, Chao Cui, Xiaorui Su, Zaixi Zhang, Alejandro Velez-Arce, Jianming Wang, Xiang Cheng Shi, Yan Bing Zhang, Jie Wu, Yu Zong Chen, et al.

    Modeling molecular activity and quantitative structure-activity relationships of chemical compounds is critical in drug design. Graph neural networks, which utilize molecular structures as frames, have shown success in assessing the biological activity of chemical compounds, guiding the selection and optimization of candidates for further development. However, current models often overlook activity cliffs (ACs)—cases where structurally similar molecules exhibit different bioactivities—due to latent spaces primarily optimized for structural features. Here, we introduce AC-awareness (ACA), an inductive bias designed to enhance molecular representation learning for activity modeling. The ACA jointly optimizes metric learning in the latent space and task performance in the target space, making models more sensitive to ACs. We develop \name, an AC-informed contrastive learning approach that can be integrated with any graph neural network. Experiments on 39 benchmark datasets demonstrate that AC-informed representations of chemical compounds consistently outperform standard models in bioactivity prediction across both regression and classification tasks. AC-informed models show strong performance in predicting pharmacokinetic and safety-relevant molecular properties. ACA paves the way toward activity-informed molecular representations, providing a valuable tool for the early stages of lead compound identification, refinement, and virtual screening.