AI and Cancer Research ── Cancer biology ── 2026-10-26
Cancer molecular biology and AI
- Abstract
Single-cell glycan and RNA sequencing (scGR-seq) enables characterization of cellular glycan heterogeneity. Here, we present scGlycoAI, an interpretable machine learning framework that integrates scGR-seq data for glycan-centered biological discovery. Using lectin-binding profiles as input, scGlycoAI employs hierarchical machine learning to accurately classify cell types and subtypes within complex tumor microenvironments. Applied to pancreatic ductal adenocarcinoma (PDAC) datasets, scGlycoAI achieved an overall eight-class classification accuracy of 90.3% under leave-one-patient-out validation using glycan features alone. These results demonstrate that glycan signatures, despite their lower dimensionality compared with transcriptomic profiles, are sufficient for robust cell-type annotation and represent cell-type-specific molecular signatures. Feature importance analysis identified PVL lectin binding, which reflects recognition of sialylated and/or GlcNAc-containing glycans, as a key feature of aggressive cancer states. A neural-network model combined with SHAP-based interpretation further revealed that elevated PVL binding was associated with transcriptional programs linked to epithelial-mesenchymal transition, KRAS signaling, extracellular matrix remodeling, and inflammatory responses. Higher expression of the PVL-associated gene signature was associated with poorer overall survival in the TCGA pancreatic adenocarcinoma cohort. Collectively, these findings suggest that glycan-based machine learning enables accurate cell classification while providing biological insights into malignancy-associated glycan states at single-cell resolution. This framework provides a foundation for glycan-centered biomarker discovery and glycan-targeted therapeutic development.
Journal IF-equivalent: 2.3 (OpenAlex 2-year mean citedness, value as of 2026-10-25, retrieved 2026-10-26; not the official Clarivate IF)Reference: Anh DXT, Tateno H. scGlycoAI: Machine learning analysis of single-cell glycan and RNA sequencing data for cell-type annotation and malignancy-associated glycan discovery. Glycobiology. 2026 Oct 24 [Epub ahead of print]. doi:10.1093/glycob/cwag100. PMID: 42925266.Checked: Abstract only - Abstract
In clinical liver MRI, diffusion-weighted imaging (DWI) and dynamic contrast-enhanced (DCE) imaging are commonly used and complementary sequences, particularly for liver tumor diagnosis and assessment. Accurate registration between the two modalities facilitates subsequent quantitative analysis and multi-modal image-based evaluation. Although deep learning has achieved remarkable progress in image registration, considerable challenges remain in multi-parametric liver imaging, mainly due to substantial contrast differences, pronounced respiratory motion, limited single-center training data, and inter-center heterogeneity. This study proposes a cross-modal registration framework for liver DWI and DCE across multi-center datasets. First, a multi-task synthesis-registration network (SynReg) is introduced, which integrates image synthesis and registration within a unified architecture to reduce cross-modal appearance discrepancy and facilitate deformable registration. To further improve personalized learning under cross-center heterogeneity, we propose a personalized decentralized distribution matching (PDDM) framework with SynReg as the backbone. PDDM performs compatibility-aware selective collaboration by identifying compatible external training cases and constructing client-specific collaboration relationships. Signal-to-noise ratio (SNR) distributions are used as a lightweight compatibility descriptor, while the synthesis components remain local and the registration network is selectively aggregated among compatible clients. The framework was evaluated on six datasets collected from five medical centers and acquired on MRI systems from three vendors under both low- and high-b-value settings. PDDM consistently achieved more favorable registration performance than single-center training and representative personalized federated learning methods, while remaining closely competitive with centralized pooled-data training. In comparisons with alternative matching strategies, PDDM showed the most consistently favorable overall performance across centers, b-value settings, and registration metrics. These findings support compatibility-aware selective collaboration as an effective strategy for personalized multi-center liver image registration. Code will be available at https://github.com/ZIC-skyzrc/PDDM-SynReg.
Journal IF-equivalent: 12.5 (OpenAlex 2-year mean citedness, value as of 2026-10-25, retrieved 2026-10-26; not the official Clarivate IF)Reference: Zheng R, Yu H, Wu X, Xia X, Li R, Wang Q, et al. Distribution matching based personalized federated learning for multi-center, cross-modal liver image registration. Medical Image Analysis. 2027 Feb;116:104365. doi:10.1016/j.media.2026.104365.Checked: Abstract only