Hongxin Xiang | Search PubMed

PubMed2026

SurvGRN: a multi-feature fusion framework for bladder cancer survival prediction.

Bladder cancer survival outcomes exhibit significant heterogeneity, influenced by multifaceted factors. While digital pathology-based survival models leveraging artificial intelligence show promise, they often overlook complementary data sources. Conversely, imaging lacks cellular detail, and genomics/proteomics entail complexity and cost. To integrate multidimensional data for enhanced survival prediction, we propose SurvGRN, a multi-feature fusion framework. SurvGRN synergistically combines clinical variables, transcriptomics, and digital pathology slides using a gated residual network architecture. Pathological features are extracted via multiple instance learning, while clinical and transcriptomic data are processed as static inputs. These features are dynamically fused using a long short-term memory (LSTM) network for comprehensive survival risk assessment. Evaluated on 400 bladder cancer patients, SurvGRN significantly outperformed existing methods: improving the C-index by 12.6% over DeepMISL; 20.6% and 7.1% over graph-based models (DeepGraphConv and Patch-GCN); and 5.4% and 4.0% over attention-based approaches (Surformer and HVTSurv). Ablation studies confirmed the contributions of pathology features (extracted via ResNet-50 pre-trained on bladder tissue), clinical/transcriptomic data, and the LSTM fusion. SurvGRN also enabled significant stratification of patients into distinct risk cohorts. This work demonstrates that holistic integration of multi-source data through tailored fusion architectures substantially improves bladder cancer survival prediction.

bladder cancer↗

PubMed2025

An image-based protein-ligand binding representation learning framework via multi-level flexible dynamics trajectory pre-training.

MOTIVATION: Accurate prediction of protein-ligand binding (PLB) relationships plays a crucial role in drug discovery, which helps identify drugs that modulate the activity of specific targets. Traditional biological assays for measuring PLB relationships are time consuming and costly. In addition, models for predicting PLB relationships have been developed and widely used in drug discovery tasks. However, learning more accurate PLB representations is essential to meet the stringent standards required for drug discovery. RESULTS: We propose an image-based PLB representation learning framework, called ImagePLB, which equips ligand representation learner (LRL) and protein representation learner (PRL) to accept 3D multi-view ligand images and protein graphs as input, respectively, and learns rich interaction information between ligand and protein through a binding representation learner (BRL). Considering the scarcity of protein-ligand pairs, we further propose a multi-level next trajectory prediction (MLNTP) task to pre-train ImagePLB on the 4D flexible dynamics trajectory of 16 972 complexes, including ligand level, protein level, and complex level, to learn information related to trajectories. Besides, by introducing trajectory regularization (TR), we effectively alleviate the problem of high (even almost identical) feature similarity caused by adjacent trajectories. Compared with the current state-of-the-art methods, ImagePLB has achieved competitive improvements on PLB-related prediction tasks, including protein-ligand affinity and efficacy prediction tasks. This study opens the door to the image-based PLB learning paradigm. AVAILABILITY AND IMPLEMENTATION: All data and implementation details of code can be obtained from https://github.com/HongxinXiang/ImagePLB.

Ligands↗