While Large Language Models (LLMs) are emerging as a promising direction in computational pathology, the substantial computational cost of giga-pixel Whole Slide Images (WSIs) necessitates the use of Multi-Instance Learning (MIL) to enable effective modeling. A key challenge is that pathological tasks typically provide only bag-level labels, while instance-level descriptions generated by LLMs often suffer from bias due to a lack of fine-grained medical knowledge. To address this, we propose that constructing task-specific pathological entity prototypes is crucial for learning generalizable features and enhancing model interpretability. Furthermore, existing vision-language MIL methods often employ unidirectional guidance, limiting cross-modal synergy. In this paper, we introduce a novel approach, Multimodal Prototype-based Multi-Instance Learning, that promotes bidirectional interaction through a balanced information compression scheme. Specifically, we leverage a frozen LLM to generate task-specific pathological entity descriptions, which are learned as text prototypes. Concurrently, the vision branch learns instance-level prototypes to mitigate the model's reliance on redundant data. For the fusion stage, we employ the Stereoscopic Optimal Transport (SOT) algorithm, which is based on a similarity metric, thereby facilitating broader semantic alignment in a higher-dimensional space. We conduct few-shot classification and explainability experiments on three distinct cancer datasets, and the results demonstrate the superior generalization capabilities of our proposed method.

OBJECTIVE: We aim to improve Drug-Drug Interactions (DDIs) by explicitly injecting medicinal-chemistry knowledge of functional groups (FGs) into graph neural network (GNN) message passing, in both transductive and inductive settings. Our goal is to (i) encode FG priors in a trainable way that enhances representation quality without handcrafting features, and (ii) yield interpretable attributions that align learned weights with pharmacologically meaningful FG patterns. METHODS: We introduce FG-DDI, a dual-view GNN that augments both intra- and inter-molecular reasoning. At the intra-molecular level, atom/bond messages are scaled by FG enrichment weights derived from detected FG motifs within each drug graph. At the inter-molecular level, a bipartite message-passing layer between a drug pair is modulated by FG-FG enrichment scores that reflect empirical co-occurrence in known DDIs. Enrichment is computed as odds ratios from corpus statistics and injected via learnable gates, ensuring differentiability and allowing data to override noisy priors. We couple this with standard supervision on interaction labels and report accuracy (ACC), AUROC, average precision (AP), and F1. Experiments use DrugBank (1706 drugs; 86 interaction types) and TwoSides (filtered triplets) under transductive and inductive splits (one unseen; both unseen). We perform ablations removing each FG term to isolate contributions and assess stability across splits. RESULTS: Comprehensive experiments on DrugBank and TwoSides datasets demonstrate that FG-DDI achieves superior performance compared to state-of-the-art methods. For DrugBank, the accuracy improves by 0.36% in transductive settings and by 0.46% and 1.42% in inductive settings, respectively for S1 and S2 partitioning. CONCLUSION: By systematically integrating chemical domain knowledge into deep learning architectures, this approach enables better generalization to unseen drug combinations while maintaining computational efficiency, making it particularly valuable for real-world pharmaceutical applications where new drugs continuously enter the market.

Significant disparities between the features of natural images and those inherent to histopathological images make it challenging to directly apply and transfer pre-trained models from natural images to histopathology tasks. Moreover, the frequent lack of annotations in histopathology patch images has driven researchers to explore self-supervised learning methods like mask reconstruction for learning representations from large amounts of unlabeled data. Crucially, previous mask-based efforts in self-supervised learning have often overlooked the spatial interactions among entities, which are essential for constructing accurate representations of pathological entities. To address these challenges, constructing graphs of entities is a promising approach. In addition, the diffusion reconstruction strategy has recently shown superior performance through its random intensity noise addition technique to enhance the robust learned representation. Therefore, we introduce H-MGDM, a novel self-supervised Histopathology image representation learning method through the Dynamic Entity-Masked Graph Diffusion Model. Specifically, we propose to use complementary subgraphs as latent diffusion conditions and self-supervised targets respectively during pre-training. We note that the graph can embed entities' topological relationships and enhance representation. Dynamic conditions and targets can improve pathological fine reconstruction. Our model has conducted pretraining experiments on three large histopathological datasets. The advanced predictive performance and interpretability of H-MGDM are clearly evaluated on comprehensive downstream tasks such as classification and survival analysis on six datasets.

The distribution of data in the real world is often imbalanced, with a small number of classes having a large number of instances, while most other classes have relatively few sample instances, resulting in a long-tailed distribution. Supervised contrastive learning still tends to favor classes with more samples, making it challenging to learn discriminative and robust features in imbalanced settings. In our work, we theoretically analyze the advantages of hard samples in contrastive learning and propose an end-to-end training framework based on class-aware perturbation. Specifically, we introduce different levels of perturbation based on class priors to construct hard samples, improve the disadvantaged position of tail classes, balance the feature space distribution, and dynamically adjust the classifier output with a combination of local batch priors and global priors. Experimental results show that our method achieves competitive performance across multiple long-tailed benchmark datasets.

In the real world, the data collected naturally are often long-tailed, which inevitably leads to class-imbalanced prediction and performance degradation. As a simple and effective data augmentation method, Mixup has been proven to be beneficial for the tail class in recent long-tail learning studies. However, samples selected by Mixup are still imbalanced, which could exacerbate the imbalance problem. Existing work always considers adjusting the input distribution to alleviate this problem, while this may lead to over-fitting in the tail class. In this paper, we detail the theoretical analysis of the data imbalance caused by Mixup, and propose a novel Balanced Mixup (BaMix) loss function from the output perspective. By adding a balance term, we theoretically prove the BaMix loss can overcome the imbalance caused by Mixup. In the experiments, our solution achieves the state-of-the-art performance on CIFAR-LT, ImageNet-LT, and iNaturalist 2018.