Graph Neural Networks (GNNs) have been widely adopted for Protein Representation Learning (PRL), as residue interaction networks can be naturally represented as graphs. Current GNN-based PRL methods typically rely on single-perspective graph construction strategies, which capture partial properties of residue interactions, resulting in incomplete protein representations. To address this limitation, we propose MMPG, a framework that constructs protein graphs from multiple perspectives and adaptively fuses them via Mixture of Experts (MoE) for PRL. MMPG constructs graphs from physical, chemical, and geometric perspectives to characterize different properties of residue interactions. To capture both perspective-specific features and their synergies, we develop an MoE module, which dynamically routes perspectives to specialized experts, where experts learn intrinsic features and cross-perspective interactions. We quantitatively verify that MoE automatically specializes experts in modeling distinct levels of interaction—from individual representations, to pairwise inter-perspective synergies, and ultimately to a global consensus across all perspectives. Through integrating this multi-level information, MMPG produces superior protein representations and achieves advanced performance on four different downstream protein tasks.

An informative, creative title prefix is memorable, capable of capturing the attention of readers, and significantly enhances the potential for increased citations. In this work, we pioneer the exploration of the significance of title prefixes in academic papers and propose a controllable title prefix generation model based on curriculum learning. Specifically, we introduce a dedicated dataset named TPOA to compensate for the lack of training data for this emerging task. To make the model capture relevant patterns and language structure, we design three title prefix generation tasks (abstract-based, title-based, and title&abstract-based) to train a ByT5 model into a curriculum learning structure as a generator. After that, we fine-tune another ByT5 model on a target style corpora as a discriminator to control the style of the generated title prefix. Through extensive experiments, our proposed model outperforms existing methods on both human evaluation and automatic evaluation, demonstrating its effectiveness.

Graph Contrastive Learning (GCL) improves Graph Neural Network (GNN)-based protein representation learning by enhancing its generalization and robustness. Existing GCL approaches for protein representation learning rely on 2D topology, where graph augmentation is solely based on topological features, ignoring the intrinsic biological properties of proteins. Besides, 3D structure-based protein graph augmentation remains unexplored, despite proteins inherently exhibiting 3D structures. To bridge this gap, we propose novel biology-aware graph augmentation strategies from the perspective of invariance and integrate them into the protein GCL framework. Specifically, we introduce Functional Community Invariance (FCI)-based graph augmentation, which employs spectral constraints to preserve topology-driven community structures while incorporating residue-level chemical similarity as edge weights to guide edge sampling and maintain functional communities. Furthermore, we propose 3D Protein Structure Invariance (3-PSI)-based graph augmentation, leveraging dihedral angle perturbations and secondary structure rotations to retain critical 3D structural information of proteins while diversifying graph views. Extensive experiments on four different protein-related tasks demonstrate the superiority of our proposed GCL protein representation learning framework.

Multi-modal emotion recognition aims to recognize different emotions precisely by utilizing different modality information. Current multi-modal methods always suffer the pain of the modality gap. Besides, many works use attention mechanisms to fuse different modalities features, which reduces the diversity of the modalities. To deal with these issues, we propose an inter-modality and intra-sample alignment framework for multi-modal emotion recognition. Specifically, we first map text features and audio features to a shared subspace and align them using a cross-modal self-attention mechanism and Maximum Mean Discrepancy matrix. To preserve the diversity of each modality, we extract modality-specific features using different encoders. Finally, we employ supervised contrastive learning techniques to align the features at a sample-level. Extensive experiments indicate that our method achieves state-of-the-art performance and effectiveness.