Preference optimization has made significant progress recently, with numerous methods developed to align language models with human preferences. This paper introduces $f$-divergence Preference Optimization ($f$-PO), a novel framework that generalizes and extends existing approaches. $f$-PO minimizes $f$-divergences between the optimized policy and the optimal policy, encompassing a broad family of alignment methods using various divergences. Our approach unifies previous algorithms like DPO and EXO, while offering new variants through different choices of $f$-divergences. We provide theoretical analysis of $f$-PO’s properties and conduct extensive experiments on state-of-the-art language models using benchmark datasets. Results demonstrate $f$-PO’s effectiveness across various tasks, achieving superior performance compared to existing methods on popular benchmarks such as AlpacaEval 2, Arena-Hard, MT-Bench, and Open LLM Leaderboard v2. Additionally, we present ablation studies exploring the impact of different $f$-divergences, offering insights into the trade-offs between regularization and performance in offline preference optimization. Our work contributes both practical algorithms and theoretical understanding to the field of language model alignment. Code is available at \url{https://github.com/MinkaiXu/fPO.}

Geometric diffusion models have shown remarkable success in molecular dynamics and structure generation. However, efficiently fine-tuning them for downstream tasks with varying geometric controls remains underexplored. In this work, we propose an SE(3)-equivariant adapter framework (GeoAda) that enables flexible and parameter-efficient fine-tuning for controlled generative tasks without modifying the original model architecture. GeoAda introduces a structured adapter design: control signals are first encoded through coupling operators, then processed by a trainable copy of selected base model layers, and finally projected back via decoupling operators followed by an equivariant zero-initialized convolution. By fine-tuning only these lightweight adapter modules, GeoAda preserves the model’s geometric consistency while mitigating overfitting and catastrophic forgetting. We theoretically prove that the proposed adapters maintain SE(3)-equivariance, ensuring that the geometric inductive biases of the pretrained diffusion model remain intact during adaptation. We demonstrate the wide applicability of \method across diverse geometric control types, including frame control, global control, subgraph control, and a broad range of application domains such as particle dynamics, molecular dynamics, human motion prediction, and molecule generation. Empirical results show that GeoAda achieves state-of-the-art fine-tuning performance while preserving original task accuracy, whereas other baselines experience significant performance degradation due to overfitting and catastrophic forgetting.

Protein-protein interactions (PPIs) are fundamental to biological processes, yet experimental determination of PPIs remains costly and labor-intensive. While computational methods have emerged as promising alternatives, sequence-based approaches face critical challenges: (1) effectively capturing long-range dependencies and critical biochemical patterns in variable-length sequences, and (2) balancing computational efficiency with sensitivity to subtle residue-level interactions. Here, we present Dual Protein Embedding-based Graph Model (DPEG), which leverages dynamic graph attention networks to enable robust sequence-driven PPI prediction. Unlike structure-dependent methods, DPEG operates solely on sequence data, bypassing the need for structural or domain annotations. Specifically, we employ ESM-2 to transform sequences into residue-level graphs, preserving evolutionary and physicochemical context. To address variable sequence lengths, we design a module that can represent protein sequences of arbitrary lengths as graph networks at the amino acid level. Further, a gated attention mechanism is introduced to adaptively refining residue representations. Finally, a dynamic attention mechanism prioritizes functionally critical motifs within the graph. Evaluated on four diverse PPI datasets spanning different species and interaction types, DPEG achieves state-of-the-art performance and demonstrates strong cross-dataset generalizability. By integrating deep sequence semantics with graph-based interaction modeling, DPEG advances sequence-only PPI prediction, offering a scalable and biologically plausible framework for proteome-wide studies.

Recent advancements in Large Language Models (LLMs) have significantly improved text generation capabilities, but these systems are still known to hallucinate, and granular uncertainty estimation for long-form LLM generations remains challenging. In this work, we propose Graph Uncertainty -- which represents the relationship between LLM generations and claims within them as a bipartite graph and estimates the claim-level uncertainty with a family of graph centrality metrics. Under this view, existing uncertainty estimation methods based on the concept of self-consistency can be viewed as using degree centrality as an uncertainty measure, and we show that more sophisticated alternatives such as closeness centrality provide consistent gains at claim-level uncertainty estimation. Moreover, we present uncertainty-aware decoding techniques that leverage both the graph structure and uncertainty estimates to improve the factuality of LLM generations by preserving only the most reliable claims. Compared to existing methods, our graph-based uncertainty metrics lead to an average of 6.8% relative gains on AUPRC across various long-form generation settings, and our end-to-end system provides consistent 2-4% gains in factuality over existing decoding techniques while significantly improving the informativeness of generated responses.

Predicting drug-target affinity (DTA) is a crucial step in the process of drug discovery. Efficient and accurate prediction of DTA would greatly reduce the time and economic cost of new drug development, which has encouraged the emergence of a large number of deep learning-based DTA prediction methods. In terms of the representation of target proteins, current methods can be classified into 1D sequence- and 2D-protein graph-based methods. However, both two approaches focused only on the inherent properties of the target protein, but neglected the broad prior knowledge regarding protein interactions that have been clearly elucidated in past decades. Aiming at the above issue, this work presents an end-to-end DTA prediction method named MSF-DTA (Multi-Source Feature Fusion-based Drug-Target Affinity). The contributions can be summarized as follows. First, MSF-DTA adopts a novel "neighboring feature"-based protein representation. Instead of utilizing only the inherent features of a target protein, MSF-DTA gathers additional information for the target protein from its biologically related "neighboring" proteins in PPI (i.e., protein-protein interaction) and SSN (i.e., sequence similarity) networks to get prior knowledge. Second, the representation was learned using an advanced graph pre-training framework, VGAE, which could not only gather node features but also learn topological connections, therefore contributing to a richer protein representation and benefiting the downstream DTA prediction task. This study provides new perspective for the DTA prediction task, and evaluation results demonstrated that MSF-DTA obtained superior performances compared to current state-of-the-art methods.

Deep learning shortens the cycle of the drug discovery for its success in extracting features of molecules and proteins. Generating new molecules with deep learning methods could enlarge the molecule space and obtain molecules with specific properties. However, it is also a challenging task considering that the connections between atoms are constrained by chemical rules. Aiming at generating and optimizing new valid molecules, this article proposed Molecular Substructure Tree Generative Model, in which the molecule is generated by adding substructure gradually. The proposed model is based on the Variational Auto-Encoder architecture, which uses the encoder to map molecules to the latent vector space, and then builds an autoregressive generative model as a decoder to generate new molecules from Gaussian distribution. At the same time, for the molecular optimization task, a molecular optimization model based on CycleGAN was constructed. Experiments showed that the model could generate valid and novel molecules, and the optimized model effectively improves the molecular properties.

Air pollution has been long recognized as a hazardous factor for the human cardiovascular system. Sulfur dioxide (SO2) is a common ambient air pollutant that is able to cause detrimental effects on hearts. Though the cardiotoxicity effects by sulfur dioxide were well documented in epidemiological reports, however, the underlying mechanisms remain unclear owing to the technical limitations that exist in traditional experimental measures. In this article, we developed a multi-scale virtual ventricular tissue, which incorporated electrophysiological activities from subcellular to tissue levels and could provide comprehensive records and insightful mechanisms of the SO2induced ventricular arrhythmias. Based on the available cellular and molecular experimental data, our findings provide a rationale at tissue level in support of epidemiologic studies pointing to the deleterious effects of SO2pollution on cardiac function.

BACKGROUND: Binding sites are the pockets of proteins that can bind drugs; the discovery of these pockets is a critical step in drug design. With the help of computers, protein pockets prediction can save manpower and financial resources. RESULTS: In this paper, a novel protein descriptor for the prediction of binding sites is proposed. Information on non-bonded interactions in the three-dimensional structure of a protein is captured by a combination of geometry-based and energy-based methods. Moreover, due to the rapid development of deep learning, all binding features are extracted to generate three-dimensional grids that are fed into a convolution neural network. Two datasets were introduced into the experiment. The sc-PDB dataset was used for descriptor extraction and binding site prediction, and the PDBbind dataset was used only for testing and verification of the generalization of the method. The comparison with previous methods shows that the proposed descriptor is effective in predicting the binding sites. CONCLUSIONS: A new protein descriptor is proposed for the prediction of the drug binding sites of proteins. This method combines the three-dimensional structure of a protein and non-bonded interactions with small molecules to involve important factors influencing the formation of binding site. Analysis of the experiments indicates that the descriptor is robust for site prediction.