arXiv subjects

Qianqian Song

Publications and source records attributed to Qianqian Song.

16 recordsLinked to original sources

arXiv2026

Signed Rescue Routing: Harm-Aware Cascades for Efficient LLM Inference

Large language model (LLM) cascades answer easy requests with a small model and escalate selected requests to a larger model. Most routers prioritize examples on which the small model appears uncertain or likely to be wrong. This proxy ignores a decisive fact: escalation is useful only when the large model corrects the small model, and it is harmful when the large model replaces a correct answer with an incorrect one. We introduce Signed Rescue Routing (SRR), a budgeted routing method that predicts these two events separately and ranks requests by their difference. We show that this signed conditional gain is the Bayes-optimal routing score under a fixed escalation budget. SRR requires only the small model's output statistics at deployment and adds a lightweight two-head router. We evaluate SRR with Qwen3-4B and Qwen3-8B on TBD examples from MMLU, HellaSwag, and ARC-Challenge. Across the accuracy-compute curve, SRR reaches an area of TBD, compared with TBD for a learned small-model error predictor and TBD for entropy routing. These results show that predicting incremental value, rather than model uncertainty, is a simple and effective objective for efficient LLM cascades.

cs.CL↗

arXiv2026

UniD$^3$: A Knowledge Graph-Enhanced RAG Framework for Drug-Disease Discovery and Reasoning

Systematic characterization of drug-disease relationships is essential for drug discovery and repurposing, yet is hindered by the heterogeneity and rapid growth of biomedical literature. Existing datasets rely on labor-intensive curation and are often incomplete, while LLM-only approaches suffer from hallucination and weak evidence grounding. We introduce UniD$^3$, a unified framework that integrates Large Language Models with Knowledge Graph-enhanced Retrieval-Augmented Generation (KG-RAG) to extract, organize, and validate drug-disease knowledge across Drug-Disease Matching (DDM), Drug Effectiveness Assessment (DEA), and Drug-Target Analysis (DTA). UniD$^3$ processes 157,849 PubMed articles with Llama 3.3-70B and constructs knowledge graphs via a dual-stage strategy combining paper-level extraction with KG-level consolidation centered on drug and disease entities. These graphs support KG-RAG-based generation of structured datasets, evaluated through external benchmarks, fuzzy matching with curated resources, and clinician review. UniD$^3$ produces six knowledge graphs and large-scale datasets, including 28,915 DDM, 15,042 DEA, and over 4,000 DTA QA pairs. External validation shows strong performance (F1: 0.85-0.87 for DDM/DEA; 0.82 for DTA), with clinician review confirming high reliability (AUROC = 0.90). KG-RAG-augmented models outperform standalone LLMs, and the UniD$^3$ chatbot enables interpretable, citation-supported exploration of drug-disease relationships. UniD$^3$ provides a scalable, extensible framework for transforming unstructured biomedical literature into high-quality, structured drug-disease knowledge, supporting AI-driven discovery, repurposing, and precision medicine.

cs.CL↗

arXiv2026

DrugClaw and DrugAudit: A Primary-Source-Grounded Agent and Authority-Aware Benchmark for Drug-Information Question Answering

Drug-information question answering is a high-stakes setting where hallucinated facts can mislead clinical decision-making and the provenance of each cited fact matters as much as the fact itself. We present DrugClaw, a multi-agent retrieval-augmented system that queries a registry of drug and pharmacovigilance skills via a reflection-driven state-machine workflow and returns answers grounded in primary regulatory or peer-reviewed records. We also contribute DrugAudit, a 3,772-item authority-aware benchmark with an evaluation panel that scores upstream-of-gold source match, token-level semantic snippet overlap, and citation faithfulness under a dual-judge LLM-as-judge protocol with inter-judge kappa = 0.88 (almost-perfect). Across DrugAudit plus drug-related subsets of MedQA (751) and PubMedQA (512), DrugClaw is top-1 on every column of the headline table: composite Evidence Index under both judges, judge-mediated answer correctness, primary-source rate (0.918, +10.1 pp over next-best), faithfulness (0.887, +5.9 pp), MedQA (0.920), and PubMedQA (0.693).

cs.CL↗

arXiv2025

SoK: Can Fully Homomorphic Encryption Support General AI Computation? A Functional and Cost Analysis

Artificial intelligence (AI) increasingly powers sensitive applications in domains such as healthcare and finance, relying on both linear operations (e.g., matrix multiplications in large language models) and non-linear operations (e.g., sorting in retrieval-augmented generation). Fully homomorphic encryption (FHE) has emerged as a promising tool for privacy-preserving computation, but it remains unclear whether existing methods can support the full spectrum of AI workloads that combine these operations. In this SoK, we ask: Can FHE support general AI computation? We provide both a functional analysis and a cost analysis. First, we categorize ten distinct FHE approaches and evaluate their ability to support general computation. We then identify three promising candidates and benchmark workloads that mix linear and non-linear operations across different bit lengths and SIMD parallelization settings. Finally, we evaluate five real-world, privacy-sensitive AI applications that instantiate these workloads. Our results quantify the costs of achieving general computation in FHE and offer practical guidance on selecting FHE methods that best fit specific AI application requirements. Our codes are available at https://github.com/UCF-ML-Research/FHE-AI-Generality.

cs.CR↗

arXiv2025

scDrugMap: Benchmarking Large Foundation Models for Drug Response Prediction

Drug resistance presents a major challenge in cancer therapy. Single cell profiling offers insights into cellular heterogeneity, yet the application of large-scale foundation models for predicting drug response in single cell data remains underexplored. To address this, we developed scDrugMap, an integrated framework featuring both a Python command-line interface and a web server for drug response prediction. scDrugMap evaluates a wide range of foundation models, including eight single-cell models and two large language models, using a curated dataset of over 326,000 cells in the primary collection and 18,800 cells in the validation set, spanning 36 datasets and diverse tissue and cancer types. We benchmarked model performance under pooled-data and cross-data evaluation settings, employing both layer freezing and Low-Rank Adaptation (LoRA) fine-tuning strategies. In the pooled-data scenario, scFoundation achieved the best performance, with mean F1 scores of 0.971 (layer freezing) and 0.947 (fine-tuning), outperforming the lowest-performing model by over 50%. In the cross-data setting, UCE excelled post fine-tuning (mean F1: 0.774), while scGPT led in zero-shot learning (mean F1: 0.858). Overall, scDrugMap provides the first large-scale benchmark of foundation models for drug response prediction in single-cell data and serves as a user-friendly, flexible platform for advancing drug discovery and translational research.

cs.AI↗

arXiv2025

PhenoProfiler: Advancing Phenotypic Learning for Image-based Drug Discovery

In the field of image-based drug discovery, capturing the phenotypic response of cells to various drug treatments and perturbations is a crucial step. However, existing methods require computationally extensive and complex multi-step procedures, which can introduce inefficiencies, limit generalizability, and increase potential errors. To address these challenges, we present PhenoProfiler, an innovative model designed to efficiently and effectively extract morphological representations, enabling the elucidation of phenotypic changes induced by treatments. PhenoProfiler is designed as an end-to-end tool that processes whole-slide multi-channel images directly into low-dimensional quantitative representations, eliminating the extensive computational steps required by existing methods. It also includes a multi-objective learning module to enhance robustness, accuracy, and generalization in morphological representation learning. PhenoProfiler is rigorously evaluated on large-scale publicly available datasets, including over 230,000 whole-slide multi-channel images in end-to-end scenarios and more than 8.42 million single-cell images in non-end-to-end settings. Across these benchmarks, PhenoProfiler consistently outperforms state-of-the-art methods by up to 20%, demonstrating substantial improvements in both accuracy and robustness. Furthermore, PhenoProfiler uses a tailored phenotype correction strategy to emphasize relative phenotypic changes under treatments, facilitating the detection of biologically meaningful signals. UMAP visualizations of treatment profiles demonstrate PhenoProfiler ability to effectively cluster treatments with similar biological annotations, thereby enhancing interpretability. These findings establish PhenoProfiler as a scalable, generalizable, and robust tool for phenotypic learning.

cs.LG↗

arXiv2025

HECLIP: Histology-Enhanced Contrastive Learning for Imputation of Transcriptomics Profiles

Histopathology, particularly hematoxylin and eosin (H\&E) staining, plays a critical role in diagnosing and characterizing pathological conditions by highlighting tissue morphology. However, H\&E-stained images inherently lack molecular information, requiring costly and resource-intensive methods like spatial transcriptomics to map gene expression with spatial resolution. To address these challenges, we introduce HECLIP (Histology-Enhanced Contrastive Learning for Imputation of Profiles), an innovative deep learning framework that bridges the gap between histological imaging and molecular profiling. HECLIP is specifically designed to infer gene expression profiles directly from H\&E-stained images, eliminating the need for expensive spatial transcriptomics assays. HECLIP leverages an advanced image-centric contrastive loss function to optimize image representation learning, ensuring that critical morphological patterns in histology images are effectively captured and translated into accurate gene expression profiles. This design enhances the predictive power of the image modality while minimizing reliance on gene expression data. Through extensive benchmarking on publicly available datasets, HECLIP demonstrates superior performance compared to existing approaches, delivering robust and biologically meaningful predictions. Detailed ablation studies further underscore its effectiveness in extracting molecular insights from histology images. Additionally, HECLIP's scalable and cost-efficient approach positions it as a transformative tool for both research and clinical applications, driving advancements in precision medicine. The source code for HECLIP is openly available at https://github.com/QSong-github/HECLIP.

cs.CE↗

arXiv2024

Disentangle VAE for Molecular Generation

Automatic molecule generation plays an important role on drug discovery and has received a great deal of attention in recent years thanks to deep learning successful use. Graph-based neural network represents state of the art methods on automatic molecule generation. However, it is still challenging to generate molecule with desired properties, which is a core task in drug discovery. In this paper, we focus on this task and propose a Controllable Junction Tree Variational Autoencoder (C JTVAE), adding an extractor module into VAE framework to describe some properties of molecule. Our method is able to generate similar molecular with desired property given an input molecule. Experimental results is encouraging.

cs.CE↗

arXiv2024

TrajVis: a visual clinical decision support system to translate artificial intelligence trajectory models in the precision management of chronic kidney disease

Objective: Our objective is to develop and validate TrajVis, an interactive tool that assists clinicians in using artificial intelligence (AI) models to leverage patients' longitudinal electronic medical records (EMR) for personalized precision management of chronic disease progression. Methods: We first perform requirement analysis with clinicians and data scientists to determine the visual analytics tasks of the TrajVis system as well as its design and functionalities. A graph AI model for chronic kidney disease (CKD) trajectory inference named DEPOT is used for system development and demonstration. TrajVis is implemented as a full-stack web application with synthetic EMR data derived from the Atrium Health Wake Forest Baptist Translational Data Warehouse and the Indiana Network for Patient Care research database. A case study with a nephrologist and a user experience survey of clinicians and data scientists are conducted to evaluate the TrajVis system. Results: The TrajVis clinical information system is composed of four panels: the Patient View for demographic and clinical information, the Trajectory View to visualize the DEPOT-derived CKD trajectories in latent space, the Clinical Indicator View to elucidate longitudinal patterns of clinical features and interpret DEPOT predictions, and the Analysis View to demonstrate personal CKD progression trajectories. System evaluations suggest that TrajVis supports clinicians in summarizing clinical data, identifying individualized risk predictors, and visualizing patient disease progression trajectories, overcoming the barriers of AI implementation in healthcare. Conclusion: TrajVis bridges the gap between the fast-growing AI/ML modeling and the clinical use of such models for personalized and precision management of chronic diseases.

cs.HC↗

arXiv2023

Pathology Steered Stratification Network for Subtype Identification in Alzheimer's Disease

Alzheimer's disease (AD) is a heterogeneous, multifactorial neurodegenerative disorder characterized by beta-amyloid, pathologic tau, and neurodegeneration. There are no effective treatments for Alzheimer's disease at a late stage, urging for early intervention. However, existing statistical inference approaches of AD subtype identification ignore the pathological domain knowledge, which could lead to ill-posed results that are sometimes inconsistent with the essential neurological principles. Integrating systems biology modeling with machine learning, we propose a novel pathology steered stratification network (PSSN) that incorporates established domain knowledge in AD pathology through a reaction-diffusion model, where we consider non-linear interactions between major biomarkers and diffusion along brain structural network. Trained on longitudinal multimodal neuroimaging data, the biological model predicts long-term trajectories that capture individual progression pattern, filling in the gaps between sparse imaging data available. A deep predictive neural network is then built to exploit spatiotemporal dynamics, link neurological examinations with clinical profiles, and generate subtype assignment probability on an individual basis. We further identify an evolutionary disease graph to quantify subtype transition probabilities through extensive simulations. Our stratification achieves superior performance in both inter-cluster heterogeneity and intra-cluster homogeneity of various clinical scores. Applying our approach to enriched samples of aging populations, we identify six subtypes spanning AD spectrum, where each subtype exhibits a distinctive biomarker pattern that is consistent with its clinical outcome. PSSN provides insights into pre-symptomatic diagnosis and practical guidance on clinical treatments, which may be further generalized to other neurodegenerative diseases.

q-bio.QM↗

arXiv2023

Uncovering the System Vulnerability and Criticality of Human Brain under Dynamical Neuropathological Events in Alzheimer's Disease

Background: Despite the striking efforts in investigating neurobiological factors behind the acquisition of amyloid-\b{eta} (A), protein tau (T), and neurodegeneration ([N]) biomarkers, the mechanistic pathways of how AT[N] biomarkers spreading throughout the brain remain elusive. Objectives: To disentangle the massive heterogeneities in AD progressions and identify vulnerable/critical brain regions to AD pathology. Methods: In this work, we characterized the interaction of AT[N] biomarkers and their propagation across brain networks using a novel bistable reaction-diffusion model, which allows us to establish a new systems biology underpinning of Alzheimer's disease (AD) progression. We applied our model to large-scale longitudinal neuroimages from the ADNI database and studied the systematic vulnerability and criticality of brains. Results: Our model yields long term prediction that is statistically significant linear correlated with temporal imaging data, produces clinically consistent risk prediction, and captures the Braak-like spreading pattern of AT[N] biomarkers in AD development. Conclusion: Our major findings include (i) tau is a stronger indicator of regional risk compared to amyloid, (ii) temporal lobe exhibits higher vulnerability to AD-related pathologies, (iii) proposed critical brain regions outperform hub nodes in transmitting disease factors across the brain, and (iv) comparing the spread of neuropathological burdens caused by amyloid-\b{eta} and tau diffusions, disruption of metabolic balance is the most determinant factor contributing to the initiation and progression of Alzheimer's disease.

q-bio.QM↗

arXiv2022

Uncovering the dynamic effects of DEX treatment on lung cancer by integrating bioinformatic inference and multiscale modeling of scRNA-seq and proteomics data

Motivation: Lung cancer is one of the leading causes for cancer-related death, with a five-year survival rate of 18%. It is a priority for us to understand the underlying mechanisms that affect the implementation and effectiveness of lung cancer therapeutics. In this study, we combine the power of Bioinformatics and Systems Biology to comprehensively uncover functional and signaling pathways of drug treatment using bioinformatics inference and multiscale modeling of both scRNA-seq data and proteomics data. The innovative and cross-disciplinary approach can be further applied to other computational studies in tumorigenesis and oncotherapy. Results: A time series of lung adenocarcinoma-derived A549 cells after DEX treatment were analysed. (1) We first discovered the differentially expressed genes in those lung cancer cells. Then through the interrogation of their regulatory network, we identified key hub genes including TGF-\b{eta}, MYC, and SMAD3 varied underlie DEX treatment. Further enrichment analysis revealed the TGF-\b{eta} signaling pathway as the top enriched term. Those genes involved in the TGF-\b{eta} pathway and their crosstalk with the ERBB pathway presented a strong survival prognosis in clinical lung cancer samples. (2) Based on biological validation and further curation, a multiscale model of tumor regulation centered on both TGF-\b{eta}-induced and ERBB-amplified signaling pathways was developed to characterize the dynamics effects of DEX therapy on lung cancer cells. Our simulation results were well matched to available data of SMAD2, FOXO3, TGF\b{eta}1, and TGF\b{eta}R1 over the time course. Moreover, we provided predictions of different doses to illustrate the trend and therapeutic potential of DEX treatment.

q-bio.QM↗

arXiv2022

Approximate Inference via Clustering

In recent years, large-scale Bayesian learning draws a great deal of attention. However, in big-data era, the amount of data we face is growing much faster than our ability to deal with it. Fortunately, it is observed that large-scale datasets usually own rich internal structure and is somewhat redundant. In this paper, we attempt to simplify the Bayesian posterior via exploiting this structure. Specifically, we restrict our interest to the so-called well-clustered datasets and construct an \emph{approximate posterior} according to the clustering information. Fortunately, the clustering structure can be efficiently obtained via a particular clustering algorithm. When constructing the approximate posterior, the data points in the same cluster are all replaced by the centroid of the cluster. As a result, the posterior can be significantly simplified. Theoretically, we show that under certain conditions the approximate posterior we construct is close (measured by KL divergence) to the exact posterior. Furthermore, thorough experiments are conducted to validate the fact that the constructed posterior is a good approximation to the true posterior and much easier to sample from.

cs.LG↗

arXiv2022

Splitting numerical integration for matrix completion

Low rank matrix approximation is a popular topic in machine learning. In this paper, we propose a new algorithm for this topic by minimizing the least-squares estimation over the Riemannian manifold of fixed-rank matrices. The algorithm is an adaptation of classical gradient descent within the framework of optimization on manifolds. In particular, we reformulate an unconstrained optimization problem on a low-rank manifold into a differential dynamic system. We develop a splitting numerical integration method by applying a splitting integration scheme to the dynamic system. We conduct the convergence analysis of our splitting numerical integration algorithm. It can be guaranteed that the error between the recovered matrix and true result is monotonically decreasing in the Frobenius norm. Moreover, our splitting numerical integration can be adapted into matrix completion scenarios. Experimental results show that our approach has good scalability for large-scale problems with satisfactory accuracy

cs.LG↗

arXiv2021

Flexible Principal Component Analysis for Exponential Family Distributions

Traditional principal component analysis (PCA) is well known in high-dimensional data analysis, but it requires to express data by a matrix with observations to be continuous. To overcome the limitations, a new method called flexible PCA (FPCA) for exponential family distributions is proposed. The goal is to ensure that it can be implemented to arbitrary shaped region for either count or continuous observations. The methodology of FPCA is developed under the framework of generalized linear models. It provides statistical models for FPCA not limited to matrix expressions of the data. A maximum likelihood approach is proposed to derive the decomposition when the number of principal components (PCs) is known. This naturally induces a penalized likelihood approach to determine the number of PCs when it is unknown. By modifying it for missing data problems, the proposed method is compared with previous PCA methods for missing data. The simulation study shows that the performance of FPCA is always better than its competitors. The application uses the proposed method to reduce the dimensionality of arbitrary shaped sub-regions of images and the global spread patterns of COVID-19 under normal and Poisson distributions, respectively.

stat.ME↗

arXiv2014

The Academic Social Network

Through academic publications, the authors of these publications form a social network. Instead of sharing casual thoughts and photos (as in Facebook), authors pick co-authors and reference papers written by other authors. Thanks to various efforts (such as Microsoft Libra and DBLP), the data necessary for analyzing the academic social network is becoming more available on the Internet. What type of information and queries would be useful for users to find out, beyond the search queries already available from services such as Google Scholar? In this paper, we explore this question by defining a variety of ranking metrics on different entities -authors, publication venues and institutions. We go beyond traditional metrics such as paper counts, citations and h-index. Specifically, we define metrics such as influence, connections and exposure for authors. An author gains influence by receiving more citations, but also citations from influential authors. An author increases his/her connections by co-authoring with other authors, and specially from other authors with high connections. An author receives exposure by publishing in selective venues where publications received high citations in the past, and the selectivity of these venues also depends on the influence of the authors who publish there. We discuss the computation aspects of these metrics, and similarity between different metrics. With additional information of author-institution relationships, we are able to study institution rankings based on the corresponding authors' rankings for each type of metric as well as different domains. We are prepared to demonstrate these ideas with a web site (http://pubstat.org) built from millions of publications and authors.

cs.SI↗