Searcharxiv⌕ Search

arXiv subjects

Xinghua Lu

Publications and source records attributed to Xinghua Lu.

16 recordsLinked to original sources
arXiv2025

Knowledge-guided Contextual Gene Set Analysis Using Large Language Models

Gene set analysis (GSA) is a foundational approach for interpreting genomic data of diseases by linking genes to biological processes. However, conventional GSA methods overlook clinical context of the analyses, often generating long lists of enriched pathways with redundant, nonspecific, or irrelevant results. Interpreting these requires extensive, ad-hoc manual effort, reducing both reliability and reproducibility. To address this limitation, we introduce cGSA, a novel AI-driven framework that enhances GSA by incorporating context-aware pathway prioritization. cGSA integrates gene cluster detection, enrichment analysis, and large language models to identify pathways that are not only statistically significant but also biologically meaningful. Benchmarking on 102 manually curated gene sets across 19 diseases and ten disease-related biological mechanisms shows that cGSA outperforms baseline methods by over 30%, with expert validation confirming its increased precision and interpretability. Two independent case studies in melanoma and breast cancer further demonstrate its potential to uncover context-specific insights and support targeted hypothesis generation.

q-bio.GN↗
arXiv2021

Understanding Heart-Failure Patients EHR Clinical Features via SHAP Interpretation of Tree-Based Machine Learning Model Predictions

Heart failure (HF) is a major cause of mortality. Accurately monitoring HF progress and adjust therapies are critical for improving patient outcomes. An experienced cardiologist can make accurate HF stage diagnoses based on combination of symptoms, signs, and lab results from the electronic health records (EHR) of a patient, without directly measuring heart function. We examined whether machine learning models, more specifically the XGBoost model, can accurately predict patient stage based on EHR, and we further applied the SHapley Additive exPlanations (SHAP) framework to identify informative features and their interpretations. Our results indicate that based on structured data from EHR, our models could predict patients' ejection fraction (EF) scores with moderate accuracy. SHAP analyses identified informative features and revealed potential clinical subtypes of HF. Our findings provide insights on how to design computing systems to accurately monitor disease progression of HF patients through continuously mining patients' EHR data.

cs.LG↗
arXiv2020

Learning Latent Causal Structures with a Redundant Input Neural Network

Most causal discovery algorithms find causal structure among a set of observed variables. Learning the causal structure among latent variables remains an important open problem, particularly when using high-dimensional data. In this paper, we address a problem for which it is known that inputs cause outputs, and these causal relationships are encoded by a causal network among a set of an unknown number of latent variables. We developed a deep learning model, which we call a redundant input neural network (RINN), with a modified architecture and a regularized objective function to find causal relationships between input, hidden, and output variables. More specifically, our model allows input variables to directly interact with all latent variables in a neural network to influence what information the latent variables should encode in order to generate the output variables accurately. In this setting, the direct connections between input and latent variables makes the latent variables partially interpretable; furthermore, the connectivity among the latent variables in the neural network serves to model their potential causal relationships to each other and to the output variables. A series of simulation experiments provide support that the RINN method can successfully recover latent causal structure between input and output variables.

cs.LG↗
arXiv2019

Supervised Vector Quantized Variational Autoencoder for Learning Interpretable Global Representations

Learning interpretable representations of data remains a central challenge in deep learning. When training a deep generative model, the observed data are often associated with certain categorical labels, and, in parallel with learning to regenerate data and simulate new data, learning an interpretable representation of each class of data is also a process of acquiring knowledge. Here, we present a novel generative model, referred to as the Supervised Vector Quantized Variational AutoEncoder (S-VQ-VAE), which combines the power of supervised and unsupervised learning to obtain a unique, interpretable global representation for each class of data. Compared with conventional generative models, our model has three key advantages: first, it is an integrative model that can simultaneously learn a feature representation for individual data point and a global representation for each class of data; second, the learning of global representations with embedding codes is guided by supervised information, which clearly defines the interpretation of each code; and third, the global representations capture crucial characteristics of different classes, which reveal similarity and differences of statistical structures underlying different groups of data. We evaluated the utility of S-VQ-VAE on a machine learning benchmark dataset, the MNIST dataset, and on gene expression data from the Library of Integrated Network-Based Cellular Signatures (LINCS). We proved that S-VQ-VAE was able to learn the global genetic characteristics of samples perturbed by the same class of perturbagen (PCL), and further revealed the mechanism correlations between PCLs. Such knowledge is crucial for promoting new drug development for complex diseases like cancer.

cs.LG↗
arXiv2019

From genome to phenome: Predicting multiple cancer phenotypes based on somatic genomic alterations via the genomic impact transformer

Cancers are mainly caused by somatic genomic alterations (SGAs) that perturb cellular signaling systems and eventually activate oncogenic processes. Therefore, understanding the functional impact of SGAs is a fundamental task in cancer biology and precision oncology. Here, we present a deep neural network model with encoder-decoder architecture, referred to as genomic impact transformer (GIT), to infer the functional impact of SGAs on cellular signaling systems through modeling the statistical relationships between SGA events and differentially expressed genes (DEGs) in tumors. The model utilizes a multi-head self-attention mechanism to identify SGAs that likely cause DEGs, or in other words, differentiating potential driver SGAs from passenger ones in a tumor. GIT model learns a vector (gene embedding) as an abstract representation of functional impact for each SGA-affected gene. Given SGAs of a tumor, the model can instantiate the states of the hidden layer, providing an abstract representation (tumor embedding) reflecting characteristics of perturbed molecular/cellular processes in the tumor, which in turn can be used to predict multiple phenotypes. We apply the GIT model to 4,468 tumors profiled by The Cancer Genome Atlas (TCGA) project. The attention mechanism enables the model to better capture the statistical relationship between SGAs and DEGs than conventional methods, and distinguishes cancer drivers from passengers. The learned gene embeddings capture the functional similarity of SGAs perturbing common pathways. The tumor embeddings are shown to be useful for tumor status representation, and phenotype prediction including patient survival time and drug response of cancer cell lines.

q-bio.MN↗
arXiv2019

PubMedQA: A Dataset for Biomedical Research Question Answering

We introduce PubMedQA, a novel biomedical question answering (QA) dataset collected from PubMed abstracts. The task of PubMedQA is to answer research questions with yes/no/maybe (e.g.: Do preoperative statins reduce atrial fibrillation after coronary artery bypass grafting?) using the corresponding abstracts. PubMedQA has 1k expert-annotated, 61.2k unlabeled and 211.3k artificially generated QA instances. Each PubMedQA instance is composed of (1) a question which is either an existing research article title or derived from one, (2) a context which is the corresponding abstract without its conclusion, (3) a long answer, which is the conclusion of the abstract and, presumably, answers the research question, and (4) a yes/no/maybe answer which summarizes the conclusion. PubMedQA is the first QA dataset where reasoning over biomedical research texts, especially their quantitative contents, is required to answer the questions. Our best performing model, multi-phase fine-tuning of BioBERT with long answer bag-of-word statistics as additional supervision, achieves 68.1% accuracy, compared to single human performance of 78.0% accuracy and majority-baseline of 55.2% accuracy, leaving much room for improvement. PubMedQA is publicly available at https://pubmedqa.github.io.

cs.CL↗
arXiv2019

Deep Contextualized Biomedical Abbreviation Expansion

Automatic identification and expansion of ambiguous abbreviations are essential for biomedical natural language processing applications, such as information retrieval and question answering systems. In this paper, we present DEep Contextualized Biomedical. Abbreviation Expansion (DECBAE) model. DECBAE automatically collects substantial and relatively clean annotated contexts for 950 ambiguous abbreviations from PubMed abstracts using a simple heuristic. Then it utilizes BioELMo to extract the contextualized features of words, and feed those features to abbreviation-specific bidirectional LSTMs, where the hidden states of the ambiguous abbreviations are used to assign the exact definitions. Our DECBAE model outperforms other baselines by large margins, achieving average accuracy of 0.961 and macro-F1 of 0.917 on the dataset. It also surpasses human performance for expanding a sample abbreviation, and remains robust in imbalanced, low-resources and clinical settings.

cs.CL↗
arXiv2019

Probing Biomedical Embeddings from Language Models

Contextualized word embeddings derived from pre-trained language models (LMs) show significant improvements on downstream NLP tasks. Pre-training on domain-specific corpora, such as biomedical articles, further improves their performance. In this paper, we conduct probing experiments to determine what additional information is carried intrinsically by the in-domain trained contextualized embeddings. For this we use the pre-trained LMs as fixed feature extractors and restrict the downstream task models to not have additional sequence modeling layers. We compare BERT, ELMo, BioBERT and BioELMo, a biomedical version of ELMo trained on 10M PubMed abstracts. Surprisingly, while fine-tuned BioBERT is better than BioELMo in biomedical NER and NLI tasks, as a fixed feature extractor BioELMo outperforms BioBERT in our probing tasks. We use visualization and nearest neighbor analysis to show that better encoding of entity-type and relational information leads to this superiority.

cs.CL↗
arXiv2018

Optical Stark Effect of a Single Defect on TiO2(110) Surface

Probing optical Stark effect at the single-molecule or atomic scale is crucial for understanding many photo-induced chemical and physical processes on surfaces. Here we report a study about optical Stark effect of single atomic defects on TiO2(110) surface with photo-assisted scanning tunneling spectroscopy. When a laser is coupled into the tunneling junction, the mid-gap state of OH-O2 defects changes remarkably in the differential conductance spectra. As laser power gradually increases, the energy of the mid-gap state shifts away from the Fermi level with increase in intensity and broadening of peak width. The observation can be explained as optical Stark effect with the Autler-Townes formula. This large optical Stark effect is due to the tip-enhancement and the strong dipole moment in the transient charged state during electron tunneling. Our study provides new aspects in exploring electron-photon interactions at the microscopic scale.

cond-mat.mes-hall↗
arXiv2014

An exact algorithm for the weighed mutually exclusive maximum set cover problem

In this paper, we introduce an exact algorithm with a time complexity of $O^*(1.325^m)$ for the {\sc weighted mutually exclusive maximum set cover} problem, where $m$ is the number of subsets in the problem. This is an NP-hard motivated and abstracted from a bioinformatics problem of identifying signaling pathways based gene mutations. Currently, this problem is addressed using heuristic algorithms, which cannot guarantee the performance of the solution. By providing a relatively efficient exact algorithm, our approach will like increase the capability of finding better solutions in the application of cancer research.

cs.DS↗
arXiv2013

Anomalously Strong 2D Band Intensity in Twisted Bilayer Graphene: Raman Evidence for Doubly Degenerate Dirac Band

We report the observation of anomalously strong 2D band in twisted bilayer graphene (tBLG) with large rotation angles under 638-nm and 532-nm visible laser excitation. The 2D band of tBLG can reach four times as opposed to two times as strong as that of single layer graphene. The same tBLG samples also exhibit rotation dependent G-line resonances and folded phonons under 364-nm UV laser excitation. We attribute this 2D band Raman enhancement to the constructive quantum interference between two double-resonance Raman pathways which are enabled by nearly degenerate Dirac band in tBLG Moiré superlattices.

cond-mat.mes-hall↗
arXiv2013

From data towards knowledge: Revealing the architecture of signaling systems by unifying knowledge mining and data mining of systematic perturbation data

Genetic and pharmacological perturbation experiments, such as deleting a gene and monitoring gene expression responses, are powerful tools for studying cellular signal transduction pathways. However, it remains a challenge to automatically derive knowledge of a cellular signaling system at a conceptual level from systematic perturbation-response data. In this study, we explored a framework that unifies knowledge mining and data mining approaches towards the goal. The framework consists of the following automated processes: 1) applying an ontology-driven knowledge mining approach to identify functional modules among the genes responding to a perturbation in order to reveal potential signals affected by the perturbation; 2) applying a graph-based data mining approach to search for perturbations that affect a common signal with respect to a functional module, and 3) revealing the architecture of a signaling system organize signaling units into a hierarchy based on their relationships. Applying this framework to a compendium of yeast perturbation-response data, we have successfully recovered many well-known signal transduction pathways; in addition, our analysis have led to many hypotheses regarding the yeast signal transduction system; finally, our analysis automatically organized perturbed genes as a graph reflecting the architect of the yeast signaling system. Importantly, this framework transformed molecular findings from a gene level to a conceptual level, which readily can be translated into computable knowledge in the form of rules regarding the yeast signaling system, such as "if genes involved in MAPK signaling are perturbed, genes involved in pheromone responses will be differentially expressed".

q-bio.MN↗
arXiv2013

Twisted Bilayer Graphene Superlattices

Twisted bilayer graphene (tBLG) provides us with a large rotational freedom to explore new physics and novel device applications, but many of its basic properties remain unresolved. Here we report the synthesis and systematic Raman study of tBLG. Chemical vapor deposition was used to synthesize hexagon- shaped tBLG with a rotation angle that can be conveniently determined by relative edge misalignment. Superlattice structures are revealed by the observation of two distinctive Raman features: folded optical phonons and enhanced intensity of the 2D-band. Both signatures are strongly correlated with G-line resonance, rotation angle and laser excitation energy. The frequency of folded phonons decreases with the increase of the rotation angle due to increasing size of the reduced Brillouin zone (rBZ) and the zone folding of transverse optic (TO) phonons to the rBZ of superlattices. The anomalous enhancement of 2D-band intensity is ascribed to the constructive quantum interference between two Raman paths enabled by a near-degenerate Dirac cone. The fabrication and Raman identification of superlattices pave the way for further basic study and new applications of tBLG.

cond-mat.mtrl-sci↗
arXiv2009

Phase Separation and Charge Transfer in a K-doped C60 Monolayer on Ag(001)

We have performed a scanning tunneling microscopy and spectroscopy study of potassium-doped C60 monolayers (KxC60) on Ag(001) in the regime of x ~ 1. Low temperature annealing (640 K) leads to the formation of two well-ordered KxC60 phases that exhibit differing levels of electron charge transfer. Further annealing (710 K) distills out the higher electron-doped phase from the lower electron-doped phase, leaving behind a third C60 phase completely devoid of K. Spectroscopic measurements indicate that the electron-doping level of the higher electron-doped KC60 phase is anomalously large.

cond-mat.mtrl-sci↗
arXiv2007

Spatially-resolved electronic and vibronic properties of single diamondoid molecules

Diamondoids are a unique form of carbon nanostructure best described as hydrogen-terminated diamond molecules. Their diamond-cage structures and tetrahedral sp3 hybrid bonding create new possibilities for tuning electronic band gaps, optical properties, thermal transport, and mechanical strength at the nanoscale. The recently-discovered higher diamondoids (each containing more than three diamond cells) have thus generated much excitement in regards to their potential versatility as nanoscale devices. Despite this excitement, however, very little is known about the properties of isolated diamondoids on metal surfaces, a very relevant system for molecular electronics. Here we report the first molecular scale study of individual tetramantane diamondoids on Au(111) using scanning tunneling microscopy and spectroscopy. We find that both the diamondoid electronic structure and electron-vibrational coupling exhibit unique spatial distributions characterized by pronounced line nodes across the molecular surfaces. Ab-initio pseudopotential density functional calculations reveal that the observed dominant electronic and vibronic properties of diamondoids are determined by surface hydrogen terminations, a feature having important implications for designing diamondoid-based molecular devices.

cond-mat.mtrl-sci↗