Biology subjects

Chung, V.

Publications and source records attributed to Chung, V..

9 recordsLinked to original sources

bioRxiv2026

The SEA-AD DREAM Challenge: Community benchmarking human and AI agent solutions for Alzheimer's disease neuropathology prediction from single-nucleus transcriptomics

Single-nucleus transcriptomic atlases offer an unprecedented opportunity to connect cellular molecular states with Alzheimer's disease (AD) neuropathology, but whether these profiles encode reproducible, predictive information about pathological burden remains unclear. We present the SEA-AD DREAM Challenge, an open, international, model-to-data competition built on the Seattle Alzheimer's Disease Brain Cell Atlas to predict Alzheimer's disease neuropathological severity from single-nucleus RNA-sequencing data. Participants developed containerized models to predict categorical neuropathological staging, including overall Alzheimer's disease neuropathologic change, Braak stage, Thal phase, and CERAD score, as well as quantitative amyloid-{beta} and phospho-tau burden measured by 6E10 and AT8 immunohistochemistry. Across 17 eligible teams from 15 countries, the crowdsourcing framework enabled systematic comparison of diverse computational approaches and surfaced a broad landscape of modeling strategies and candidate predictive features. Top-performing methods achieved near-perfect prediction of categorical staging, with the best submission reaching a quadratic weighted kappa of 1.0 for the Overall AD Neuropathological Change score (ADNC), and competitive prediction of quantitative pathological burden in held-out data, with a best concordance correlation coefficient of 0.48. Post hoc perturbation analyses revealed that top categorical-stage predictions relied heavily on donor-level metadata-driven signals rather than transcriptomic features, whereas quantitative pathology prediction was more robust and supported by transcriptomic and cell-type-associated features with potential biological relevance to AD progression. The challenge also introduced the first AI Agent Track in a DREAM Challenge, providing an early benchmark for autonomous and human-guided agentic model development in single-cell neuroscience. This work demonstrates that single-nucleus transcriptomes encode substantial information about Alzheimer's disease pathology, establishes a reproducible benchmark for molecular neuropathology prediction, and highlights critical principles for designing privacy-preserving, leakage-aware community challenges using deeply phenotyped human brain data.

neuroscience↗

bioRxiv2025

Fluphenazine, an antipsychotic compound, ameliorates Alzheimer's disease by clearing amyloid beta accumulation in

The ageing population worldwide faces an increasing burden of age-related conditions, with Alzheimers disease being a prominent neurodegenerative concern. Drug repurposing, the practice of identifying new therapeutic applications for existing drugs, offers a promising avenue for accelerated intervention. In this study, we utilized the yeast S. cerevisiae to screen a library of 1,760 FDA-approved compounds, both with and without rapamycin, to assess potential synergistic effects on yeast growth. We identified 87 compounds that showed synergistic effect with rapamycin and caused growth defects in yeast. These compounds were further screened for their effects on paralysis in a C. elegans model of Alzheimers disease. We found that three compounds synergistically delayed paralysis in combination with rapamycin. Additionally, four other compounds delayed paralysis when tested at different concentrations. Moreover, we tested fluphenazine, an antipsychotic drug identified in our screen and found that it enhanced the overall health of treated worms. Western blot and X-34 staining confirmed that fluphenazine reduced amyloid-beta accumulation. These results indicate that repurposed drugs have significant potential to accelerate Alzheimers disease drug discovery.

molecular biology↗

bioRxiv2022

A Community Challenge to Predict Clinical Outcomes After Immune Checkpoint Blockade in Non-Small Cell Lung Cancer

PurposePredictive biomarkers of immune checkpoint inhibitors (ICIs) efficacy are currently lacking for non-small cell lung cancer (NSCLC). Here, we describe the results from the Anti-PD-1 Response Prediction DREAM Challenge, a crowdsourced initiative that enabled the assessment of predictive models by using data from two randomized controlled clinical trials (RCTs) of ICIs in first-line metastatic NSCLC. MethodsParticipants developed and trained models using public resources. These were evaluated with data from the CheckMate 026 trial (NCT02041533), according to the model-to-data paradigm to maintain patient confidentiality. The generalizability of the models with the best predictive performance was assessed using data from the CheckMate 227 trial (NCT02477826). Both trials were phase III RCTs with a chemotherapy control arm, which supported the differentiation between predictive and prognostic models. Isolated model containers were evaluated using a bespoke strategy that considered the challenges of handling transcriptome data from clinical trials. ResultsA total of 59 teams participated, with 417 models submitted. Multiple predictive models, as opposed to a prognostic model, were generated for predicting overall survival, progression-free survival, and progressive disease status with ICIs. Variables within the models submitted by participants included tumor mutational burden (TMB), programmed death ligand 1 (PD-L1) expression, and gene-expression-based signatures. The bestperforming models showed improved predictive power over reference variables, including TMB or PD-L1. ConclusionThis DREAM Challenge is the first successful attempt to use protected phase III clinical data for a crowdsourced effort towards generating predictive models for ICIs clinical outcomes and could serve as a blueprint for similar efforts in other tumor types and disease states, setting a benchmark for future studies aiming to identify biomarkers predictive of ICIs efficacy. Context summaryO_ST_ABSKey objectiveC_ST_ABSNot all patients with non-small cell lung cancer (NSCLC) eligible for immune checkpoint inhibitor (ICIs) respond to treatment, but accurate predictive biomarkers of ICIs clinical outcomes are currently lacking. This crowdsourced initiative enabled the robust assessment of predictive models using data from two randomized clinical trials of first-line ICI in metastatic NSCLC. Knowledge generatedModels submitted indicate that a combination of programmed death ligand 1 (PD-L1), tumor mutational burden (TMB), and immune gene signatures might be able to identify patients more likely to respond to ICIs. TMB and PD-L1 seemed important to predict progression-free survival and overall survival. Mechanisms including apoptosis, T-cell crosstalk, and adaptive immune resistance appeared essential to predict response. Relevance

cancer biology↗

bioRxiv2022

Omics Imagification: Transforming High-throughput Molecular Representation of a Cell into an Image

Different omics profiles, depending on the underlying technology, encompass measurements of several hundred to several thousand molecules in a biological sample or a cell. This study develops upon the concept of "omics imagification" as a process of transforming a vector representing these numerical measurements into an image with a one-to-one relationship with the corresponding sample. The proposed imagification process transforms a high-dimensional vector of molecular measurements into a two-dimensional RGB image to enable holistic molecular representation of a biological sample and to improve the classification of different biological phenotypes using automated image recognition methods in computer vision. A transformed image represents 2D coordinates of molecules in a neighbour-embedded space representing molecular abundance and gene intensity. The proposed method was applied to a single-cell RNA sequencing (scRNA-seq) data to "imagify" gene expression profiles of individual cells. Our results show that a simple convolutional neural network trained on single-cell transcriptomics images accurately classifies diverse cell types outperforming the best-performing scRNA-seq classifiers such as support vector machine and random forest.

bioinformatics↗

bioRxiv2021

Cell-to-cell and type-to-type heterogeneity of signaling networks: Insights from the crowd

Recent technological developments allow us to measure the status of dozens of proteins in individual cells. This opens the way to understand the heterogeneity of complex multi-signaling networks across cells and cell-types, with important implications to understand and treat diseases such as cancer. These technologies are however limited to proteins for which antibodies are available and are fairly costly, making predictions of new markers and of existing markers under new conditions a valuable alternative. To assess our capacity to make such predictions and boost further methodological development, we organised the Single Cell Signaling in Breast Cancer DREAM challenge. We used a mass cytometry data set, covering 36 markers in over 4,000 conditions totalling 80 million single cells across 67 breast cancer cell lines. Through four increasingly difficult subchallenges, the participants predicted missing markers, new conditions, and the time course response of single cells to stimuli in the presence and absence of kinase inhibitors. The challenge results show that despite the stochastic nature of signal transduction in single cells, the signaling events are tightly controlled and machine learning methods can accurately predict new experimental data. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=58 SRC="FIGDIR/small/436603v2_ufig1.gif" ALT="Figure 1"> View larger version (24K): org.highwire.dtl.DTLVardef@a95337org.highwire.dtl.DTLVardef@966d5aorg.highwire.dtl.DTLVardef@1e55b18org.highwire.dtl.DTLVardef@bfb5d7_HPS_FORMAT_FIGEXP M_FIG C_FIG Key pointsO_LIOver 80 million single-cell multiplexed measurements across 67 cell lines, 54 conditions and 10 time points to benchmark predictive models of single cell signaling C_LIO_LI73 approaches from 27 teams for predicting response to kinase inhibitors on single cell level, and dynamic response from unperturbed basal omics data C_LIO_LIPredictions of single marker models correlate with measurements with a correlation coefficient of 0.76 C_LIO_LITop models of whole signaling response models perform almost as well as a biological replicate C_LIO_LICell-line specific variation in dynamics can be predicted from basal omics C_LI

bioinformatics↗

bioRxiv2020

A Community Challenge for Pancancer Drug Mechanism of Action Inference from Perturbational Profile Data

The Columbia Cancer Target Discovery and Development (CTD2) Center has developed PANACEA (PANcancer Analysis of Chemical Entity Activity), a collection of dose-response curves and perturbational profiles for 400 clinical oncology drugs in cell lines selected to optimally represent 19 cancer subtypes. This resource, developed to study tumor-specific drug mechanism of action, was instrumental in hosting a DREAM Challenge to assess computational models for de novo drug polypharmacology prediction. Dose-response and perturbational profiles for 32 kinase inhibitors were provided to 21 participating teams, who did not know the identity or nature of the compounds, and they were asked to predict high-affinity binding among ~1,300 possible protein targets. Best performing methods leveraged both gene expression profile similarity analysis, and deep-learning methodologies trained on individual datasets. This study lays the foundation for future integrative analyses of pharmacogenomic data, reconciliation of polypharmacology effects in different tumor contexts, and insights into network-based assessment of context-specific drug mechanism of action.

cancer biology↗

bioRxiv2020

BioLitMine: advanced mining of biomedical and biological literature about human genes and genes from major model organisms

The accumulation of biological and biomedical literature outpaces the ability of most researchers and clinicians to stay abreast of their own immediate fields, let alone a broader range of topics. Although available search tools support identification of relevant literature, finding relevant and key publications is not always straightforward. For example, important publications might be missed in searches with an official gene name due to gene synonyms. Moreover, ambiguity of gene names can result in retrieval of a large number of irrelevant publications. To address these issues and help researchers and physicians quickly identify relevant publications, we developed BioLitMine, an advanced literature mining tool that takes advantage of the medical subject heading (MeSH) index and gene-to-publication annotations already available for PubMed literature. Using BioLitMine, a user can identify what MeSH terms are represented in the set of publications associated with a given gene of the interest, or start with a term and identify relevant publications. Users can also use the tool to find co-cited genes and a build a literature co-citation network. In addition, BioLitMine can help users build a gene list relevant to a MeSH terms, such as a list of genes relevant to "stem cells" or "breast neoplasms." Users can also start with a gene or pathway of interest and identify authors associated with that gene or pathway, a feature that makes it easier to identify experts who might serve as collaborators or reviewers. Altogether, BioLitMine extends the value of PubMed-indexed literature and its existing expert curation by providing a robust and gene-centric approach to retrieval of relevant information.

bioinformatics↗

bioRxiv2019

SNP-CRISPR: a web tool for SNP-specific genome editing

CRISPR-Cas9 is a powerful genome editing technology in which a single guide RNA (sgRNA) confers target site specificity to achieve Cas9-mediated genome editing. Numerous sgRNA design tools have been developed based on reference genomes for humans and model organisms. However, existing resources are not optimal as genetic mutations or single nucleotide polymorphisms (SNPs) within the targeting region affect the efficiency of CRISPR-based approaches by interfering with guide-target complementarity. To facilitate identification of sgRNAs (1) in non-reference genomes, (2) across varying genetic backgrounds, or (3) for specific targeting of SNP-containing alleles, for example, disease relevant mutations, we developed a web tool, SNP-CRISPR (https://www.flyrnai.org/tools/snp_crispr/). SNP-CRISPR can be used to design sgRNAs based on public variant data sets or user-identified variants. In addition, the tool computes efficiency and specificity scores for sgRNA designs targeting both the variant and the reference. Moreover, SNP-CRISPR provides the option to upload multiple SNPs and target single or multiple nearby base changes simultaneously with a single sgRNA design. Given these capabilities, SNP-CRISPR has a wide range of potential research applications in model systems and for design of sgRNAs for disease-associated variant correction.

bioinformatics↗

bioRxiv2018

iProteinDB: an integrative database of Drosophila post-translational modifications

Post-translational modification (PTM) serves as a regulatory mechanism for protein function, influencing stability, protein interactions, activity and localization, and is critical in many signaling pathways. The best characterized PTM is phosphorylation, whereby a phosphate is added to an acceptor residue, commonly serine, threonine and tyrosine. As proteins are often phosphorylated at multiple sites, identifying those sites that are important for function is a challenging problem. Considering that many phosphorylation sites may be non-functional, prioritizing evolutionarily conserved phosphosites provides a general strategy to identify the putative functional sites with regards to regulation and function. To facilitate the identification of conserved phosphosites, we generated a large-scale phosphoproteomics dataset from Drosophila embryos collected from six closely-related species. We built iProteinDB (https://www.flyrnai.org/tools/iproteindb/), a resource integrating these data with other high-throughput PTM datasets, including vertebrates, and manually curated information for Drosophila. At iProteinDB, scientists can view the PTM landscape for any Drosophila protein and identify predicted functional phosphosites based on a comparative analysis of data from closely-related Drosophila species. Further, iProteinDB enables comparison of PTM data from Drosophila to that of orthologous proteins from other model organisms, including human, mouse, rat, Xenopus laevis, Danio rerio, and Caenorhabditis elegans.

bioinformatics↗