Biology subjects

Butte, A. J.

Publications and source records attributed to Butte, A. J..

16 recordsLinked to original sources

bioRxiv2025

Benchmarking Large Language Models for Predictive Modeling in Biomedical Research With a Focus on Reproductive Health

Generative AI, particularly large language models (LLMs), is increasingly being used in computational biology to support code generation for data analysis. In this study, we evaluated the ability of LLMs to generate functional R and Python code for predictive modeling tasks, leveraging standardized molecular datasets from several recent DREAM (Dialogue for Reverse Engineering Assessments and Methods) Challenges focused on reproductive health. We assessed LLM performance across four predictive tasks derived from three DREAM challenges: gestational age regression from gene expression, gestational age regression from DNA methylation profiles, and classification of preterm birth and early preterm birth from microbiome data. LLMs were prompted with task descriptions, data locations, and target outcomes. LLM-generated code was then run to fit and apply prediction models and generate graphics, and they were ranked based on their success in completing the tasks and achieving strong test set performance. Among the eight LLMs tested, o3-mini-high, 4o, DeepseekR1 and Gemini 2.0 completed at least one task without error. Overall, R code generation was more successful (14/16 tasks) than Python (7/16), attributed to the utility of Bioconductor packages for querying Gene Expression Omnibus data. OpenAIs o3-mini-high outperformed others, completing 7/8 tasks. Test set performance of the top LLM matched or exceeded top-performing teams from the original DREAM challenges. These findings underscore the potential of LLMs to enhance exploratory analysis and democratize access to predictive modeling in omics by automating key components of analysis pipelines, and highlight the potential to increase research output when conducting analyses of standardized datasets from public repositories.

bioinformatics↗

bioRxiv2024

HIV-SEQ REVEALS GLOBAL HOST GENE EXPRESSION DIFFERENCES BETWEEN HIV-TRANSCRIBING CELLS FROM VIREMIC AND SUPPRESSED PEOPLE WITH HIV

"Active" reservoir cells transcribing HIV can perpetuate chronic inflammation in virally suppressed people with HIV (PWH) and likely contribute to viral rebound after antiretroviral therapy (ART) interruption, so they represent an important target for new therapies. These cells, however, are difficult to study using single-cell RNA-seq (scRNA-seq) due to their low frequency and low levels of HIV transcripts, which are usually not polyadenylated. Here, we developed "HIV-seq" to enable more efficient capture of HIV transcripts - including non-polyadenylated ones - for scRNA-seq analysis of cells from PWH. By spiking in a set of custom-designed capture sequences targeting conserved regions of the HIV genome during scRNA-seq, we increased our ability to find HIV RNA+ cells from PWH by up to 44%. Implementing HIV-seq in conjunction with surface phenotyping by CITE-seq on paired blood specimens from PWH before vs. after ART suppression, we found that HIV RNA+ cells were enriched among T effector memory (Tem) cells during both viremia and ART suppression, but exhibited a cytotoxic signature during viremia only. By contrast, HIV RNA+ cells from the ART-suppressed timepoints exhibited a distinct anti-inflammatory signature involving elevated TGF-{beta} and diminished IFN signaling. Overall, these findings demonstrate that active reservoir cells exhibit transcriptional features distinct from HIV RNA+ cells during viremia, and underscore HIV-seq as a useful tool to better understand the mechanisms by which HIV-transcribing cells can persist during ART.

immunology↗

bioRxiv2024

Immunological Profiling in Knee Osteoarthritis: Treg Dysfunction as Key Driver of Pain

Pain is the hallmark symptom of osteoarthritis (OA) and its biological drivers remain poorly understood. While the role of innate immunity in OA has been extensively studied, the involvement of adaptive immunity, in particular regulatory T cells (Tregs), is not well understood. Using a comprehensive multi-omic approach on the peripheral blood from 46 knee OA patients with similar radiographic stage, including deep immunophenotyping, cytokine profiling, transcriptomic and T-cell receptor analysis on sorted CD4 Tregs and effector T cells (Teff), we identified an immunological signature associated with OA-related pain. Cytokines promoting Treg expansion and activation (with increases of sIL2-RA, sTNFR1, sTNFR2) were correlated with the Western Ontario and McMaster Universities Arthritis Index (WOMAC) pain subscore, suggesting a potential Treg dysfunction. Nineteen T cell subsets were correlated with WOMAC pain. Notably, we found a negative correlation of cell subsets associated with Treg expansion and activation (FoxP3+CTLA4+, CD4+CD57+, Treg CD95+, CD4 Treg CD45RA-). Differential gene expression analysis between patients with low and high WOMAC pain intensity (threshold [≥] 40/100) revealed an upregulation of inflammasome-related genes such as IL1RL1, IL31RA, IFITM3, NLRP3, IFNG in Tregs. Functional enrichment analysis highlighted an overrepresentation of innate immune response, IL-8, and interferon activation pathways suggesting a pro-inflammatory state in Tregs of patients with high pain intensity. Collectively, our systems immunology approach highlights multiple associations between Treg dysfunctionality and OA-related pain, providing new insights into the adaptive immune systems contribution to OA-related pain.

immunology↗

bioRxiv2023

Deciphering Cell-types and Gene Signatures Associated with Disease Activity in Rheumatoid Arthritis using Single Cell RNA-sequencing

ObjectiveSingle cell profiling of synovial tissue has previously identified gene signatures associated with rheumatoid arthritis (RA) pathophysiology, but synovial tissue is difficult to obtain. This study leverages single cell sequencing of peripheral blood mononuclear cells (PBMCs) from patients with RA and matched healthy controls to identify disease relevant cell subsets and cell type specific signatures of disease. MethodsSingle-cell RNA sequencing (scRNAseq) was performed on peripheral blood mononuclear cells (PBMCs) from 18 RA patients and 18 matched controls, accounting for age, gender, race, and ethnicity). Samples were processed using standard CellRanger and Scanpy pipelines, pseudobulk differential gene expression analysis was performed using DESeq2, and cell-cell communication analysis using CellChat. ResultsWe identified 18 distinct PBMC subsets, including a novel IFITM3+ monocyte subset. CD4+ T effector memory cells were increased in patients with moderate to high disease activity (DAS28-CRP [≥] 3.2), while non-classical monocytes were decreased in patients with low disease activity or remission (DAS28-CRP < 3.2). Differential gene expression analysis identified RA-associated genes in IFITM3+ and non-classical monocyte subsets, and downregulation of pro-inflammatory genes in the V{delta} subset. Additionally, we identified gene signatures associated with disease activity, characterized by upregulation of pro-inflammatory genes TNF, JUN, EGR1, IFIT2, MAFB, G0S2, and downregulation of HLA-DQB1, HLA-DRB5, TNFSF13B. Notably, cell-cell communication analysis revealed upregulation of immune-associated signaling pathways, including VISTA, in patients with RA. ConclusionsWe provide a novel single-cell transcriptomics dataset of PBMCs from patients with RA, and identify insights into the systemic cellular and molecular mechanisms underlying RA disease activity.

immunology↗

bioRxiv2021

Deep learning from multiple experts improves identification of amyloid neuropathologies

Pathologists can label pathologies differently, making it challenging to yield consistent assessments in the absence of one ground truth. To address this problem, we present a DL approach that draws on a cohort of experts, weighs each contribution, and is robust to noisy labels. We collected 100,495 annotations on 20,099 candidate amyloid beta neuropathologies (cerebral amyloid angiopathy (CAA), and cored and diffuse plaques) from three institutions, independently annotated by five experts. DL methods trained on a consensus-of-two strategy yielded 12.6-26% improvements by area under the precision recall curve (AUPRC) when compared to those that learned individualized annotations. This strategy surpassed individual-expert models, even when unfairly assessed on benchmarks favoring them. Moreover, ensembling over individual models was robust to hidden random annotators. In blind prospective tests of 52,555 subsequent expert-annotated images, the models labeled pathologies like their human counterparts (consensus model AUPRC=0.74 cored; 0.69 CAA). This study demonstrates a means to combine multiple ground truths into a common-ground DL model that yields consistent diagnoses informed by multiple and potentially variable expert opinion.

pathology↗

bioRxiv2021

Trans-channel fluorescence learning improves high-content screening for Alzheimer's disease therapeutics

In microscopy-based drug screens, fluorescent markers carry critical information on how compounds affect different biological processes. However, practical considerations may hinder the use of certain fluorescent markers. Here, we present a deep learning method for overcoming this limitation. We accurately generated predicted fluorescent signals from other related markers and validated this new machine learning (ML) method on two biologically distinct datasets. We used the ML method to improve the selection of biologically active compounds for Alzheimers disease (AD) from high-content high-throughput screening (HCS). The ML method identified novel compounds that effectively blocked tau aggregation, which would have been missed by traditional screening approaches unguided by ML. The method improved triaging efficiency of compound rankings over conventional rankings by raw image channels. We reproduced this ML pipeline on a biologically independent cancer-based dataset, demonstrating its generalizability. The approach is disease-agnostic and applicable across diverse fluorescence microscopy datasets.

pharmacology and toxicology↗

bioRxiv2020

Cold-associated mammokines preserve adipocyte identity

Sympathetic activation during cold exposure increases adipocyte thermogenesis via expression of mitochondrial protein uncoupling protein 1 (UCP1)1. The propensity of adipocytes to express UCP1 is under a critical influence of the adipose microenvironment and varies among various fat depots2-7. Here we report that cold-induced adipocyte UCP1 expression in female mouse subcutaneous white adipose tissue (scWAT) is regulated by mammary gland ductal epithelial cells in the adipose niche. Single cell RNA-sequencing (scRNA-seq) show that under cold condition glandular alveolar and hormone-sensing luminal epithelium subtypes express transcripts that encode secretory factors involved in regulating adipocyte UCP1 expression. We term mammary duct secretory factors as "mammokines". Using whole-tissue immunofluorescence 3D visualization, we reveal previously undescribed sympathetic nerve-ductal points of contact and show that sympathetic nerve-activated mammary ducts limit adipocyte UCP1 expression via cold-induced mammokine production. Both in vivo and ex vivo ablation of mammary ductal epithelium enhances cold-induced scWAT adipocyte thermogenic gene program. The mammary duct network extends throughout most scWATs in female mice, which under cold exposure show markedly less UCP1 expression, fat oxidation, energy expenditure, and subcutaneous fat mass loss compared to male mice. These results show a previously uncharacterized role of sympathetic nerve-activated glandular epithelium in adipocyte thermogenesis. Overall, our findings suggest an evolutionary role of mammary duct luminal cells in defending glandular adiposity during cold exposure, highlight mammary gland epithelium as a highly active metabolic cell type, and implicate a broader role of mammokines in mammary gland physiology and systemic metabolism.

cell biology↗

bioRxiv2020

A robust and interpretable, end-to-end deep learning model for cytometry data

Cytometry technologies are essential tools for immunology research, providing high-throughput measurements of the immune cells at the single-cell level. Traditional approaches in interpreting and using cytometry measurements include manual or automated gating to identify cell subsets from the cytometry data, providing highly intuitive results but may lead to significant information loss, in that additional details in measured or correlated cell signals might be missed. In this study, we propose and test a deep convolutional neural network for analyzing cytometry data in an end-to-end fashion, allowing a direct association between raw cytometry data and the clinical outcome of interest. Using nine large CyTOF studies from the open-access ImmPort database, we demonstrated that the deep convolutional neural network model can accurately diagnose the latent cytomegalovirus (CMV) in healthy individuals, even when using highly heterogeneous data from different studies. In addition, we developed a permutation-based method for interpreting the deep convolutional neural network model and identified a CD27-CD94+ CD8+ T cell population significantly associated with latent CMV infection. Finally, we provide a tutorial for creating, training and interpreting the tailored deep learning model for cytometry data using Keras and TensorFlow (github.com/hzc363/DeepLearningCyTOF).

bioinformatics↗

bioRxiv2019

EpiSort: Enumeration of cell types using targeted bisulfite sequencing

The cellular composition of tumors is now recognized as an essential phenotype, with implications to diagnostic, progression and therapy efficacy. A tool for accurate profiling of the tumor microenvironment is lacking, as single-cell methods and computational approaches are not applicable or suffer from low accuracy. Here we present EpiSort, a novel strategy based on targeted bisulfite sequencing, which allows the accurate enumeration of 23 cell types and may be applicable to cancer studies.

immunology↗

bioRxiv2018

Tracing diagnosis trajectories over millions of inpatients reveal an unexpected association between schizophrenia and rhabdomyolysis

While it has been technically feasible to create longitudinal representations of individual health at a nationwide scale, the use of these techniques to identify novel disease associations for the risk stratification of patients has had limited success. Here, we created a large-scale US longitudinal disease network of traced readmission patterns (i.e., disease trajectories), merging data from over 10.4 million inpatients from 350 California hospitals through the Healthcare Cost and Utilization Project between 1980 and 2010. We were able to create longitudinal representations of disease progression mapping over 300 common diseases, including the well-known complication of heart failure after acute myocardial infarction. Surprisingly, out of these generated disease trajectories, we discovered an unknown association between schizophrenia, a chronic mental disorder, and rhabdomyolysis, a rare disease of muscle breakdown. It was found that 92 of 3674 patients (2.5%) with schizophrenia were readmitted for rhabdomyolysis (relative risk, 2.21 [1.80-2.71, confidence interval = 0.95] P-value 9.54E-15), which has a general population incidence of 1 in 10,000. We validated this association using independent electronic health records from over 830,000 patients treated over seven years at the University of California, San Francisco (UCSF) medical center. A case review of 29 patients at UCSF who were treated for schizophrenia and who went on to develop rhabdomyolysis demonstrated that the majority of cases (62%) are idiopathic, which suggests a biological connection between these two diseases. Together, these findings demonstrate the power of using public disease registries in combination with electronic medical records to discover novel disease associations.\n\nOne Sentence SummaryBased on the longitudinal health records from millions of California inpatient discharges, we created a temporal network that enabled us to understand statewide patterns of hospital readmissions, which led to the novel finding that hospitalization for schizophrenia is significantly associated with rehospitalization for rhabdomyolysis.

bioinformatics↗

bioRxiv2018

Single-cell RNA-seq reveals profibrotic macrophages in lung fibrosis

Myeloid cells localize to peripheral tissues in a wide range of pathologic contexts. However, appreciation of distinct myeloid subtypes has been limited by the signal averaging inherent to bulk sequencing approaches. Here we applied single-cell RNA sequencing (scRNA-seq) to map cellular heterogeneity in lung fibrosis induced by bleomycin injury in mice. We first developed a computational framework that enables unbiased, granular cell-type annotation of scRNA-seq. This approach identified a macrophage subpopulation that was specific to injured lung and notable for high expression of Cx3cr1+ and MHCII genes. We found that these macrophages, which bear a gene expression profile consistent with monocytic origin, progressively acquire alveolar macrophage identity and localize to sites of fibroblast accumulation. Probing their functional role, in vitro studies showed a trophic effect of these cells on fibroblast activation, and ablation of Cx3cr1-expressing cells suppressed fibrosis in vivo. We also found by gene set analysis and immunofluorescence that markers of these macrophages were upregulated in samples from patients with lung fibrosis compared with healthy controls. Taken together, our results uncover a specific pathologic subgroup of macrophages with markers that could enable their therapeutic targeting for fibrosis.

immunology↗

bioRxiv2018

Tumor cell-adipocyte gap junctions activate lipolysis and are essential for breast tumorigenesis

A pro-tumorigenic role for adipocytes has been identified in breast cancer, and reliance on fatty acid catabolism found in aggressive tumors. The molecular mechanisms by which tumor cells coopt neighboring adipocytes, however, remain elusive. Here, we describe a direct interaction linking tumorigenesis to adjacent adipocytes. We examine breast tumors and their normal adjacent tissue from several patient cohorts, patient-derived xenografts and mouse models, and find that lipolysis and lipolytic signaling are activated in neighboring adipose tissue. We find that functional gap junctions form between breast cancer cells and adipocytes. As a result, cAMP is transferred from breast cancer cells to adipocytes and activates lipolysis in a gap junction-dependent manner. We identify connexin 31 (GJB3), which promotes receptor triple negative breast cancer growth and activation of lipolysis in vivo. Thus, direct tumor cell-adipocyte interaction contributes to tumorigenesis and may serve as a new therapeutic target in breast cancer. One sentence summaryGap junctions between breast cancer cells and adipocytes transfer cAMP and activate lipolysis in the breast tumor microenvironment to support growth.

cancer biology↗

bioRxiv2017

Creating ethnicity-specific reference intervals for lab tests from EHR data

The results of clinical lab tests are an essential component of medical decision-making. To guide interpretation, test results are returned with reference intervals defined by the range in which 95% of values occur in healthy individuals. Clinical laboratories often set their own reference intervals to accommodate local population and instruments variations. This approach is costly and can be biased. We describe a novel data-driven method for using electronic health record data to extract healthy patients information to define reference intervals. We found that the distributions of many clinical lab tests differ among self-identified racial and ethnic groups (SIREs) in healthy patients. Finally, we derived SIRE-specific reference intervals and provide evidence that these intervals have clinical prognostic value. Specifically, we show that for two lab tests, serum creatinine level and hemoglobin A1C, SIRE-specific reference intervals are more predictive for need for dialysis and development type 2 diabetes than existing reference intervals.\n\nOne Sentence SummaryA novel method for defining population-specific reference intervals of common clinical laboratory tests from electronical health records has better prognostic value than existing reference intervals.

bioinformatics↗

bioRxiv2017

The 10,000 Immunomes Project: A resource for human immunology

New immunological assays now enable rich measurements of human immune function, but difficulty attaining enough measurements across sufficiently large and diverse cohorts has hindered describing normal human immune physiology on a large scale. Here we present the 10,000 Immunomes Project (10KIP), a diverse human immunology reference derived from over 44,000 individuals across 242 studies from ImmPort, a publicly available resource of raw immunology study data and protocols. We carefully curated datasets, aggregating subjects from healthy/control arms and harmonizing data across studies. We demonstrate 10KIPs utility by describing variations in serum cytokines and leukocytes by age, race, and sex; defining a baseline cell-cytokine network; and using 10KIP as a common control to describe immunologic changes in pregnancy. Subject-level data is available for interactive visualization and download at http://10kImmunomes.org/. We believe 10KIP can serve as a common control cohort and will accelerate hypothesis generation by clinical and basic immunologists across diverse populations.\n\nOne Sentence SummaryAn open online resource of human immunology data from more than 10,000 normal subjects including interactive data visualization and download enables a new look at immune system differences across age and sex, rapid hypothesis generation, and creation of custom control cohorts.

immunology↗

bioRxiv2017

Meta-analysis of Cytometry Data Reveals Racial Differences in Immune Cells

While meta-analysis has demonstrated increased statistical power and more robust estimations in studies, the application of this commonly accepted methodology to cytometry data has been challenging. Different cytometry studies often involve diverse sets of markers. Moreover, the detected values of the same marker are inconsistent between studies due to different experimental designs and cytometer configurations. As a result, the cell subsets identified by existing auto-gating methods cannot be directly compared across studies. We developed MetaCyto for automated meta-analysis of both flow and mass cytometry (CyTOF) data. By combining clustering methods with a silhouette scanning method, MetaCyto is able to identify commonly labeled cell subsets across studies, thus enabling meta-analysis. Applying MetaCyto across a set of 10 heterogeneous cytometry studies totaling 2926 samples enabled us to identify multiple cell populations exhibiting differences in abundance between White and Asian adults. Software is released to the public through GitHub (github.com/hzc363/MetaCyto).

bioinformatics↗

bioRxiv2017

xCell: Digitally portraying the tissue cellular heterogeneity landscape

Tissues are complex milieu consisting of numerous cell-types. Numerous recent methods attempt to enumerate cell subsets from transcriptomes. However, available method used limited source for training and displayed only partial portrayal of the full cellular landscape. Here we present xCell, a novel gene-signature based method for inferring 64 immune and stroma cell-types. We harmonized 1,822 pure human cell-types transcriptomes from various sources, employed curve fitting approach for linear comparison of cell-types, and introduced a novel spillover compensation technique for separating between cell-types. Using extensive in silico analyses and comparison to cytometry immunophenotyping we show that xCell outperforms other methods: http://xCell.ucsf.edu/.

bioinformatics↗