Harmful Algae. 2024 Nov;139:102737. doi: 10.1016/j.hal.2024.102737. Epub 2024 Oct 15.

ABSTRACT

Algal blooms impact trophic interactions, community structure and element fluxes. Despite playing an important role in the demise of phytoplankton blooms, only few viruses infecting diatoms have been cultured. Pseudo-nitzschia is a widespread diatom genus that commonly blooms in coastal waters and contains toxin-producing species. This study describes the characterization of a novel virus infecting the toxigenic species Pseudo-nitzschia galaxiae isolated from the northern Adriatic Sea. The ssRNA virus PnGalRNAV has 29.5 nm ± 1.2 nm icosahedral virions and a genome size of 8.8 kb. It belongs to the picorna-like Marnaviridae family and shows high specificity for P. galaxiae infecting two genetically and morphologically distinct strains. We found two genetically distinct types of this virus and screening of the global virome database revealed matching sequences from the Mediterranean region and China, suggesting its global distribution. Another virus of similar shape and size infecting Pseudo-nitzschia calliantha was found, but its genome could not be determined. In addition, we have obtained and characterized a new virus that infects Chaetoceros tenuissimus. The replicase protein of this virus is very similar to the previously described ChTenDNAV type-II virus, but it has a unique genome and infection pattern. Our study is an important contribution to the collective diatom virus culture collection and will allow further investigation into how these viruses control diatom bloom termination, carbon export and toxin release in the case of Pseudo-nitzschia.

PMID:39567073 | DOI:10.1016/j.hal.2024.102737

Int J Biol Macromol. 2024 Nov 18:137668. doi: 10.1016/j.ijbiomac.2024.137668. Online ahead of print.

ABSTRACT

Lysine lactylation (Kla) is a post-translational modification (PTM) that holds significant importance in the regulation of various biological processes. While traditional experimental methods are highly accurate for identifying Kla sites, they are both time-consuming and labor-intensive. Recent machine learning advances have enabled computational models for Kla site prediction. In this study, we propose a novel framework that integrates sequence embedding with sequence descriptors to enhance the representation of protein sequence features. Our framework employs a BiGRU-Transformer architecture to capture both local and global dependencies within the sequence, while incorporating six sequence descriptors to extract biochemical properties and evolutionary patterns. Additionally, we apply a cross-attention fusion mechanism to combine sequence embeddings with descriptor-based features, enabling the model to capture complex interactions between different feature representations. Our model demonstrated excellent performance in predicting Kla sites, achieving an accuracy of 0.998 on the training set and 0.969 on the independent set. Additionally, through attention analysis and motif discovery, our model provided valuable insights into key sequence patterns and regions that are crucial for Kla modification. This work not only deepens the understanding of Kla's functional roles but also holds the potential to positively impact future research in protein modification prediction and functional annotation.

PMID:39566793 | DOI:10.1016/j.ijbiomac.2024.137668

Curr Biol. 2024 Nov 13:S0960-9822(24)01430-1. doi: 10.1016/j.cub.2024.10.041. Online ahead of print.

ABSTRACT

The ability to learn is typically attributed to animals with brains. However, the apparently simplest form of learning, habituation, in which a steadily decreasing response is exhibited to a repeated stimulus, is found not only in animals but also in single-cell organisms and individual mammalian cells. Habituation has been codified from studies in both invertebrate and vertebrate animals as having ten characteristic hallmarks, seven of which involve a single stimulus. Here, we show by mathematical modeling that simple molecular networks, based on plausible biochemistry with common motifs of negative feedback and incoherent feedforward, can robustly exhibit all single-stimulus hallmarks. The models reveal how the hallmarks arise from underlying properties of timescale separation and reversal behavior of memory variables, and they reconcile opposing views of frequency and intensity sensitivity expressed within the neuroscience and cognitive science traditions. Our results suggest that individual cells may exhibit habituation behavior as rich as that which has been codified in multi-cellular animals with central nervous systems and that the relative simplicity of the biomolecular level may enhance our understanding of the mechanisms of learning.

PMID:39566497 | DOI:10.1016/j.cub.2024.10.041

J Bone Miner Res. 2024 Nov 20:zjae191. doi: 10.1093/jbmr/zjae191. Online ahead of print.

ABSTRACT

Achieving bone union remains a significant clinical dilemma. The use of osteoinductive agents, specifically bone morphogenetic proteins (BMPs), has gained wide attention. However, multiple side effects, including increased incidence of cancer, has renewed interest in investigating alternatives that provide safer, yet effective bone regeneration. Here we demonstrate the robust bone healing capabilities of the main megakaryocyte growth factor, thrombopoietin (TPO) and second generation TPO agents using multiple animal models including mice, rats, and pigs. This bone healing activity is shown in two fracture models (critical sized defect [CSD] and closed fracture) and with local or systemic administration. Our transcriptomic analyses, cellular studies, and protein arrays demonstrate that TPO enhances multiple cellular processes important to fracture healing, particularly angiogenesis, which is required for bone union. Finally, the therapeutic potential of thrombopoietic agents is high since they are used in the clinic for other indications (e.g., thrombocytopenia) with established safety profiles and act upon a narrowly defined population of cells.

PMID:39566068 | DOI:10.1093/jbmr/zjae191

Bioinformatics. 2024 Nov 20:btae700. doi: 10.1093/bioinformatics/btae700. Online ahead of print.

ABSTRACT

SUMMARY: Transcript quantification tools efficiently map bulk RNA sequencing reads to reference transcriptomes. However, their output consists of transcript count estimates that are subject to multiple biases and cannot be readily used with existing differential gene expression analysis tools in Python.Here we present pytximport, a Python implementation of the tximport R package that supports a variety of input formats, different modes of bias correction, inferential replicates, gene-level summarization of transcript counts, transcript-level exports, transcript-to-gene mapping generation and optional filtering of transcripts by biotype. pytximport is part of the scverse ecosystem of open-source Python software packages for omics analyses and includes both a Python as well as a command-line interface.With pytximport, we propose a bulk RNA sequencing analysis workflow based on Bioconda and scverse ecosystem packages, ensuring reproducible analyses through Snakemake rules. We apply this pipeline to a publicly available RNA-sequencing dataset, demonstrating how pytximport enables the creation of Python-centric workflows capable of providing insights into transcriptomic alterations.

AVAILABILITY: pytximport is licensed under the GNU General Public License version 3. The source code is available at https://github.com/complextissue/pytximport and via Zenodo with DOI: 10.5281/zenodo.13907917. A related Snakemake workflow is available through GitHub at https://github.com/complextissue/snakemake-bulk-rna-seq-workflow and Zenodo with DOI: 10.5281/zenodo.12713811. Documentation and a vignette for new users are available at: https://pytximport.readthedocs.io.

SUPPLEMENTARY INFORMATION: Supplementary Material is available at Bioinformatics online.

PMID:39565903 | DOI:10.1093/bioinformatics/btae700

Brief Bioinform. 2024 Sep 23;25(6):bbae561. doi: 10.1093/bib/bbae561.

ABSTRACT

Graph databases are becoming increasingly popular across scientific disciplines, being highly suitable for storing and connecting complex heterogeneous data. In systems biology, they are used as a backend solution for biological data repositories, ontologies, networks, pathways, and knowledge graph databases. In this review, we analyse all publications using or mentioning graph databases retrieved from PubMed and PubMed Central full-text search, focusing on the top 16 available graph databases, Publications are categorized according to their domain and application, focusing on pathway and network biology and relevant ontologies and tools. We detail different approaches and highlight the advantages of outstanding resources, such as UniProtKB, Disease Ontology, and Reactome, which provide graph-based solutions. We discuss ongoing efforts of the systems biology community to standardize and harmonize knowledge graph creation and the maintenance of integrated resources. Outlining prospects, including the use of graph databases as a way of communication between biological data repositories, we conclude that efficient design, querying, and maintenance of graph databases will be key for knowledge generation in systems biology and other research fields with heterogeneous data.

PMID:39565895 | DOI:10.1093/bib/bbae561

Sci Transl Med. 2024 Nov 20;16(774):eadq5091. doi: 10.1126/scitranslmed.adq5091. Epub 2024 Nov 20.

ABSTRACT

Macrophages are key drivers of inflammation and tissue damage in autoimmune diseases including rheumatoid arthritis. The rate-limiting step for transcription of more than 70% of inducible genes in macrophages is RNA polymerase II (Pol II) promoter-proximal pause release; however, the specific role of Pol II early elongation control in inflammation, and whether it can be modulated therapeutically, is unknown. Genetic ablation of a pause-stabilizing negative elongation factor (NELF) in macrophages did not affect baseline Pol II occupancy but enhanced the transcriptional response of paused anti-inflammatory genes to lipopolysaccharide followed by secondary attenuation of inflammatory signaling in vitro and in the K/BxN serum transfer mouse model of arthritis. To pharmacologically disrupt the Pol II transcription cycle, we used two covalent inhibitors of the transcription factor II H-associated cyclin-dependent kinase 7 (CDK7), THZ1 and YKL-5-124. Both reduced Pol II pausing in murine and human macrophages, broadly suppressed induction of pro- but not anti-inflammatory genes, and rapidly reversed preestablished inflammatory macrophage polarization. In mice, CDK7 inhibition ameliorated both acute and chronic progressive inflammatory arthritis. Lastly, CDK7 inhibition down-regulated a pathogenic gene expression signature in synovial explants from patients with rheumatoid arthritis. We propose that interfering with Pol II early elongation by targeting CDK7 represents a therapeutic opportunity for rheumatoid arthritis and other inflammatory diseases.

PMID:39565872 | DOI:10.1126/scitranslmed.adq5091

STAR Protoc. 2024 Nov 19;5(4):103457. doi: 10.1016/j.xpro.2024.103457. Online ahead of print.

ABSTRACT

The availability of multi-omics data applied to profile cancer cohorts is rapidly increasing. Here, we present a protocol for Multiomics2Targets, a computational pipeline that can identify driver cell signaling pathways, protein kinases, and cell-surface targets for immunotherapy. We describe steps for preparing the data, uploading files, and tuning parameters. We then detail procedures for running the workflow, visualizing the results, and exporting and sharing reports containing the analysis. For complete details on the use and execution of this protocol, please refer to Deng et al.1.

PMID:39565691 | DOI:10.1016/j.xpro.2024.103457

ACS Synth Biol. 2024 Nov 20. doi: 10.1021/acssynbio.4c00265. Online ahead of print.

ABSTRACT

Repeat expansion disorders, exemplified by myotonic dystrophy type 1 (DM1), present challenges in diagnostic quantification because of the variability and complexity of repeat lengths. Traditional diagnostic methods, including PCR and Southern blotting, exhibit limitations in sensitivity and specificity, necessitating the development of innovative approaches for precise and rapid diagnosis. Here, we introduce a CRISPR-based diagnostic method, REPLICA (repeat-primed locating of inherited disease by Cas3), for the quantification and rapid diagnosis of DM1. This method, using in vitro-assembled CRISPR-Cas3, demonstrates superior sensitivity and specificity in quantifying CTG repeat expansion lengths, correlated with disease severity. We also validate the robustness and accuracy of CRISPR diagnostics in quantitatively diagnosing DM1 using patient genomes. Furthermore, we optimize a REPLICA-based assay for point-of-care-testing using lateral flow test strips, facilitating rapid screening and detection. In summary, REPLICA-based CRISPR diagnostics offer precise and rapid detection of repeat expansion disorders, promising personalized treatment strategies.

PMID:39565688 | DOI:10.1021/acssynbio.4c00265

Nucleic Acids Res. 2024 Nov 20:gkae1024. doi: 10.1093/nar/gkae1024. Online ahead of print.

ABSTRACT

Heterochromatin plays a critical role in regulating gene expression and maintaining genome integrity. While structural and enzymatic components have been linked to heterochromatin establishment, a comprehensive view of the underlying pathways at diverse heterochromatin domains remains elusive. Here, we developed a systematic approach to identify factors involved in heterochromatin silencing at pericentromeres, subtelomeres and the silent mating type locus in Schizosaccharomyces pombe. Using quantitative measures, iterative genetic screening and domain-specific heterochromatin reporters, we identified 369 mutants with different degrees of reduced or enhanced silencing. As expected, mutations in the core heterochromatin machinery globally decreased silencing. However, most other mutants exhibited distinct qualitative and quantitative profiles that indicate heterochromatin domain-specific functions, as seen for example for metabolic pathways affecting primarily subtelomere silencing. Moreover, similar phenotypic profiles revealed shared functions for subunits within complexes. We further discovered that the uncharacterized protein Dhm2 plays a crucial role in heterochromatin maintenance, affecting the inheritance of H3K9 methylation and the clonal propagation of the repressed state. Additionally, Dhm2 loss resulted in delayed S-phase progression and replication stress. Collectively, our systematic approach unveiled a landscape of domain-specific heterochromatin regulators controlling distinct states and identified Dhm2 as a previously unknown factor linked to heterochromatin inheritance and replication fidelity.

PMID:39565189 | DOI:10.1093/nar/gkae1024

mSystems. 2024 Nov 20:e0139524. doi: 10.1128/msystems.01395-24. Online ahead of print.

ABSTRACT

The tumor microbiome, a complex community of microbes found in tumors, has been found to be linked to cancer development, progression, and treatment outcome. However, it remains a bottleneck in distangling the relationship between the tumor microbiome and host gene expressions in tumor microenvironment, as well as their concert effects on patient survival. In this study, we aimed to decode this complex relationship by developing ASD-cancer (autoencoder-based subtypes detector for cancer), a semi-supervised deep learning framework that could extract survival-related features from tumor microbiome and transcriptome data, and identify patients' survival subtypes. By using tissue samples from The Cancer Genome Atlas database, we identified two statistically distinct survival subtypes across all 20 types of cancer Our framework provided improved risk stratification (e.g., for liver hepatocellular carcinoma, [LIHC], log-rank test, P = 8.12E-6) compared to PCA (e.g., for LIHC, log-rank test, P = 0.87), predicted survival subtypes accurately, and identified biomarkers for survival subtypes. Additionally, we identified potential interactions between microbes and host genes that may play roles in survival. For instance, in LIHC, Arcobacter, Methylocella, and Isoptericola may regulate host survival through interactions with host genes enriched in the HIF-1 signaling pathway, indicating these species as potential therapy targets. Further experiments on validation data sets have also supported these patterns. Collectively, ASD-cancer has enabled accurate survival subtyping and biomarker discovery, which could facilitate personalized treatment for broad-spectrum types of cancers.IMPORTANCEUnraveling the intricate relationship between the tumor microbiome, host gene expressions, and their collective impact on cancer outcomes is paramount for advancing personalized treatment strategies. Our study introduces ASD-cancer, a cutting-edge autoencoder-based subtype detector. ASD-cancer decodes the complexities within the tumor microenvironment, successfully identifying distinct survival subtypes across 20 cancer types. Its superior risk stratification, demonstrated by significant improvements over traditional methods like principal component analysis, holds promise for refining patient prognosis. Accurate survival subtype predictions, biomarker discovery, and insights into microbe-host gene interactions elevate ASD-cancer as a powerful tool for advancing precision medicine. These findings not only contribute to a deeper understanding of the tumor microenvironment but also open avenues for personalized interventions across diverse cancer types, underscoring the transformative potential of ASD-cancer in shaping the future of cancer care.

PMID:39565103 | DOI:10.1128/msystems.01395-24

Front Microbiol. 2024 Nov 5;15:1481702. doi: 10.3389/fmicb.2024.1481702. eCollection 2024.

ABSTRACT

A metagenomic time series from Arctic seawater was obtained from Dease Strait, to analyse the changes in bacterioplankton caused by the summer phytoplankton bloom. Bacterial clades specialized in the metabolism of polysaccharides, such as Bacteroidetes, became dominant along the bloom. These specialized taxa quickly displaced the microbial clades that dominate nutrient-poor waters during early spring, such as Archaea, Alpha-and Gammaproteobacteria. At the functional level, phyla Bacteroidetes, Planctomycetes, and Verrucomicrobia showed higher contents of polysaccharide-degradation functions. The Bacteroidetes community shifted toward species with higher polysaccharide-degrading capabilities, targeting algal polysaccharides in summer. Regarding transporters, Bacteroidetes dominated SusC-TonB transporters and had an exclusive family of glycoside-binding proteins (SusD). These proteins were used to identify polysaccharide-utilization loci that clustered transporters and polysaccharide-active enzymes, showing a higher level of specialization toward polysaccharide use. Altogether, these genomic features point to the genetic adaptations that promote the dominance of Bacteroidetes during phytoplankton blooms.

PMID:39564488 | PMC:PMC11573768 | DOI:10.3389/fmicb.2024.1481702

BMC Med Inform Decis Mak. 2024 Nov 19;24(1):346. doi: 10.1186/s12911-024-02763-1.

ABSTRACT

BACKGROUND: Multiscale sample entropy (MSE) is a prevalent complexity metric to characterize a time series and has been extensively applied to the physiological signal analysis. However, for a short-term time series, the likelihood of identifying comparable subsequences decreases, leading to higher variability in the Sample Entropy (SampEn) calculation. Additionally, as the scale factor increases in the MSE calculation, the coarse-graining process further shortens the time series. Consequently, each newly generated time series at a larger scale consists of fewer data points, potentially resulting in unreliable or undefined entropy values, particularly at higher scales. To overcome the shortcoming, a modified multiscale Renyi distribution entropy (MMRDis) was proposed in our present work.

METHODS: The MMRDis method uses a moving-averaging procedure to acquire a family of time series, each of which quantify the dynamic behaviors of the short-term time series over the multiple temporal scales. Then, MMRDis is constructed for the original and the coarse-grained time series.

RESULTS: The MMRDis method demonstrated superior computational stability on simulated Gaussian white and 1/f noise time series, effectively avoiding undefined measurements in short-term time series. Analysis of short-term heart rate variability (HRV) signals from healthy elderly individuals, healthy young people, and subjects with congestive heart failure and atrial fibrillation revealed that MMRDis complexity measurement values decreased with aging and disease. Additionally, MMRDis exhibited better distinction capability for short-term HRV physiological/pathological signals compared to several recently proposed complexity metrics.

CONCLUSIONS: MMRDis was a promising measurement for screening cardiovascular condition within a short time.

PMID:39563351 | DOI:10.1186/s12911-024-02763-1

Sci Rep. 2024 Nov 19;14(1):28630. doi: 10.1038/s41598-024-79724-1.

ABSTRACT

Mammalian milk contains milk-derived extracellular vesicles (MEVs), a group of biological nanovesicles that transport macromolecules. Their ability to cross the blood brain barrier and the presence of cargo capable of modifying gene function have led to the hypothesis that MEVs may play a role in brain function and development. Here, we investigated the uptake of MEVs by human microglia cells in vitro and explored the functional outcomes of MEV uptake. We examined the expression of the miR-148/152 family, highly abundant MEV microRNAs, that directly suppress the translation of DNA methyltransferase (DNMT) enzymes crucial for catalyzing DNA methylation modifications. We also measured phenotypic and inflammatory gene expression in baseline homeostatic and IFN-γ primed microglia to determine if MEVs induce anti-inflammatory effects. We found that MEVs are taken up and localize in baseline and primed microglia. In baseline microglia, MEV supplementation reduced miR-148a-5P levels, increased DNMT1 transcript, protein abundance, and enzymatic activity, compared to cells that did not receive MEVs. In primed microglia, MEV supplementation decreased miR-148a-5P levels and increased DNMT1 protein abundance, but DNMT1 transcript and enzymatic levels remained unchanged. Contrary to predictions, MEV supplementation failed to attenuate pro-inflammatory IL1β expression in primed microglia. This study provides the first evidence of MEV uptake by a brain macrophage, suggesting a potential role in regulating epigenetic machinery and neuroimmune modulation.

PMID:39562680 | DOI:10.1038/s41598-024-79724-1

Nat Commun. 2024 Nov 19;15(1):10022. doi: 10.1038/s41467-024-54209-x.

NO ABSTRACT

PMID:39562561 | DOI:10.1038/s41467-024-54209-x

Am J Hum Genet. 2024 Nov 13:S0002-9297(24)00380-X. doi: 10.1016/j.ajhg.2024.10.011. Online ahead of print.

ABSTRACT

There has been considerable progress in building models to predict the effect of missense substitutions in protein-coding genes, fueled in large part by progress in applying deep learning methods to sequence data. These models have the potential to enable clinical variant annotation on a large scale and hence increase the impact of patient sequencing in guiding diagnosis and treatment. To realize this potential, it is essential to provide reliable assessments of model performance, scope of applicability, and robustness. As a response to this need, the ClinGen Sequence Variant Interpretation Working Group, Pejaver et al., recently proposed a strategy for validation and calibration of in-silico predictions in the context of guidelines for variant annotation. While this work marks an important step forward, the strategy presented still has important limitations. We propose core principles and recommendations to overcome these limitations that can enable both more reliable and more impactful use of variant effect prediction models in the future.

PMID:39561772 | DOI:10.1016/j.ajhg.2024.10.011

Cell Rep Methods. 2024 Nov 18;4(11):100905. doi: 10.1016/j.crmeth.2024.100905.

ABSTRACT

Single-cell RNA sequencing has transformed the study of biological tissues by enabling transcriptomic characterizations of their constituent cell states. Computational methods for gene expression deconvolution use this information to infer the cell composition of related tissues profiled at the bulk level. However, current deconvolution methods are restricted to discrete cell types and have limited power to make inferences about continuous cellular processes such as cell differentiation or immune cell activation. We present ConDecon, a clustering-independent method for inferring the likelihood for each cell in a single-cell dataset to be present in a bulk tissue. ConDecon represents an improvement in phenotypic resolution and functionality with respect to regression-based methods. Using ConDecon, we discover the implication of neurodegenerative microglia inflammatory pathways in the mesenchymal transformation of pediatric ependymoma and characterize their spatial trajectories of activation. The generality of this approach enables the deconvolution of other data modalities, such as bulk ATAC-seq data.

PMID:39561717 | DOI:10.1016/j.crmeth.2024.100905

Cell Rep Methods. 2024 Nov 18;4(11):100906. doi: 10.1016/j.crmeth.2024.100906.

ABSTRACT

We developed a Bayesian-based algorithm to infer gene expression states in individual samples and incorporated it into a workflow to identify tumor-associated antigens (TAAs) across 33 cancer types using RNA sequencing (RNA-seq) data from the Genotype-Tissue Expression (GTEx) and The Cancer Genome Atlas (TCGA). Our analysis identified 212 candidate TAAs, with 78 validated in independent RNA-seq datasets spanning seven cancer types. Eighteen of these TAAs were further corroborated by proteomics data, including 10 linked to liver cancer. We predicted that 38 peptides derived from these 10 TAAs would bind strongly to HLA-A02, the most common HLA allele. Experimental validation confirmed significant binding affinity and immunogenicity for 21 of these peptides. Notably, approximately 64% of liver tumors expressed one or more TAAs associated with these 21 peptides, positioning them as promising candidates for liver cancer therapies, such as peptide vaccines or T cell receptor (TCR)-T cell treatments. This study highlights the power of integrating computational and experimental approaches to discover TAAs for immunotherapy.

PMID:39561714 | DOI:10.1016/j.crmeth.2024.100906

J Neurol Sci. 2024 Nov 10;467:123303. doi: 10.1016/j.jns.2024.123303. Online ahead of print.

ABSTRACT

The study examined changes in the plasma proteome, metabolome, and lipidome of N = 14 patients with relapsing-remitting multiple sclerosis (RRMS) initiating treatment with ocrelizumab, assayed at baseline, 6 months, and 12 months. Analyses of >4000 circulating biomarkers identified depletion of B-cell associated proteins as the early effect observed following ocrelizumab (OCR) initiation, accompanied by the reduction in plasma abundance of cytokines and cytotoxic proteins, markers of neuronaxonal damage, and biologically active lipids including ceramides and lysophospholipids, at 6 months. B-cell depletion was accompanied by decreases in B-cell receptor and cytokine signaling but a pronounced increase in circulating plasma B-cell activating factor (BAFF). This was followed by an upregulation of a number of signaling and metabolic pathways at 12 months. Patients with higher baseline brain MRI lesion load demonstrated both higher levels of cytotoxic and structural proteins in plasma at baseline and more pronounced biomarker change trajectories over time. Digital cytometry identified a putative increase in myeloid cells and a pro-inflammatory subset of T-cells. Therapeutic effects of ocrelizumab extend beyond CD20-mediated B-cell lysis and implicate metabolic reprogramming, juxtaposing the early normalization of immune activation, cytokine signaling and metabolite and lipid turnover in periphery with changes in the dynamics of immune cell activation or composition. We identify BAFF increase following CD20 depletion as a tentative compensatory mechanism that contributes to the reconstitution of targeted B-cells, necessitating further research.

PMID:39561535 | DOI:10.1016/j.jns.2024.123303

J Med Internet Res. 2024 Nov 19;26:e63445. doi: 10.2196/63445.

ABSTRACT

BACKGROUND: Social determinants of health (SDoH) such as housing insecurity are known to be intricately linked to patients' health status. More efficient methods for abstracting structured data on SDoH can help accelerate the inclusion of exposome variables in biomedical research and support health care systems in identifying patients who could benefit from proactive outreach. Large language models (LLMs) developed from Generative Pre-trained Transformers (GPTs) have shown potential for performing complex abstraction tasks on unstructured clinical notes.

OBJECTIVE: Here, we assess the performance of GPTs on identifying temporal aspects of housing insecurity and compare results between both original and deidentified notes.

METHODS: We compared the ability of GPT-3.5 and GPT-4 to identify instances of both current and past housing instability, as well as general housing status, from 25,217 notes from 795 pregnant women. Results were compared with manual abstraction, a named entity recognition model, and regular expressions.

RESULTS: Compared with GPT-3.5 and the named entity recognition model, GPT-4 had the highest performance and had a much higher recall (0.924) than human abstractors (0.702) in identifying patients experiencing current or past housing instability, although precision was lower (0.850) compared with human abstractors (0.971). GPT-4's precision improved slightly (0.936 original, 0.939 deidentified) on deidentified versions of the same notes, while recall dropped (0.781 original, 0.704 deidentified).

CONCLUSIONS: This work demonstrates that while manual abstraction is likely to yield slightly more accurate results overall, LLMs can provide a scalable, cost-effective solution with the advantage of greater recall. This could support semiautomated abstraction, but given the potential risk for harm, human review would be essential before using results for any patient engagement or care decisions. Furthermore, recall was lower when notes were deidentified prior to LLM abstraction.

PMID:39561354 | DOI:10.2196/63445