Automatic extraction of social determinants of health from medical notes of chronic lower back pain patients.

OBJECTIVE: We applied natural language processing and inference methods to extract social determinants of health (SDoH) information from clinical notes of patients with chronic low back pain (cLBP) to enhance future analyses of the associations between SDoH disparities and cLBP outcomes. MATERIALS AND METHODS: Clinical notes for patients with cLBP were annotated for 7 SDoH domains, as well as depression, anxiety, and pain scores, resulting in 626 notes with at least one annotated entity for 364 patients. We used a 2-tier taxonomy with these 10 first-level classes (domains) and 52 second-level classes. We developed and validated named entity recognition (NER) systems based on both rule-based and machine learning approaches and validated an entailment model. RESULTS: Annotators achieved a high interrater agreement (Cohen's kappa of 95.3% at document level). A rule-based system (cTAKES), RoBERTa NER, and a hybrid model (combining rules and logistic regression) achieved performance of F1 = 47.1%, 84.4%, and 80.3%, respectively, for first-level classes. DISCUSSION: While the hybrid model had a lower F1 performance, it matched or outperformed RoBERTa NER model in terms of recall and had lower computational requirements. Applying an untuned RoBERTa entailment model, we detected many challenging wordings missed by NER systems. Still, the entailment model may be sensitive to hypothesis wording. CONCLUSION: This study developed a corpus of annotated clinical notes covering a broad spectrum of SDoH classes. This corpus provides a basis for training machine learning models and serves as a benchmark for predictive models for NER for SDoH and knowledge extraction from clinical texts.

Machine Learning Prediction of Liver Allograft Utilization From Deceased Organ Donors Using the National Donor Management Goals Registry.

Early prediction of whether a liver allograft will be utilized for transplantation may allow better resource deployment during donor management and improve organ allocation. The national donor management goals (DMG) registry contains critical care data collected during donor management. We developed a machine learning model to predict transplantation of a liver graft based on data from the DMG registry. METHODS: Several machine learning classifiers were trained to predict transplantation of a liver graft. We utilized 127 variables available in the DMG dataset. We included data from potential deceased organ donors between April 2012 and January 2019. The outcome was defined as liver recovery for transplantation in the operating room. The prediction was made based on data available 12-18 h after the time of authorization for transplantation. The data were randomly separated into training (60%), validation (20%), and test sets (20%). We compared the performance of our models to the Liver Discard Risk Index. RESULTS: Of 13 629 donors in the dataset, 9255 (68%) livers were recovered and transplanted, 1519 recovered but used for research or discarded, 2855 were not recovered. The optimized gradient boosting machine classifier achieved an area under the curve of the receiver operator characteristic of 0.84 on the test set, outperforming all other classifiers. CONCLUSIONS: This model predicts successful liver recovery for transplantation in the operating room, using data available early during donor management. It performs favorably when compared to existing models. It may provide real-time decision support during organ donor management and transplant logistics.

Automatic Labeling of Special Diagnostic Mammography Views from Images and DICOM Headers.

Applying state-of-the-art machine learning techniques to medical images requires a thorough selection and normalization of input data. One of such steps in digital mammography screening for breast cancer is the labeling and removal of special diagnostic views, in which diagnostic tools or magnification are applied to assist in assessment of suspicious initial findings. As a common task in medical informatics is prediction of disease and its stage, these special diagnostic views, which are only enriched among the cohort of diseased cases, will bias machine learning disease predictions. In order to automate this process, here, we develop a machine learning pipeline that utilizes both DICOM headers and images to predict such views in an automatic manner, allowing for their removal and the generation of unbiased datasets. We achieve AUC of 99.72% in predicting special mammogram views when combining both types of models. Finally, we apply these models to clean up a dataset of about 772,000 images with expected sensitivity of 99.0%. The pipeline presented in this paper can be applied to other datasets to obtain high-quality image sets suitable to train algorithms for disease detection.

Author Correction: Discovery of stimulation-responsive immune enhancers with CRISPR activation.

In this Letter, analysis of steady-state regulatory T (Treg) cell percentages from Il2ra enhancer deletion (EDEL) and wild-type (WT) mice revealed no differences between them (Extended Data Fig. 9d). This analysis included two mice whose genotypes were incorrectly assigned. Even after correction of the genotypes, no significant differences in Treg cell percentages were seen when data across experimental cohorts were averaged (as was done in Extended Data Fig. 9d). However, if we normalize the corrected data to account for variation among experimental cohorts, a subtle decrease in EDEL Treg cell percentages is revealed and, using the corrected and normalized data, we have redrawn Extended Data Fig. 9d in Supplementary Fig. 1. The Supplementary Information to this Amendment contains the corrected and reanalysed Extended Data Fig. 9d. The sentence "This enhancer deletion (EDEL) strain also had no obvious T cell phenotypes at steady state (Extended Data Fig. 9)." should read: "This enhancer deletion (EDEL) strain had a small decrease in the percentage of Treg cells (Extended Data Fig. 9).". This error does not affect any of the main figures in the Letter or the data from mice with the human autoimmune-associated single nucleotide polymorphism (SNP) knocked in or with a 12-base-pair deletion at the site (12DEL). In addition, we stated in the Methods that we observed consistent immunophenotypes of EDEL mice across three founders, but in fact, we observed consistent phenotypes in mice from two founders. This does not change any of our conclusions and the original Letter has not been corrected.

Discovery of stimulation-responsive immune enhancers with CRISPR activation.

The majority of genetic variants associated with common human diseases map to enhancers, non-coding elements that shape cell-type-specific transcriptional programs and responses to extracellular cues. Systematic mapping of functional enhancers and their biological contexts is required to understand the mechanisms by which variation in non-coding genetic sequences contributes to disease. Functional enhancers can be mapped by genomic sequence disruption, but this approach is limited to the subset of enhancers that are necessary in the particular cellular context being studied. We hypothesized that recruitment of a strong transcriptional activator to an enhancer would be sufficient to drive target gene expression, even if that enhancer was not currently active in the assayed cells. Here we describe a discovery platform that can identify stimulus-responsive enhancers for a target gene independent of stimulus exposure. We used tiled CRISPR activation (CRISPRa) to synthetically recruit a transcriptional activator to sites across large genomic regions (more than 100 kilobases) surrounding two key autoimmunity risk loci, CD69 and IL2RA. We identified several CRISPRa-responsive elements with chromatin features of stimulus-responsive enhancers, including an IL2RA enhancer that harbours an autoimmunity risk variant. Using engineered mouse models, we found that sequence perturbation of the disease-associated Il2ra enhancer did not entirely block Il2ra expression, but rather delayed the timing of gene activation in response to specific extracellular signals. Enhancer deletion skewed polarization of naive T cells towards a pro-inflammatory T helper (TH17) cell state and away from a regulatory T cell state. This integrated approach identifies functional enhancers and reveals how non-coding variation associated with human immune dysfunction alters context-specific gene programs.

Receptor-like cytoplasmic kinase MARIS functions downstream of CrRLK1L-dependent signaling during tip growth.

Growing plant cells need to rigorously coordinate external signals with internal processes. For instance, the maintenance of cell wall (CW) integrity requires the coordination of CW sensing with CW remodeling and biosynthesis to avoid growth arrest or integrity loss. Despite the involvement of receptor-like kinases (RLKs) of the Catharanthus roseus RLK1-like (CrRLK1L) subfamily and the reactive oxygen species-producing NADPH oxidases, it remains largely unknown how this coordination is achieved. ANXUR1 (ANX1) and ANX2, two redundant members of the CrRLK1L subfamily, are required for tip growth of the pollen tube (PT), and their closest homolog, FERONIA, controls root-hair tip growth. Previously, we showed that ANX1 overexpression mildly inhibits PT growth by oversecretion of CW material, whereas pollen tubes of anx1 anx2 double mutants burst spontaneously after germination. Here, we report the identification of suppressor mutants with improved fertility caused by the rescue of anx1 anx2 pollen tube bursting. Mapping of one these mutants revealed an R240C nonsynonymous substitution in the activation loop of a receptor-like cytoplasmic kinase (RLCK), which we named MARIS (MRI). We show that MRI is a plasma membrane-localized member of the RLCK-VIII subfamily and is preferentially expressed in both PTs and root hairs. Interestingly, mri-knockout mutants display spontaneous PT and root-hair bursting. Moreover, expression of the MRI(R240C) mutant, but not its wild-type form, partially rescues the bursting phenotypes of anx1 anx2 PTs and fer root hairs but strongly inhibits wild-type tip growth. Thus, our findings identify a novel positive component of the CrRLK1L-dependent signaling cascade that coordinates CW integrity and tip growth.

A calcium dialog mediated by the FERONIA signal transduction pathway controls plant sperm delivery.

Sperm delivery for double fertilization of flowering plants relies on interactions between the pollen tube (PT) and two synergids, leading to programmed cell death (PCD) of the PT and one synergid. The mechanisms underlying the communication among these cells during PT reception is unknown. We discovered that the synergids control this process by coordinating their distinct calcium signatures in response to the calcium dynamics and growth behavior of the PT. Induced and spontaneous aberrant calcium responses in the synergids abolish the two coordinated PCD events. Components of the FERONIA (FER) signaling pathway are required for initiating and modulating these calcium responses and for coupling the PCD events. Intriguingly, the calcium signatures are interchangeable between the two synergids, implying that their fates of death and survival are determined by reversible interactions with the PT. Thus, complex intercellular interactions involving a receptor kinase pathway and calcium-mediated signaling control sperm delivery in plants.

Patterning of the angiosperm female gametophyte through the prism of theoretical paradigms.

The FG (female gametophyte) of flowering plants (angiosperms) is a simple highly polar structure composed of only a few cell types. The FG develops from a single cell through mitotic divisions to generate, depending on the species, four to 16 nuclei in a syncytium. These nuclei are then partitioned into three or four distinct cell types. The mechanisms underlying the specification of the nuclei in the FG has been a focus of research over the last decade. Nevertheless, we are far from understanding the patterning mechanisms that govern cell specification. Although some results were previously interpreted in terms of static positional information, several lines of evidence now show that local interactions are important. In the present article, we revisit the available data on developmental mutants and cell fate markers in the light of theoretical frameworks for biological patterning. We argue that a further dissection of the mechanisms may be impeded by the combinatorial and dynamical nature of developmental cues. However, accounting for these properties of developing systems is necessary to disentangle the diversity of the phenotypic manifestations of the underlying molecular interactions.

ANXUR receptor-like kinases coordinate cell wall integrity with growth at the pollen tube tip via NADPH oxidases.

It has become increasingly apparent that the extracellular matrix (ECM), which in plants corresponds to the cell wall, can influence intracellular activities in ways that go far beyond their supposedly passive mechanical support. In plants, growing cells use mechanisms sensing cell wall integrity to coordinate cell wall performance with the internal growth machinery to avoid growth cessation or loss of integrity. How this coordination precisely works is unknown. Previously, we reported that in the tip-growing pollen tube the ANXUR receptor-like kinases (RLKs) of the CrRLK1L subfamily are essential to sustain growth without loss of cell wall integrity in Arabidopsis. Here, we show that over-expression of the ANXUR RLKs inhibits growth by over-activating exocytosis and the over-accumulation of secreted cell wall material. Moreover, the characterization of mutations in two partially redundant pollen-expressed NADPH oxidases coupled with genetic interaction studies demonstrate that the ANXUR RLKs function upstream of these NADPH oxidases. Using the H2O2-sensitive HyPer and the Ca²âº-sensitive YC3.60 sensors in NADPH oxidase-deficient mutants, we reveal that NADPH oxidases generate tip-localized, pulsating H2O2 production that functions, possibly through Ca²âº channel activation, to maintain a steady tip-focused Ca²âº gradient during growth. Our findings support a model where ECM-sensing receptors regulate reactive oxygen species production, Ca²âº homeostasis, and exocytosis to coordinate ECM-performance with the internal growth machinery.

Theoretical and experimental evidence indicates that there is no detectable auxin gradient in the angiosperm female gametophyte.

The plant life cycle alternates between a diploid sporophytic and a haploid gametophytic generation. The female gametophyte (FG) of flowering plants is typically formed through three syncytial mitoses, followed by cellularisation that forms seven cells belonging to four cell types. The specification of cell fates in the FG has been suggested to depend on positional information provided by an intrinsic auxin concentration gradient. The goal of this study was to develop mathematical models that explain the formation of this gradient in a syncytium. Two factors were proposed to contribute to the maintenance of the auxin gradient in Arabidopsis FGs: polar influx at early stages and localised auxin synthesis at later stages. However, no gradient could be generated using classical, one-dimensional theoretical models under these assumptions. Thus, we tested other hypotheses, including spatial confinement by the large central vacuole, background efflux and localised degradation, and investigated the robustness of cell specification under different parameters and assumptions. None of the models led to the generation of an auxin gradient that was steep enough to allow sufficiently robust patterning. This led us to re-examine the response to an auxin gradient in developing FGs using various auxin reporters, including a novel degron-based reporter system. In agreement with the predictions of our models, auxin responses were not detectable within the FG of Arabidopsis or maize, suggesting that the effects of manipulating auxin production and response on cell fate determination might be indirect.