ABSTRACT
Ikaros transcription factors are essential for adaptive lymphocyte function, yet their role in innate lymphopoiesis is unknown. Using conditional genetic inactivation, we show that Ikzf1/Ikaros is essential for normal natural killer (NK) cell lymphopoiesis and IKZF1 directly represses Cish, a negative regulator of interleukin-15 receptor resulting in impaired interleukin-15 receptor signaling. Both Bcl2l11 and BIM levels, and intrinsic apoptosis were increased in Ikzf1-null NK cells, which in part accounts for NK lymphopenia as both were restored to normal levels when Ikzf1 and Bcl2l11 were co-deleted. Ikzf1-null NK cells presented extensive transcriptional alterations with reduced AP-1 transcriptional complex expression and increased expression of Ikzf2/Helios and Ikzf3/Aiolos. IKZF1 and IKZF3 directly bound AP-1 family members and deletion of both Ikzf1 and Ikzf3 in NK cells resulted in further reductions in Jun/Fos expression and complete loss of peripheral NK cells. Collectively, we show that Ikaros family members are important regulators of apoptosis, cytokine responsiveness and AP-1 transcriptional activity.
Subject(s)
Killer Cells, Natural , Transcription Factor AP-1 , Transcription Factor AP-1/genetics , Killer Cells, Natural/metabolism , Receptors, Interleukin-15 , Ikaros Transcription Factor/genetics , Ikaros Transcription Factor/metabolismABSTRACT
Group 2 innate lymphoid cells (ILC2s) are essential to maintain tissue homeostasis. In cancer, ILC2s can harbor both pro-tumorigenic and anti-tumorigenic functions, but we know little about their underlying mechanisms or whether they could be clinically relevant or targeted to improve patient outcomes. Here, we found that high ILC2 infiltration in human melanoma was associated with a good clinical prognosis. ILC2s are critical producers of the cytokine granulocyte-macrophage colony-stimulating factor, which coordinates the recruitment and activation of eosinophils to enhance antitumor responses. Tumor-infiltrating ILC2s expressed programmed cell death protein-1, which limited their intratumoral accumulation, proliferation and antitumor effector functions. This inhibition could be overcome in vivo by combining interleukin-33-driven ILC2 activation with programmed cell death protein-1 blockade to significantly increase antitumor responses. Together, our results identified ILC2s as a critical immune cell type involved in melanoma immunity and revealed a potential synergistic approach to harness ILC2 function for antitumor immunotherapies.
Subject(s)
Antibodies/pharmacology , Antineoplastic Combined Chemotherapy Protocols/pharmacology , Immune Checkpoint Inhibitors/pharmacology , Interleukin-33/pharmacology , Lymphocytes/drug effects , Melanoma, Experimental/drug therapy , Programmed Cell Death 1 Receptor/antagonists & inhibitors , Skin Neoplasms/drug therapy , Animals , Cell Line, Tumor , Chemotaxis, Leukocyte/drug effects , Cytotoxicity, Immunologic/drug effects , Eosinophils/drug effects , Eosinophils/immunology , Eosinophils/metabolism , Female , Granulocyte-Macrophage Colony-Stimulating Factor/genetics , Granulocyte-Macrophage Colony-Stimulating Factor/metabolism , Humans , Lymphocytes/immunology , Lymphocytes/metabolism , Male , Melanoma, Experimental/genetics , Melanoma, Experimental/immunology , Melanoma, Experimental/metabolism , Mice, Inbred C57BL , Mice, Knockout , Phenotype , Programmed Cell Death 1 Receptor/genetics , Programmed Cell Death 1 Receptor/metabolism , Skin Neoplasms/genetics , Skin Neoplasms/immunology , Skin Neoplasms/metabolismABSTRACT
The last decade has seen rapid progress in the use of genomic tests, including gene panels, whole-exome sequencing, and whole-genome sequencing, in research and clinical cancer care. These advances have created expansive opportunities to characterize the molecular attributes of cancer, revealing a subset of cancer-associated aberrations called driver mutations. The identification of these driver mutations can unearth vulnerabilities of cancer cells to targeted therapeutics, which has led to the development and approval of novel diagnostics and personalized interventions in various malignancies. The applications of this modern approach, often referred to as precision oncology or precision cancer medicine, are already becoming a staple in cancer care and will expand exponentially over the coming years. Although genomic tests can lead to better outcomes by informing cancer risk, prognosis, and therapeutic selection, they remain underutilized in routine cancer care. A contributing factor is a lack of understanding of their clinical utility and the difficulty of results interpretation by the broad oncology community. Practical guidelines on how to interpret and integrate genomic information in the clinical setting, addressed to clinicians without expertise in cancer genomics, are currently limited. Building upon the genomic foundations of cancer and the concept of precision oncology, the authors have developed practical guidance to aid the interpretation of genomic test results that help inform clinical decision making for patients with cancer. They also discuss the challenges that prevent the wider implementation of precision oncology.
Subject(s)
Genetic Testing , Genomics , Neoplasms , Precision Medicine , Humans , Neoplasms/genetics , Neoplasms/therapy , Neoplasms/diagnosis , Precision Medicine/methods , Genomics/methods , Genetic Testing/methods , Practice Guidelines as Topic , Biomarkers, Tumor/genetics , MutationABSTRACT
Highlighting the Distinguished Speakers Symposium on "The Future of Human Genetics and Genomics," this collection of articles is based on presentations at the ASHG 2023 Annual Meeting in Washington, DC, in celebration of all our field has accomplished in the past 75 years, since the founding of ASHG in 1948.
Subject(s)
Breast Neoplasms , Humans , Breast Neoplasms/genetics , Female , Genomics/methodsABSTRACT
To gain a better understanding of the complexity of gene expression in normal and diseased tissues it is important to account for the spatial context and identity of cells in situ. State-of-the-art spatial profiling technologies, such as the Nanostring GeoMx Digital Spatial Profiler (DSP), now allow quantitative spatially resolved measurement of the transcriptome in tissues. However, the bioinformatics pipelines currently used to analyse GeoMx data often fail to successfully account for the technical variability within the data and the complexity of experimental designs, thus limiting the accuracy and reliability of the subsequent analysis. Carefully designed quality control workflows, that include in-depth experiment-specific investigations into technical variation and appropriate adjustment for such variation can address this issue. Here, we present standR, an R/Bioconductor package that enables an end-to-end analysis of GeoMx DSP data. With four case studies from previously published experiments, we demonstrate how the standR workflow can enhance the statistical power of GeoMx DSP data analysis and how the application of standR enables scientists to develop in-depth insights into the biology of interest.
Subject(s)
Gene Expression Profiling , Software , Transcriptome , Computational Biology , Reproducibility of Results , Workflow , Intracellular Space/geneticsABSTRACT
Mass spectrometry (MS) enables high-throughput identification and quantification of proteins in complex biological samples and can provide insights into the global function of biological systems. Label-free quantification is cost-effective and suitable for the analysis of human samples. Despite rapid developments in label-free data acquisition workflows, the number of proteins quantified across samples can be limited by technical and biological variability. This variation can result in missing values which can in turn challenge downstream data analysis tasks. General purpose or gene expression-specific imputation algorithms are widely used to improve data completeness. Here, we propose an imputation algorithm designated for label-free MS data that is aware of the type of missingness affecting data. On published datasets acquired by data-dependent and data-independent acquisition workflows with variable degrees of biological complexity, we demonstrate that the proposed missing value estimation procedure by barycenter computation competes closely with the state-of-the-art imputation algorithms in differential abundance tasks while outperforming them in the accuracy of variance estimates of the peptide abundance measurements, and better controls the false discovery rate in label-free MS experiments. The barycenter estimation procedure is implemented in the msImpute software package and is available from the Bioconductor repository.
Subject(s)
Algorithms , Peptides , Humans , Peptides/analysis , Proteins , Mass Spectrometry/methodsABSTRACT
Gene-set analysis (GSA) dominates the functional interpretation of omics data and downstream hypothesis generation. Despite its ability to summarise thousands of measurements into semantically interpretable components, GSA often results in hundreds of significantly enriched gene-sets. However, summarisation and effective visualisation of GSA results to facilitate hypothesis generation is still lacking. While some webservers provide gene-set visualization tools, there is still a need for tools that can effectively summarize and guide exploration of GSA results. To enable versatility, webservers accept gene lists as input, however, none provide end-to-end solutions for emerging data types such as single-cell and spatial omics. Here, we present vissE.Cloud, a webserver for end-to-end gene-set analysis, offering gene-set summarisation and highly interactive visualisation. vissE.Cloud uses algorithms from our earlier R package vissE to summarise GSA results by identifying biological themes. We maintain versatility by allowing analysis of gene lists, as well as, analysis of raw single-cell and spatial omics data, including CosMx and Xenium data, making vissE.Cloud the first webserver to provide end-to-end gene-set analysis of sub-cellular localised spatial data. Structuring the results hierarchically allows swift interactive investigations of results at the gene, gene-set, and clusters level. vissE.Cloud is freely available at https://www.vissE.Cloud.
Subject(s)
Computational Biology , Data Visualization , Software , Algorithms , Phenotype , Internet , Computational Biology/instrumentation , Computational Biology/methodsABSTRACT
Functional analysis of high throughput experiments using pathway analysis is now ubiquitous. Though powerful, these methods often produce thousands of redundant results owing to knowledgebase redundancies upstream. This scale of results hinders extensive exploration by biologists and can lead to investigator biases due to previous knowledge and expectations. To address this issue, we present vissE, a flexible network-based analysis and visualisation tool that organises information into semantic categories and provides various visualisation modules to characterise them with respect to the underlying data, thus providing a comprehensive view of the biological system. We demonstrate vissE's versatility by applying it to three different technologies: bulk, single-cell and spatial transcriptomics. Applying vissE to a factor analysis of a breast cancer spatial transcriptomic data, we identified stromal phenotypes that support tumour dissemination. Its adaptability allows vissE to enhance all existing gene-set enrichment and pathway analysis workflows, empowering biologists during molecular discovery.
Subject(s)
Breast Neoplasms , Gene Expression Profiling , Humans , Female , Breast Neoplasms/genetics , Breast Neoplasms/pathology , Transcriptome , PhenotypeABSTRACT
BACKGROUND: Through whole-exome sequencing of 60 formalin-fixed paraffin-embedded Nigerian (NGRn) benign prostatic hyperplasia (BPH) samples, we identified germline and somatic alterations in apoptotic pathways impacting BPH development and progression. Prostate enlargement is a common occurrence in male aging; however, this enlargement can lead to lower urinary tract symptoms that negatively impact quality of life. This impact is disproportionately present in men of African ancestry. BPH pathophysiology is poorly understood and studies examining non-European populations are lacking. METHODS: In this study, NGRn BPH, normal prostate, and prostate cancer (PCa) tumor samples were sequenced and compared to characterize genetic alterations in NGRn BPH. RESULTS: Two hundred and two nonbenign, ClinVar-annotated germline variants were present in NGRn BPH samples. Six genes [BRCA1 (92%), HSD3B1 (85%), TP53 (37%), PMS2 (23%), BARD1 (20%), and BRCA2 (17%)] were altered in at least 10% of samples; however, compared to NGRn normal and tumor, the frequency of alterations in BPH samples showed no significant differences at the gene or variant level. BRCA2_rs11571831 and TP53_rs1042522 germline alterations had a statistically significant co-occurrence interaction in BPH samples. In at least two BPH samples, 173 genes harbored somatic variants known to be clinically actionable. Three genes (COL18A1, KIF16B, and LRP1) showed a statistically significant (p < 0.05) higher frequency in BPH. NGRn BPH also had five gene pairs (PKD1/KIAA0100, PKHD1/PKD1, DNAH9/LRP1B, NWD1/DCHS2, and TCERG1/LMTK2) with statistically significant co-occurring interactions. Two hundred and seventy-nine genes contained novel somatic variants in NGRn BPH. Three genes (CABP1, FKBP1C, and RP11-595B24.2) had a statistically significant (p < 0.05) higher alteration frequency in NGRn BPH and three were significantly higher in NGRn tumor (CACNA1A, DMKN, and CACNA2D2). Pairwise Fisher's exact tests showed 14 gene pairs with statistically significant (p < 0.05) interactions and four interactions approaching significance (p < 0.10). Mutational patterns in NGRn BPH were similar to COSMIC (Catalog of Somatic Mutations in Cancer) signatures associated with aging and dysfunctional DNA damage repair. CONCLUSIONS: NGRn BPH contained significant germline alteration interactions (BRCA2_rs11571831 and TP53_rs1042522) and increased somatic alteration frequencies (LMTK2, LRP1, COL18A1, CABP1, and FKBP1C) that impact apoptosis. Normal prostate development is maintained by balancing apoptotic and proliferative activity. Dysfunction in either mechanism can lead to abnormal prostate growth. This work is the first to examine genomic sequencing in NGRn BPH and provides data that fill known gaps in the understanding BPH and how it impacts men of African ancestry.
Subject(s)
Prostatic Hyperplasia , Prostatic Neoplasms , Humans , Male , Prostatic Hyperplasia/genetics , Prostatic Hyperplasia/pathology , Exome Sequencing , Quality of Life , Prostatic Neoplasms/genetics , Prostatic Neoplasms/pathology , Prostate/pathology , Axonemal Dyneins/genetics , Transcriptional Elongation Factors/genetics , Kinesins/geneticsABSTRACT
The significant role of dairy caretakers in maintaining animal welfare on dairy farms emphasizes the necessity of appropriate training and education to ensure the implementation of practices that promote good animal welfare. This study explored the potential of case-based learning as a novel approach to training for dairy caretakers by investigating dairy caretakers' perspectives on case study discussions. Additionally, this study sought to understand thoughts and feelings of caretakers during case study discussions to help identify information that caretakers use to evaluate cases and make decisions. Two case studies were developed and presented to participants, and thematic analysis of case study discussion transcripts was performed. Pre- and post-training questionnaires for 21 caretakers (n = 21) were summarized. The study found that caretaker reactions to case studies were generally positive. Thematic analysis revealed that caretakers use previous knowledge to make treatment decisions for cattle, and valued discussion with coworkers. The results of this study suggest the need for further investigation into the use of case studies and other activities that provide opportunities for critical thinking as training opportunities on dairy farms.
Subject(s)
Animal Welfare , Communication , Humans , Cattle , Animals , Farms , Dairying/methodsABSTRACT
PURPOSE: The use of peer learning methods in radiology continues to grow as a means to constructively learn from past mistakes. This study examined whether emergency radiologists receive a disproportionate amount of peer learning feedback entered as potential learning opportunities (PLO), which could play a significant role in stress and career satisfaction. Our institution offers 24/7 attending coverage, with emergency radiologists interpreting a wide range of X-ray, ultrasound and CT exams on both adults and pediatric patients. MATERIALS AND METHODS: Peer learning submissions entered as PLO at a single large academic medical center over a span of 3 years were assessed by subspecialty distribution and correlated with the number of attending radiologists in each section. Total number of studies performed on emergency department patients and throughout the hospital system were obtained for comparison purposes. Data was assessed using analysis of variance and post hoc analysis. RESULTS: Emergency radiologists received significantly more (2.5 times) PLO submissions than the next closest subspeciality division and received more yearly PLO submissions per attending compared to other subspeciality divisions. This was found to still be true when normalizing for increased case volumes; Emergency radiologists received more PLO submissions per 1000 studies compared to other divisions in our department (1.59 vs. 0.85, p = 0.04). CONCLUSION: Emergency radiologists were found to receive significantly more PLO submissions than their non-emergency colleagues. Presumed causes for this discrepancy may include a higher error rate secondary to wider range of studies interpreted, demand for shorter turn-around times, higher volumes of exams read per shift, and hindsight bias in the setting of follow-up review.
Subject(s)
Radiology , Humans , Child , Radiology/education , Radiologists , Clinical Competence , Academic Medical CentersABSTRACT
PURPOSE: The AAST Organ Injury Scale is widely adopted for splenic injury severity but suffers from only moderate inter-rater agreement. This work assesses SpleenPro, a prototype interactive explainable artificial intelligence/machine learning (AI/ML) diagnostic aid to support AAST grading, for effects on radiologist dwell time, agreement, clinical utility, and user acceptance. METHODS: Two trauma radiology ad hoc expert panelists independently performed timed AAST grading on 76 admission CT studies with blunt splenic injury, first without AI/ML assistance, and after a 2-month washout period and randomization, with AI/ML assistance. To evaluate user acceptance, three versions of the SpleenPro user interface with increasing explainability were presented to four independent expert panelists with four example cases each. A structured interview consisting of Likert scales and free responses was conducted, with specific questions regarding dimensions of diagnostic utility (DU); mental support (MS); effort, workload, and frustration (EWF); trust and reliability (TR); and likelihood of future use (LFU). RESULTS: SpleenPro significantly decreased interpretation times for both raters. Weighted Cohen's kappa increased from 0.53 to 0.70 with AI/ML assistance. During user acceptance interviews, increasing explainability was associated with improvement in Likert scores for MS, EWF, TR, and LFU. Expert panelists indicated the need for a combined early notification and grading functionality, PACS integration, and report autopopulation to improve DU. CONCLUSIONS: SpleenPro was useful for improving objectivity of AAST grading and increasing mental support. Formative user research identified generalizable concepts including the need for a combined detection and grading pipeline and integration with the clinical workflow.
Subject(s)
Tomography, X-Ray Computed , Wounds, Nonpenetrating , Humans , Tomography, X-Ray Computed/methods , Artificial Intelligence , Reproducibility of Results , Machine LearningABSTRACT
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is known to present with pulmonary and extra-pulmonary organ complications. In comparison with the 2009 pandemic (pH1N1), SARS-CoV-2 infection is likely to lead to more severe disease, with multi-organ effects, including cardiovascular disease. SARS-CoV-2 has been associated with acute and long-term cardiovascular disease, but the molecular changes that govern this remain unknown. In this study, we investigated the host transcriptome landscape of cardiac tissues collected at rapid autopsy from seven SARS-CoV-2, two pH1N1, and six control patients using targeted spatial transcriptomics approaches. Although SARS-CoV-2 was not detected in cardiac tissue, host transcriptomics showed upregulation of genes associated with DNA damage and repair, heat shock, and M1-like macrophage infiltration in the cardiac tissues of COVID-19 patients. The DNA damage present in the SARS-CoV-2 patient samples, were further confirmed by γ-H2Ax immunohistochemistry. In comparison, pH1N1 showed upregulation of interferon-stimulated genes, in particular interferon and complement pathways, when compared with COVID-19 patients. These data demonstrate the emergence of distinct transcriptomic profiles in cardiac tissues of SARS-CoV-2 and pH1N1 influenza infection supporting the need for a greater understanding of the effects on extra-pulmonary organs, including the cardiovascular system of COVID-19 patients, to delineate the immunopathobiology of SARS-CoV-2 infection, and long term impact on health.
Subject(s)
COVID-19 , Cardiovascular Diseases , Humans , SARS-CoV-2 , Transcriptome , InterferonsABSTRACT
BACKGROUND: Despite increased utilization of contralateral prophylactic mastectomy (CPM), there is insufficient evidence that it improves survival in average-risk women with unilateral breast cancer. CPM may be of heightened interest to patients with triple negative breast cancer (TNBC) because these patients are more likely to have BRCA1 mutation-associated disease and are not candidates for the chemoprevention benefits of adjuvant endocrine therapy. METHODS: Survival and recurrence outcomes were evaluated for all TNBC patients from a multi-institutional database (1999-2018) at two academic cancer programs in two metropolitan cities of the Northeast and Midwest. Median follow-up time was 3.7 years. RESULTS: Seven hundred and nighty six TNBC patients were evaluated and 15.45% underwent CPM. Women undergoing CPM were more likely to be white (p < 0.001), younger (p < 0.001), and underwent genetic testing (p < 0.001). A borderline survival benefit was observed for TNBC patients undergoing CPM (5-year overall survival 95.1% vs. 85.0%; p = 0.05). There was no difference in survival when BRCA mutation carriers were excluded (5-year overall survival 94.1% vs. 85.2%; p = 0.12). For BRCA mutation carriers, a numeric trend was observed for improved survival for patients undergoing CPM (5-year overall survival 97.2% vs. 84.1%; p = 0.35). Among patients not undergoing CPM, the rate of developing a new primary breast cancer was 2.2% (15/673). Among these 15 patients, 20% (3/15) were known BRCA mutation carriers. CONCLUSIONS: Our data demonstrate no survival benefit for TNBC patients without BRCA1/2 mutations undergoing CPM.
Subject(s)
Breast Neoplasms , Prophylactic Mastectomy , Triple Negative Breast Neoplasms , Humans , Female , Mastectomy , BRCA1 Protein/genetics , Triple Negative Breast Neoplasms/surgery , Breast Neoplasms/genetics , Breast Neoplasms/surgery , Mutation , BRCA2 Protein/geneticsABSTRACT
Artificial intelligence (AI) holds promise for helping patients access new and individualized health care pathways while increasing efficiencies for health care practitioners. Radiology has been at the forefront of this technology in medicine; many radiology practices are implementing and trialing AI-focused products. AI also holds great promise for reducing health disparities and promoting health equity. Radiology is ideally positioned to help reduce disparities given its central and critical role in patient care. The purposes of this article are to discuss the potential benefits and pitfalls of deploying AI algorithms in radiology, specifically highlighting the impact of AI on health equity; to explore ways to mitigate drivers of inequity; and to enhance pathways for creating better health care for all individuals, centering on a practical framework that helps radiologists address health equity during deployment of new tools.
Subject(s)
Health Equity , Radiology , Humans , Artificial Intelligence , Radiologists , Radiology/methods , AlgorithmsABSTRACT
The importance of developing a robust remote workforce in academic radiology has come to the forefront due to several converging factors. COVID-19, and the abrupt transformation it precipitated in terms of how radiologists worked, has been the biggest impetus for change; concurrent factors such as increasing examination volumes and radiologist burnout have also contributed. How to best advance the most desirable and favorable aspects of remote work while preserving an academic environment that fulfills the tripartite mission is a critical challenge that nearly all academic institutions face today. In this article, we discuss current challenges in academic radiology, including effects of the COVID-19 pandemic, from three perspectives-the radiologist, the learner, and the health system-addressing the following topics: productivity, recruitment, wellness, clinical supervision, mentorship and research, educational engagement, radiologist access, investments in technology, and radiologist value. Throughout, we focus on the opportunities and drawbacks of remote work, to help guide its effective and reliable integration into academic radiology practices.
ABSTRACT
Radiology procedure workflow is a summation of individual workflows for scheduling, precertification, preprocedure clinic visits, and day of procedure, representing a complex total process with many opportunities for inefficiencies and waste. At the authors' institution, a lack of standard work and communication gaps in a pre- and postprocedure care area (PPCA) workflow were identified as factors in bottlenecks, waits and delays, and staff and patient frustrations. Using "lean" process improvement tools, these workflows were targeted in a rapid improvement event (RIE). A cross-functional team was formed to work on the PPCA workflow RIE. Using lean management principles, process gaps were identified and changes were instituted to improve patient and information flow. Three projects were implemented over a course of 4 months. These included a 5S, a lean methodology of workplace organization to optimize supply cabinets; standardization of nursing preprocedure documentation and process; and standard work confirmation in daily management system huddles. At baseline, 45% of patients were prepared within 60 minutes of their arrival in the PPCA. After the RIE and instituting the changes from the RIE, 80% of patients were prepared within 60 minutes of their arrival in the PPCA. Implementing lean management strategies, such as daily management systems and huddles, and establishing standard work confirmation help to eliminate waste and create systems and teams that sustain and improve complex workflows. © RSNA, 2022.
Subject(s)
Hospitals , Quality Improvement , Humans , Workflow , Efficiency, OrganizationalABSTRACT
Improving detection and follow-up of recommendations made in radiology reports is a critical unmet need. The long and unstructured nature of radiology reports limits the ability of clinicians to assimilate the full report and identify all the pertinent information for prioritizing the critical cases. We developed an automated NLP pipeline using a transformer-based ClinicalBERT++ model which was fine-tuned on 3 M radiology reports and compared against the traditional BERT model. We validated the models on both internal hold-out ED cases from EUH as well as external cases from Mayo Clinic. We also evaluated the model by combining different sections of the radiology reports. On the internal test set of 3819 reports, the ClinicalBERT++ model achieved 0.96 f1-score while the BERT also achieved the same performance using the reason for exam and impression sections. However, ClinicalBERT++ outperformed BERT on the external test dataset of 2039 reports and achieved the highest performance for classifying critical finding reports (0.81 precision and 0.54 recall). The ClinicalBERT++ model has been successfully applied to large-scale radiology reports from 5 different sites. Automated NLP system that can analyze free-text radiology reports, along with the reason for the exam, to identify critical radiology findings and recommendations could enable automated alert notifications to clinicians about the need for clinical follow-up. The clinical significance of our proposed model is that it could be used as an additional layer of safeguard to clinical practice and reduce the chance of important findings reported in a radiology report is not overlooked by clinicians as well as provide a way to retrospectively track large hospital databases for evaluating the documentation of the critical findings.
Subject(s)
Natural Language Processing , Radiology , Humans , Retrospective Studies , Radiography , Research ReportABSTRACT
MOTIVATION: Protein-protein interactions (PPIs) are critical to normal cellular function and are related to many disease pathways. A range of protein functions are mediated and regulated by protein interactions through post-translational modifications (PTM). However, only 4% of PPIs are annotated with PTMs in biological knowledge databases such as IntAct, mainly performed through manual curation, which is neither time- nor cost-effective. Here we aim to facilitate annotation by extracting PPIs along with their pairwise PTM from the literature by using distantly supervised training data using deep learning to aid human curation. METHOD: We use the IntAct PPI database to create a distant supervised dataset annotated with interacting protein pairs, their corresponding PTM type, and associated abstracts from the PubMed database. We train an ensemble of BioBERT models-dubbed PPI-BioBERT-x10-to improve confidence calibration. We extend the use of ensemble average confidence approach with confidence variation to counteract the effects of class imbalance to extract high confidence predictions. RESULTS AND CONCLUSION: The PPI-BioBERT-x10 model evaluated on the test set resulted in a modest F1-micro 41.3 (P =5 8.1, R = 32.1). However, by combining high confidence and low variation to identify high quality predictions, tuning the predictions for precision, we retained 19% of the test predictions with 100% precision. We evaluated PPI-BioBERT-x10 on 18 million PubMed abstracts and extracted 1.6 million (546507 unique PTM-PPI triplets) PTM-PPI predictions, and filter [Formula: see text] (4584 unique) high confidence predictions. Of the 5700, human evaluation on a small randomly sampled subset shows that the precision drops to 33.7% despite confidence calibration and highlights the challenges of generalisability beyond the test set even with confidence calibration. We circumvent the problem by only including predictions associated with multiple papers, improving the precision to 58.8%. In this work, we highlight the benefits and challenges of deep learning-based text mining in practice, and the need for increased emphasis on confidence calibration to facilitate human curation efforts.
Subject(s)
Data Mining , Protein Processing, Post-Translational , Humans , Proteins , PubMedABSTRACT
Cell extrusion is a morphogenetic process that is implicated in epithelial homeostasis and elicited by stimuli ranging from apoptosis to oncogenic transformation. To explore whether the morphogenetic transcription factor Snail (SNAI1) induces extrusion, we inducibly expressed a stabilized Snail6SA transgene in confluent MCF-7 monolayers. When expressed in small clusters (less than three cells) within otherwise wild-type confluent monolayers, Snail6SA expression induced apical cell extrusion. In contrast, larger clusters or homogenous cultures of Snail6SA cells did not show enhanced apical extrusion, but eventually displayed sporadic basal delamination. Transcriptomic profiling revealed that Snail6SA did not substantively alter the balance of epithelial and mesenchymal genes. However, we identified a transcriptional network that led to upregulated RhoA signalling and cortical contractility in cells expressing Snail6SA Enhanced contractility was necessary, but not sufficient, to drive extrusion, suggesting that Snail collaborates with other factors. Indeed, we found that the transcriptional downregulation of cell-matrix adhesion cooperates with contractility to mediate basal delamination. This provides a pathway for Snail to influence epithelial morphogenesis independently of classic epithelial-to-mesenchymal transition.