ABSTRACT
Accurate screening of COVID-19 infection status for symptomatic patients is a critical public health task. Although molecular and antigen tests now exist for COVID-19, in resource-limited settings, screening tests are often not available. Furthermore, during the early stages of the pandemic tests were not available in any capacity. We utilized an automated machine learning (ML) approach to train and evaluate thousands of models on a clinical dataset consisting of commonly available clinical and laboratory data, along with cytokine profiles for patients (n = 150). These models were then further tested for generalizability on an out-of-sample secondary dataset (n = 120). We were able to develop a ML model for rapid and reliable screening of patients as COVID-19 positive or negative using three approaches: commonly available clinical and laboratory data, a cytokine profile, and a combination of the common data and cytokine profile. Of the tens of thousands of models automatically tested for the three approaches, all three approaches demonstrated > 92% sensitivity and > 88 specificity while our highest performing model achieved 95.6% sensitivity and 98.1% specificity. These models represent a potential effective deployable solution for COVID-19 status classification for symptomatic patients in resource-limited settings and provide proof-of-concept for rapid development of screening tools for novel emerging infectious diseases.
Subject(s)
COVID-19 , Cytokines , Machine Learning , Humans , COVID-19/diagnosis , Cytokines/blood , SARS-CoV-2/isolation & purification , SARS-CoV-2/immunology , Mass Screening/methods , Male , Female , Sensitivity and Specificity , Middle Aged , Adult , AgedABSTRACT
High-quality medical data is critical to the development and implementation of machine learning (ML) algorithms in healthcare; however, security, and privacy concerns continue to limit access. We sought to determine the utility of "synthetic data" in training ML algorithms for the detection of tuberculosis (TB) from inflammatory biomarker profiles. A retrospective dataset (A) comprised of 278 patients was used to generate synthetic datasets (B, C, and D) for training models prior to secondary validation on a generalization dataset. ML models trained and validated on the Dataset A (real) demonstrated an accuracy of 90%, a sensitivity of 89% (95% CI, 83-94%), and a specificity of 100% (95% CI, 81-100%). Models trained using the optimal synthetic dataset B showed an accuracy of 91%, a sensitivity of 93% (95% CI, 87-96%), and a specificity of 77% (95% CI, 50-93%). Synthetic datasets C and D displayed diminished performance measures (respective accuracies of 71% and 54%). This pilot study highlights the promise of synthetic data as an expedited means for ML algorithm development.
Subject(s)
Carcinoma, Squamous Cell , Melanoma, Amelanotic , Melanoma , Skin Neoplasms , Carcinoma, Squamous Cell/diagnosis , Carcinoma, Squamous Cell/pathology , Humans , Melanoma/diagnosis , Melanoma/pathology , Melanoma, Amelanotic/diagnosis , Melanoma, Amelanotic/pathology , Skin Neoplasms/diagnosis , Skin Neoplasms/pathologyABSTRACT
Serological diagnosis of active tuberculosis (TB) is enhanced by detection of multiple antibodies due to variable immune responses among patients. Clinical interpretation of these complex datasets requires development of suitable algorithms, a time consuming and tedious undertaking addressed by the automated machine learning platform MILO (Machine Intelligence Learning Optimizer). MILO seamlessly integrates data processing, feature selection, model training, and model validation to simultaneously generate and evaluate thousands of models. These models were then further tested for generalizability on out-of-sample secondary and tertiary datasets. Out of 31 antigens evaluated, a 23-antigen model was the most robust on both the secondary dataset (TB vs healthy) and the tertiary dataset (TB vs COPD) with sensitivity of 90.5% and respective specificities of 100.0% and 74.6%. MILO represents a user-friendly, end-to-end solution for automated generation and deployment of optimized models, ideal for applications where rapid clinical implementation is critical such as emerging infectious diseases.
Subject(s)
Machine Learning , Models, Theoretical , Tuberculosis/epidemiology , Adult , Female , Humans , Male , Retrospective Studies , Young AdultABSTRACT
Artificial Intelligence (AI) and machine learning (ML) have now spawned a new field within health care and health science research. These new predictive analytics tools are starting to change various facets of our clinical care domains including the practice of laboratory medicine. Many of these ML tools and studies are also starting to populate our literature landscape as we know it but unfamiliarity of the average reader to the basic knowledge and critical concepts within AI/ML is now demanding a need to better prepare our audience to such relatively unfamiliar concepts. A fundamental knowledge of such platforms will inevitably enhance cross-disciplinary literacy and ultimately lead to enhanced integration and understanding of such tools within our discipline. In this review, we provide a general outline of AI/ML along with an overview of the fundamental concepts of ML categories, specifically supervised, unsupervised, and reinforcement learning. Additionally, since the vast majority of our current approaches within ML in laboratory medicine and health care involve supervised algorithms, we will predominantly concentrate on such platforms. Finally, the need for making such tools more accessible to the average investigator is becoming a major driving force for the need of automation within these ML platforms. This has now given rise to the automated ML (Auto-ML) world which will undoubtedly help shape the future of ML within health care. Hence, an overview of Auto-ML is also covered within this manuscript which will hopefully enrich the reader's understanding, appreciation, and the need for embracing such tools.
Subject(s)
Delivery of Health Care/methods , Machine Learning , Medical Laboratory Science/methods , Algorithms , Artificial Intelligence , Automation , Research Design , Supervised Machine Learning , WorkflowABSTRACT
Modulation of Wnt signaling has untapped potential in regenerative medicine due to its essential functions in stem cell homeostasis. However, Wnt lipidation and Wnt-Frizzled (Fzd) cross-reactivity have hindered translational Wnt applications. Here, we designed and engineered water-soluble, Fzd subtype-specific "next-generation surrogate" (NGS) Wnts that hetero-dimerize Fzd and Lrp6. NGS Wnt supports long-term expansion of multiple different types of organoids, including kidney, colon, hepatocyte, ovarian, and breast. NGS Wnts are superior to Wnt3a conditioned media in organoid expansion and single-cell organoid outgrowth. Administration of Fzd subtype-specific NGS Wnt in vivo reveals that adult intestinal crypt proliferation can be promoted by agonism of Fzd5 and/or Fzd8 receptors, while a broad spectrum of Fzd receptors can induce liver zonation. Thus, NGS Wnts offer a unified organoid expansion protocol and a laboratory "tool kit" for dissecting the functions of Fzd subtypes in stem cell biology.
Subject(s)
Frizzled Receptors , Organoids , Hepatocytes , Stem Cells , Wnt Signaling PathwayABSTRACT
Context: Point-of-care glucose meters are an integral part in the assessment of patients with altered mental status. For this reason, glucose meters are checked for interference from commonly encountered substances, including acetaminophen. The Nova StatStrip® glucose meter has previously been reported to be resistant to interference. We report a case of a very high acetaminophen concentration causing interference with this point-of-care glucose meter.Case report: A 25-year-old female presented after an overdose of acetaminophen and diphenhydramine combination product. Patient was minimally responsive, so a point-of-care glucose check was attempted using the Nova StatStrip® glucose meter. Five different meters were attempted, and each showed an error message. Laboratory analysis using Beckman Coulter® Unicel DxC 800 revealed a glucose of 180 mg/dL and an acetaminophen concentration of 465 mg/L. Serum spiked with acetaminophen at different concentrations revealed interference with the Nova StatStrip® glucose meter at a concentration of 399 mg/L and above. To our knowledge, this interference with the Nova StatStrip® glucose meter has not been reported in the medical literature.Conclusion: Very high levels of acetaminophen can interfere with point-of-care glucose meters, even those that have previously been reported to be robust such as the Nova StatStrip® glucose meter. Clinicians should be aware of this possible interference when treating patients with acetaminophen overdose.
Subject(s)
Acetaminophen/blood , Acetaminophen/poisoning , Blood Glucose/analysis , Drug Overdose/blood , Point-of-Care Systems , Reagent Strips , Adult , Female , HumansABSTRACT
To discriminate between closely related members of a protein family that differ at a limited number of spatially distant positions is a challenge for drug discovery. We describe a combined computational design and experimental selection approach for generating binders targeting functional sites with large, shape complementary interfaces to read out subtle sequence differences for subtype-specific antagonism. Repeat proteins are computationally docked against a functionally relevant region of the target protein surface that varies in the different subtypes, and the interface sequences are optimized for affinity and specificity first computationally and then experimentally. We used this approach to generate a series of human Frizzled (Fz) subtype-selective antagonists with extensive shape complementary interaction surfaces considerably larger than those of repeat proteins selected from random libraries. In vivo administration revealed that Wnt-dependent pericentral liver gene expression involves multiple Fz subtypes, while maintenance of the intestinal crypt stem cell compartment involves only a limited subset.
Subject(s)
Frizzled Receptors/antagonists & inhibitors , Frizzled Receptors/metabolism , Molecular Docking Simulation , Animals , Ankyrins/chemistry , Ankyrins/metabolism , Cell Line , Crystallography, X-Ray , Drug Discovery , Duodenum/cytology , Duodenum/metabolism , Frizzled Receptors/chemistry , Humans , Mice, Inbred C57BL , Protein Binding , Protein Conformation , Stem Cells/cytology , Stem Cells/metabolismABSTRACT
The polycomb repressive complex 2 (PRC2) histone methyltransferase plays a central role in epigenetic regulation in development and in cancer, and hence to interrogate its role in a specific developmental transition, methods are needed for disrupting function of the complex with high temporal and spatial precision. The catalytic and substrate recognition functions of PRC2 are coupled by binding of the N-terminal helix of the Ezh2 methylase to an extended groove on the EED trimethyl lysine binding subunit. Disrupting PRC2 function can in principle be achieved by blocking this single interaction, but there are few approaches for blocking specific protein-protein interactions in living cells and organisms. Here, we describe the computational design of proteins that bind to the EZH2 interaction site on EED with subnanomolar affinity in vitro and form tight and specific complexes with EED in living cells. Induction of the EED binding proteins abolishes H3K27 methylation in human embryonic stem cells (hESCs) and at all but the earliest stage blocks self-renewal, pinpointing the first critical repressive H3K27me3 marks in development.
Subject(s)
Computer Simulation , Histones/metabolism , Human Embryonic Stem Cells/metabolism , Polycomb Repressive Complex 2/metabolism , Histones/chemistry , Human Embryonic Stem Cells/cytology , Humans , Methylation , Polycomb Repressive Complex 2/chemistryABSTRACT
Wnt proteins modulate cell proliferation and differentiation and the self-renewal of stem cells by inducing ß-catenin-dependent signalling through the Wnt receptor frizzled (FZD) and the co-receptors LRP5 and LRP6 to regulate cell fate decisions and the growth and repair of several tissues. The 19 mammalian Wnt proteins are cross-reactive with the 10 FZD receptors, and this has complicated the attribution of distinct biological functions to specific FZD and Wnt subtype interactions. Furthermore, Wnt proteins are modified post-translationally by palmitoylation, which is essential for their secretion, function and interaction with FZD receptors. As a result of their acylation, Wnt proteins are very hydrophobic and require detergents for purification, which presents major obstacles to the preparation and application of recombinant Wnt proteins. This hydrophobicity has hindered the determination of the molecular mechanisms of Wnt signalling activation and the functional importance of FZD subtypes, and the use of Wnt proteins as therapeutic agents. Here we develop surrogate Wnt agonists, water-soluble FZD-LRP5/LRP6 heterodimerizers, with FZD5/FZD8-specific and broadly FZD-reactive binding domains. Similar to WNT3A, these Wnt agonists elicit a characteristic ß-catenin signalling response in a FZD-selective fashion, enhance the osteogenic lineage commitment of primary mouse and human mesenchymal stem cells, and support the growth of a broad range of primary human organoid cultures. In addition, the surrogates can be systemically expressed and exhibit Wnt activity in vivo in the mouse liver, regulating metabolic liver zonation and promoting hepatocyte proliferation, resulting in hepatomegaly. These surrogates demonstrate that canonical Wnt signalling can be activated by bi-specific ligands that induce receptor heterodimerization. Furthermore, these easily produced, non-lipidated Wnt surrogate agonists facilitate functional studies of Wnt signalling and the exploration of Wnt agonists for translational applications in regenerative medicine.
Subject(s)
Signal Transduction , Wnt Proteins/agonists , Wnt Proteins/metabolism , Wnt Signaling Pathway , beta Catenin/metabolism , Animals , Cell Differentiation , Cell Lineage , Cell Proliferation , Frizzled Receptors/metabolism , HEK293 Cells , Hepatocytes/cytology , Hepatomegaly/metabolism , Hepatomegaly/pathology , Humans , Hydrophobic and Hydrophilic Interactions , Intestines/cytology , Ligands , Liver/metabolism , Liver/pathology , Low Density Lipoprotein Receptor-Related Protein-5/metabolism , Low Density Lipoprotein Receptor-Related Protein-6/metabolism , Mesenchymal Stem Cells/cytology , Mesenchymal Stem Cells/metabolism , Mice , Models, Molecular , Organoids/cytology , Organoids/metabolism , Protein Multimerization , Solubility , Tissue Culture TechniquesABSTRACT
von Willebrand factor (VWF) is a multimeric plasma protein that mediates platelet adhesion to sites of vascular injury. The hemostatic function of VWF depends upon the formation of disulfide-linked multimers, which requires the VWF propeptide (D1D2 domains) and adjacent D'D3 domains. VWF multimer assembly occurs in the trans-Golgi at pH ~ 6.2 but not at pH 7.4, which suggests that protonation of one or more His residues (pK(a) ~6.0) mediates the pH dependence of multimerization. Alignment of 30 vertebrate VWF sequences identified 13 highly conserved His residues in the D1D2D'D3 domains, and His-to-Ala mutagenesis identified His³95 and His46° in the D2 domain as critical for VWF multimerization. Replacement of His³95 with Lys or Arg prevented multimer assembly, suggesting that reversible protonation of this His residue is essential. In contrast, replacement of His46° with Lys or Arg preserved normal multimer assembly, whereas Leu, Met, and Gln did not, indicating that the function of His46° depends primarily upon the presence of a positive charge. These results suggest that pH sensing by evolutionarily conserved His residues facilitates the assembly and packaging of VWF multimers upon arrival in the trans-Golgi.
Subject(s)
Histidine , Phylogeny , Protein Multimerization , von Willebrand Factor/chemistry , von Willebrand Factor/metabolism , Animals , Cell Line , Conserved Sequence , Humans , Hydrogen-Ion Concentration , Molecular Sequence Data , Mucins/chemistry , Mucins/metabolism , Mutagenesis , Protein Structure, Quaternary , Sequence Homology, Amino Acid , trans-Golgi Network/metabolism , von Willebrand Factor/geneticsABSTRACT
The relationship between sequence, structure, and function is examined by comparing nineteen cyclic nucleotide monophosphate binding domains of known structure from six different functional families. Comparisons are made by structure and sequence alignment and through the generation of 3610 homology models. This analysis suggests there are only weak relationships between functional families, sequence, and/or structure. However, we have identified that for cyclic nucleotide monophosphate binding domains privileged template structures occur for homology modeling. The existence of privileged template structures, capable of creating accurate modeling for a broad family of proteins, may lead to improved homology modeling protocols.
Subject(s)
Computational Biology/methods , Cyclic AMP/metabolism , Cyclic GMP/metabolism , Amino Acid Sequence , Binding Sites , Molecular Sequence Data , Molecular Structure , Protein Interaction Domains and Motifs/genetics , Protein Interaction Domains and Motifs/physiology , Sequence Homology, Amino AcidABSTRACT
Von Willebrand factor (VWF) dimerizes through C-terminal CK domains, and VWF dimers assemble into multimers in the Golgi by forming intersubunit disulfide bonds between D3 domains. This unusual oxidoreductase reaction requires the VWF propeptide (domains D1D2), which acts as an endogenous pH-dependent chaperone. The cysteines involved in multimer assembly were characterized by using a VWF construct that encodes the N-terminal D1D2D'D3 domains. Modification with thiol-specific reagents demonstrated that secreted D'D3 monomer contained reduced Cys, whereas D'D3 dimer and propeptide did not. Reduced Cys in the D'D3 monomer were alkylated with N-ethylmaleimide and analyzed by mass spectrometry. All 52 Cys within the D'D3 region were observed, and only Cys(1099) and Cys(1142) were modified by N-ethylmaleimide. When introduced into the D1D2D'D3 construct, the mutation C1099A or C1142A markedly impaired the formation of D'D3 dimers, and the double mutation prevented dimerization. In full-length VWF, the mutations C1099A and C1099A/C1142A prevented multimer assembly; the mutation C1142A allowed the formation of almost exclusively dimers, with few tetramers and no multimers larger than hexamers. Therefore, Cys(1099) and Cys(1142) are essential for the oxidoreductase mechanism of VWF multimerization. Cys(1142) is reported to form a Cys(1142)-Cys(1142) intersubunit bond, suggesting that Cys(1099) also participates in a Cys(1099)-Cys(1099) disulfide bond between D3 domains. This arrangement of intersubunit disulfide bonds implies that the dimeric N-terminal D'D3 domains of VWF subunits align in a parallel orientation within VWF multimers.