Search

Patients are complex and heterogeneous; clinical data sets are complicated by noise, missing data, and the presence of mixed-type data. Using such data sets requires understanding the high-dimensional "space of patients", composed of all measurements that define all relevant phenotypes. The current state-of-the-art merely defines spatial groupings of patients using cluster analyses. Our goal is to apply topological data analysis (TDA), a new unsupervised technique, to obtain a more complete understanding of patient space. We applied TDA to a space of 266 previously untreated patients with Chronic Lymphocytic Leukemia (CLL), using the "daisy" metric to compute distances between clinical records. We found clear evidence for both loops and voids in the CLL data. To interpret these structures, we developed novel computational and graphical methods. The most persistent loop and the most persistent void can be explained using three dichotomized, prognostically important factors in CLL: IGHV somatic mutation status, beta-2 microglobulin, and Rai stage. In conclusion, patient space turns out to be richer and more complex than current models suggest. TDA could become a powerful tool in a researcher's arsenal for interpreting high-dimensional data by providing novel insights into biological processes and improving our understanding of clinical and biological data sets.

We present a novel model of time-series analysis to learn from electronic health record (EHR) data when infection occurred in the intensive care unit (ICU) by translating methods from proteomics and Bayesian statistics. Using 48,536 patients hospitalized in an ICU, we describe each hospital course as an 'alphabet' of 23 physician actions ('events') in temporal order. We analyze these as k-mers of length 3-12 events and apply a Bayesian model of (cumulative) relative risk (RR). The log2-transformed RR (median=0.248, mean=0.226) supported the conclusion that the events selected were individually associated with increased risk of infection. Selecting from all possible cutoffs of maximum gain (MG), MG>0.0244 predicts administration of antibiotics with PPV 82.0 %, NPV 44.4 %, and AUC 0.706. Our approach holds value for retrospective analysis of other clinical syndromes for which time-of-onset is critical to analysis but poorly marked in EHRs, including delirium and decompensation., (© 2022 Wiley-VHCA AG, Zurich, Switzerland.)

Background: To examine the potential utility of five multifocal pupillographic objective perimetry (mfPOP) protocols, in the assessment of early diabetic retinopathy (DR) and generalised diabetes-related tissue injury in subjects with type 1 diabetes (T1D)., Methods: Twenty-five T1D subjects (age 41.8 ± 12.1 (SD) years, 13 male) with either no DR (n = 13) or non-proliferative DR (n = 12), and 23 age and gender-matched control subjects (age 39.7 ± 12.9 years, 9 male) were examined by mfPOP using five different stimulus methods differing in visual field eccentricity (central 30° and 60°), and colour (blue, yellow or green test-stimuli presented on, respectively, a blue, yellow or red background), each assessing 44 test-locations per eye. In the T1D subjects, we assessed 16 metabolic status and diabetes complications variables. These were summarised as three principal component analysis (PCA) factors. DR severity was assessed using Early Treatment of Diabetic Retinopathy Study (ETDRS) scores. Area under the curve (AUC) from receiver operator characteristic analyses quantified the diagnostic power of mfPOP response sensitivity and delay deviations for differentiating: (i) T1D subjects from control subjects, (ii) T1D subjects according to three levels of the identified PCA-factors from control subjects, and (iii) TID subjects with from those without non-proliferative DR., Results: The two largest PCA-factors describing the T1D subjects were associated with metabolic variables (e.g. body mass index, HbA1c), and tissue-injury variables (e.g. serum creatinine, vibration perception). Linear models showed that mfPOP per-region response delays were more strongly associated than sensitivities with the metabolic PCA-factor and ETDRS scores. Combined mfPOP amplitude and delay measures produced AUCs of 90.4 ± 8.9% (mean ± SE) for discriminating T1D subjects with DR from control subjects, and T1D subjects with DR from those without of 85.9 ± 8.8%. The yellow and green stimuli performed better than blue on most measures., Conclusions/interpretation: In T1D subjects, mfPOP testing was able to identify localised visual field functional abnormalities (retinal/neural reflex) in the absence or presence of mild DR. mfPOP responses were also associated with T1D metabolic status, but less so with early stages of non-ophthalmic diabetes complications., (© 2022. The Author(s).)

Summary: Unsupervised machine learning provides tools for researchers to uncover latent patterns in large-scale data, based on calculated distances between observations. Methods to visualize high-dimensional data based on these distances can elucidate subtypes and interactions within multi-dimensional and high-throughput data. However, researchers can select from a vast number of distance metrics and visualizations, each with their own strengths and weaknesses. The Mercator R package facilitates selection of a biologically meaningful distance from 10 metrics, together appropriate for binary, categorical and continuous data, and visualization with 5 standard and high-dimensional graphics tools. Mercator provides a user-friendly pipeline for informaticians or biologists to perform unsupervised analyses, from exploratory pattern recognition to production of publication-quality graphics., Availabilityand Implementation: Mercator is freely available at the Comprehensive R Archive Network (https://cran.r-project.org/web/packages/Mercator/index.html)., (© The Author(s) 2021. Published by Oxford University Press.)

Introduction: Clustering analyses in clinical contexts hold promise to improve the understanding of patient phenotype and disease course in chronic and acute clinical medicine. However, work remains to ensure that solutions are rigorous, valid, and reproducible. In this paper, we evaluate best practices for dissimilarity matrix calculation and clustering on mixed-type, clinical data., Methods: We simulate clinical data to represent problems in clinical trials, cohort studies, and EHR data, including single-type datasets (binary, continuous, categorical) and 4 data mixtures. We test 5 single distance metrics (Jaccard, Hamming, Gower, Manhattan, Euclidean) and 3 mixed distance metrics (DAISY, Supersom, and Mercator) with 3 clustering algorithms (hierarchical (HC), k-medoids, self-organizing maps (SOM)). We quantitatively and visually validate by Adjusted Rand Index (ARI) and silhouette width (SW). We applied our best methods to two real-world data sets: (1) 21 features collected on 247 patients with chronic lymphocytic leukemia, and (2) 40 features collected on 6000 patients admitted to an intensive care unit., Results: HC outperformed k-medoids and SOM by ARI across data types. DAISY produced the highest mean ARI for mixed data types for all mixtures except unbalanced mixtures dominated by continuous data. Compared to other methods, DAISY with HC uncovered superior, separable clusters in both real-world data sets., Discussion: Selecting an appropriate mixed-type metric allows the investigator to obtain optimal separation of patient clusters and get maximum use of their data. Superior metrics for mixed-type data handle multiple data types using multiple, type-focused distances. Better subclassification of disease opens avenues for targeted treatments, precision medicine, clinical decision support, and improved patient outcomes., (Copyright © 2021 Elsevier Inc. All rights reserved.)

Background: In the intensive care unit (ICU), delirium is a common, acute, confusional state associated with high risk for short- and long-term morbidity and mortality. Machine learning (ML) has promise to address research priorities and improve delirium outcomes. However, due to clinical and billing conventions, delirium is often inconsistently or incompletely labeled in electronic health record (EHR) datasets. Here, we identify clinical actions abstracted from clinical guidelines in electronic health records (EHR) data that indicate risk of delirium among intensive care unit (ICU) patients. We develop a novel prediction model to label patients with delirium based on a large data set and assess model performance., Methods: EHR data on 48,451 admissions from 2001 to 2012, available through Medical Information Mart for Intensive Care-III database (MIMIC-III), was used to identify features to develop our prediction models. Five binary ML classification models (Logistic Regression; Classification and Regression Trees; Random Forests; Naïve Bayes; and Support Vector Machines) were fit and ranked by Area Under the Curve (AUC) scores. We compared our best model with two models previously proposed in the literature for goodness of fit, precision, and through biological validation., Results: Our best performing model with threshold reclassification for predicting delirium was based on a multiple logistic regression using the 31 clinical actions (AUC 0.83). Our model out performed other proposed models by biological validation on clinically meaningful, delirium-associated outcomes., Conclusions: Hurdles in identifying accurate labels in large-scale datasets limit clinical applications of ML in delirium. We developed a novel labeling model for delirium in the ICU using a large, public data set. By using guideline-directed clinical actions independent from risk factors, treatments, and outcomes as model predictors, our classifier could be used as a delirium label for future clinically targeted models.

Background: There have been many recent breakthroughs in processing and analyzing large-scale data sets in biomedical informatics. For example, the CytoGPS algorithm has enabled the use of text-based karyotypes by transforming them into a binary model. However, such advances are accompanied by new problems of data sparsity, heterogeneity, and noisiness that are magnified by the large-scale multidimensional nature of the data. To address these problems, we developed the Mercator R package, which processes and visualizes binary biomedical data. We use Mercator to address biomedical questions of cytogenetic patterns relating to lymphoid hematologic malignancies, which include a broad set of leukemias and lymphomas. Karyotype data are one of the most common form of genetic data collected on lymphoid malignancies, because karyotyping is part of the standard of care in these cancers., Results: In this paper we combine the analytic power of CytoGPS and Mercator to perform a large-scale multidimensional pattern recognition study on 22,741 karyotype samples in 47 different hematologic malignancies obtained from the public Mitelman database., Conclusion: Our findings indicate that Mercator was able to identify both known and novel cytogenetic patterns across different lymphoid malignancies, furthering our understanding of the genetics of these diseases.

Karyotyping, the practice of visually examining and recording chromosomal abnormalities, is commonly used to diagnose diseases of genetic origin, including cancers. Karyotypes are recorded as text written in the International System for Human Cytogenetic Nomenclature (ISCN). Downstream analysis of karyotypes is conducted manually, due to the visual nature of analysis and the linguistic structure of the ISCN. The ISCN has not been computer-readable and, as such, prevents the full potential of these genomic data from being realized. In response, we developed CytoGPS, a platform to analyze large volumes of cytogenetic data using a Loss-Gain-Fusion model that converts the human-readable ISCN karyotypes into a machine-readable binary format. As proof of principle, we applied CytoGPS to cytogenetic data from the Mitelman Database of Chromosome Aberrations and Gene Fusions in Cancer, a National Cancer Institute hosted database of over 69,000 karyotypes of human cancers. Using the Jaccard coefficient to determine similarity between karyotypes structured as binary vectors, we were able to identify novel patterns from 4,968 Mitelman CML karyotypes, such as the co-occurrence of trisomy 19 and 21. The CytoGPS platform unlocks the potential for large-scale, comparative analysis of cytogenetic data. This methodological platform is freely available at CytoGPS.org., (Copyright © 2020. Published by Elsevier Inc.)

Objective: Unsupervised machine learning approaches hold promise for large-scale clinical data. However, the heterogeneity of clinical data raises new methodological challenges in feature selection, choosing a distance metric that captures biological meaning, and visualization. We hypothesized that clustering could discover prognostic groups from patients with chronic lymphocytic leukemia, a disease that provides biological validation through well-understood outcomes., Methods: To address this challenge, we applied k-medoids clustering with 10 distance metrics to 2 experiments ("A" and "B") with mixed clinical features collapsed to binary vectors and visualized with both multidimensional scaling and t-stochastic neighbor embedding. To assess prognostic utility, we performed survival analysis using a Cox proportional hazard model, log-rank test, and Kaplan-Meier curves., Results: In both experiments, survival analysis revealed a statistically significant association between clusters and survival outcomes (A: overall survival, P = .0164; B: time from diagnosis to treatment, P = .0039). Multidimensional scaling separated clusters along a gradient mirroring the order of overall survival. Longer survival was associated with mutated immunoglobulin heavy-chain variable region gene (IGHV) status, absent Zap 70 expression, female sex, and younger age., Conclusions: This approach to mixed-type data handling and selection of distance metric captured well-understood, binary, prognostic markers in chronic lymphocytic leukemia (sex, IGHV mutation status, ZAP70 expression status) with high fidelity., (© The Author(s) 2020. Published by Oxford University Press on behalf of the American Medical Informatics Association.)

Neuronal axons terminate as synaptic boutons that form stable yet plastic connections with their targets. Synaptic bouton development relies on an underlying network of both long-lived and dynamic microtubules that provide structural stability for the boutons while also allowing for their growth and remodeling. However, a molecular-scale mechanism that explains how neurons appropriately balance these two microtubule populations remains a mystery. We hypothesized that α-tubulin acetyltransferase (αTAT), which both stabilizes long-lived microtubules against mechanical stress via acetylation and has been implicated in promoting microtubule dynamics, could play a role in this process. Using the Drosophila neuromuscular junction as a model, we found that non-enzymatic dαTAT activity limits the growth of synaptic boutons by affecting dynamic, but not stable, microtubules. Loss of dαTAT results in the formation of ectopic boutons. These ectopic boutons can be similarly suppressed by resupplying enzyme-inactive dαTAT or by treatment with a low concentration of the microtubule-targeting agent vinblastine, which acts to suppress microtubule dynamics. Biophysical reconstitution experiments revealed that non-enzymatic αTAT1 activity destabilizes dynamic microtubules but does not substantially impact the stability of long-lived microtubules. Further, during microtubule growth, non-enzymatic αTAT1 activity results in increasingly extended tip structures, consistent with an increased rate of acceleration of catastrophe frequency with microtubule age, perhaps via tip structure remodeling. Through these mechanisms, αTAT enriches for stable microtubules at the expense of dynamic ones. We propose that the specific suppression of dynamic microtubules by non-enzymatic αTAT activity regulates the remodeling of microtubule networks during synaptic bouton development., Competing Interests: Declaration of Interests The authors declare no competing interests., (Copyright © 2019 Elsevier Ltd. All rights reserved.)

Purpose of Review: Novel technologies, such as high-definition cameras, encryption software, electronic stethoscopes, microfluidic diagnostic systems, and widely available broadband Internet have expanded the potential for telemedicine. This narrative review presents current and future uses of telemedicine in the prevention, diagnosis, treatment, stewardship, and management of infectious disease., Recent Findings: Beginning in the 1990s, early approaches to telemedicine in infectious disease focused largely on treatment of HIV/AIDS, hepatitis C, and tuberculosis. However, recent innovations allow for targeting of additional diseases and in increasingly remote settings. Telemedicine allows virtual visits between patients in the home and remote providers, permitting outpatient management of complex conditions, such as post-surgical site monitoring, and non-urgent infectious maladies, such as uncomplicated urinary tract infection. Remote provider education by videoconference and integrated clinical decision support tools create avenues to improve inpatient care, including antimicrobial stewardship. Technological strides from miniaturization of diagnostic tests to robotic telepresence physical exams improve access to infectious disease care in isolated and infrastructure-poor environments, from cargo ships to other resource-limited settings. Telemedicine in the field of infectious disease is rapidly expanding in clinical, technological, geographical, and human capacity. Recent innovations narrow gaps in access to care for populations traditionally underserved, stigmatized, isolated by remote geography, or lacking technological infrastructure. Current and future approaches will transform inpatient, outpatient, and remote care.

TPX2 is a widely conserved microtubule-associated protein that is required for mitotic spindle formation and function. Previous studies have demonstrated that TPX2 is required for the nucleation of microtubules around chromosomes; however, the molecular mechanism by which TPX2 promotes microtubule nucleation remains a mystery. In this study, we found that TPX2 acts to suppress tubulin subunit off-rates during microtubule assembly and disassembly, thus allowing for the support of unprecedentedly slow rates of plus-end microtubule growth, and also leading to a dramatically reduced microtubule shortening rate. These changes in microtubule dynamics can be explained in computational simulations by a moderate increase in tubulin-tubulin bond strength upon TPX2 association with the microtubule lattice, which in turn acts to reduce the departure rate of tubulin subunits from the microtubule ends. Thus, the direct suppression of tubulin subunit off-rates by TPX2 during microtubule growth and shortening could provide a molecular mechanism to explain the nucleation of new microtubules in the presence of TPX2., (© 2016. Published by The Company of Biologists Ltd.)

During cell division, a microtubule-based mitotic spindle mediates the faithful segregation of duplicated chromosomes into daughter cells. Proper length control of the metaphase mitotic spindle is critical to this process and is thought to be achieved through a mechanism in which spindle pole separation forces from plus-end-directed motors are balanced by forces from minus-end-directed motors that pull spindle poles together. However, in contrast to this model, metaphase mitotic spindles with inactive kinesin-14 minus-end-directed motors often have shorter spindle lengths, along with poorly aligned spindle microtubules. A mechanistic explanation for this paradox is unknown. Using computational modeling, in vitro reconstitution, live-cell fluorescence microscopy, and electron microscopy, we now find that the budding yeast kinesin-14 molecular motor Kar3-Cik1 can efficiently align spindle microtubules along the spindle axis. This then allows plus-end-directed kinesin-5 motors to efficiently exert the outward microtubule sliding forces needed for proper spindle bipolarity., (Copyright © 2014 Elsevier Inc. All rights reserved.)

Microtubules are key structural and transport elements in cells. The dynamics at microtubule ends are characterized by periods of slow growth, followed by stochastic switching events termed "catastrophes," in which microtubules suddenly undergo rapid shortening. Growing microtubules are thought to be protected from catastrophe by a GTP-tubulin "cap": GTP-tubulin subunits add to the tips of growing microtubules but are subsequently hydrolyzed to GDP-tubulin subunits once they are incorporated into the microtubule lattice. Loss of the GTP-tubulin cap exposes GDP-tubulin subunits at the microtubule tip, resulting in a catastrophe event. However, the mechanistic basis for sudden loss of the GTP cap, leading to catastrophe, is not known. To investigate microtubule catastrophe events, we performed 3D mechanochemical simulations that account for interactions between neighboring protofilaments. We found that there are two separate factors that contribute to catastrophe events in the 3D simulation: the GTP-tubulin cap size, which settles into a steady-state value that depends on the free tubulin concentration during microtubule growth, and the structure of the microtubule tip. Importantly, 3D simulations predict, and both fluorescence and electron microscopy experiments confirm, that microtubule tips become more tapered as the microtubule grows. This effect destabilizes the tip and ultimately contributes to microtubule catastrophe. Thus, the likelihood of a catastrophe event may be intimately linked to the aging physical structure of the growing microtubule tip. These results have important consequences for catastrophe regulation in cells, as microtubule-associated proteins could promote catastrophe events in part by modifying microtubule tip structures., (Copyright © 2013 Elsevier Ltd. All rights reserved.)

Recent advances in DNA synthesis technology have enabled the construction of novel genetic pathways and genomic elements, furthering our understanding of system-level phenomena. The ability to synthesize large segments of DNA allows the engineering of pathways and genomes according to arbitrary sets of design principles. Here we describe a synthetic yeast genome project, Sc2.0, and the first partially synthetic eukaryotic chromosomes, Saccharomyces cerevisiae chromosome synIXR, and semi-synVIL. We defined three design principles for a synthetic genome as follows: first, it should result in a (near) wild-type phenotype and fitness; second, it should lack destabilizing elements such as tRNA genes or transposons; and third, it should have genetic flexibility to facilitate future studies. The synthetic genome features several systemic modifications complying with the design principles, including an inducible evolution system, SCRaMbLE (synthetic chromosome rearrangement and modification by loxP-mediated evolution). We show the utility of SCRaMbLE as a novel method of combinatorial mutagenesis, capable of generating complex genotypes and a broad variety of phenotypes. When complete, the fully synthetic genome will allow massive restructuring of the yeast genome, and may open the door to a new type of combinatorial genetics based entirely on variations in gene content and copy number., (© 2011 Macmillan Publishers Limited. All rights reserved)

Mutations in the leucine-rich repeat kinase 2 (LRRK2) gene are associated with late-onset, autosomal-dominant, familial Parkinson's disease (PD) and also contribute to sporadic disease. The LRRK2 gene encodes a large protein with multiple domains, including functional Roc GTPase and protein kinase domains. Mutations in LRRK2 most likely cause disease through a toxic gain-of-function mechanism. The expression of human LRRK2 variants in cultured primary neurons induces toxicity that is dependent on intact GTP binding or kinase activities. However, the mechanism(s) underlying LRRK2-induced neuronal toxicity is poorly understood, and the contribution of GTPase and/or kinase activity to LRRK2 pathobiology is not well defined. To explore the pathobiology of LRRK2, we have developed a model of LRRK2 cytotoxicity in the baker's yeast Saccharomyces cerevisiae. Protein domain analysis in this model reveals that expression of GTPase domain-containing fragments of human LRRK2 are toxic. LRRK2 toxicity in yeast can be modulated by altering GTPase activity and is closely associated with defects in endocytic vesicular trafficking and autophagy. These truncated LRRK2 variants induce similar toxicity in both yeast and primary neuronal models and cause similar vesicular defects in yeast as full-length LRRK2 causes in primary neurons. The toxicity induced by truncated LRRK2 variants in yeast acts through a mechanism distinct from toxicity induced by human alpha-synuclein. A genome-wide genetic screen identified modifiers of LRRK2-induced toxicity in yeast including components of vesicular trafficking pathways, which can also modulate the trafficking defects caused by expression of truncated LRRK2 variants. Our results provide insight into the basic pathobiology of LRRK2 and suggest that the GTPase domain may contribute to the toxicity of LRRK2. These findings may guide future therapeutic strategies aimed at attenuating LRRK2-mediated neurodegeneration., Competing Interests: The authors have declared that no competing interests exist.

Background: Mutations within the leucine-rich repeat kinase 2 (LRRK2) gene account for a significant proportion of autosomal-dominant and some late-onset sporadic Parkinson's disease. Elucidation of LRRK2 protein function in health and disease provides an opportunity for deciphering molecular pathways important in neurodegeneration. In mammals, LRRK1 and LRRK2 protein comprise a unique family encoding a GTPase domain that controls intrinsic kinase activity. The expression profiles of the murine LRRK proteins have not been fully described and insufficiently characterized antibodies have produced conflicting results in the literature., Results: Herein, we comprehensively evaluate twenty-one commercially available antibodies to the LRRK2 protein using mouse LRRK2 and human LRRK2 expression vectors, wild-type and LRRK2-null mouse brain lysates and human brain lysates. Eleven antibodies detect over-expressed human LRRK2 while four antibodies detect endogenous human LRRK2. In contrast, two antibodies recognize over-expressed mouse LRRK2 and one antibody detected endogenous mouse LRRK2. LRRK2 protein resides in both soluble and detergent soluble protein fractions. LRRK2 and the related LRRK1 genes encode low levels of expressed mRNA species corresponding to low levels of protein both during development and in adulthood with largely redundant expression profiles., Conclusion: Despite previously published results, commercially available antibodies generally fail to recognize endogenous mouse LRRK2 protein; however, several antibodies retain the ability to detect over-expressed mouse LRRK2 protein. Over half of the commercially available antibodies tested detect over-expressed human LRRK2 protein and some have sufficient specificity to detect endogenous LRRK2 in human brain. The mammalian LRRK proteins are developmentally regulated in several tissues and coordinated expression suggest possible redundancy in the function between LRRK1 and LRRK2.

Long interspersed element type 1 (L1) retrotransposons are ubiquitous mammalian mobile elements and potential tools for in vivo mutagenesis; however, native L1 elements are relatively inactive in mice when introduced as transgenes. We have previously described a synthetic L1 element, ORFeus, containing two synonymously recoded ORFs relative to mouse L1. It is significantly more active for retrotransposition in cell culture than all native L1 elements tested. To study its activity in vivo, we developed a transgenic mouse model in which ORFeus expression was controlled by a constitutive heterologous promoter, and we established definitive evidence for ORFeus retrotransposition activity both in germ line and somatic tissues. Germ line retrotransposition frequencies resulting in 0.33 insertions per animal are seen among progeny of ORFeus donor element heterozygotes derived from a single founder, representing a >20-fold increase over native L1 elements. We observe somatic transposition events in 100% of the ORFeus donor-containing animals, and an average of 17 different insertions are easily recovered from each animal; modeling suggests that the number of somatic insertions per animal exceeds this number by perhaps several orders of magnitude. Nearly 200 insertions were precisely mapped, and their distribution in the mouse genome appears random relative to transcription units and guanine-cytosine content. The results suggest that ORFeus may be developed into useful tools for in vivo mutagenesis.

Ty1 elements are long terminal repeat (LTR) retrotransposons that reside within the genome of Saccharomyces cerevisiae. It has been known for many years that the 2'-5' phosphodiesterase Dbr1p, which debranches intron lariats, is required for efficient Ty1 transposition. A recent report suggested the intriguing possibility that Ty1 RNA forms a lariat as a transposition intermediate. We set out to further investigate the nature of the proposed Ty1 lariat branchpoint. However, using a wide range of techniques we were unable to find any evidence for the proposed lariat structure. Furthermore, we demonstrate that some of the techniques used in the initial study describing the lariat are capable of incorrectly reporting a lariat structure. Thus, the role of the Dbr1 protein in Ty1 retrotransposition remains elusive.