Using machine learning to improve anaphylaxis case identification in medical claims data.

Objectives: Anaphylaxis is a severe life-threatening allergic reaction, and its accurate identification in healthcare databases can harness the potential of "Big Data" for healthcare or public health purposes. Materials and methods: This study used claims data obtained between October 1, 2015 and February 28, 2019 from the CMS database to examine the utility of machine learning in identifying incident anaphylaxis cases. We created a feature selection pipeline to identify critical features between different datasets. Then a variety of unsupervised and supervised methods were used (eg, Sammon mapping and eXtreme Gradient Boosting) to train models on datasets of differing data quality, which reflects the varying availability and potential rarity of ground truth data in medical databases. Results: Resulting machine learning model accuracies ranged from 47.7% to 94.4% when tested on ground truth data. Finally, we found new features to help experts enhance existing case-finding algorithms. Discussion: Developing precise algorithms to detect medical outcomes in claims can be a laborious and expensive process, particularly for conditions presented and coded diversely. We found it beneficial to filter out highly potent codes used for data curation to identify underlying patterns and features. To improve rule-based algorithms where necessary, researchers could use model explainers to determine noteworthy features, which could then be shared with experts and included in the algorithm. Conclusion: Our work suggests machine learning models can perform at similar levels as a previously published expert case-finding algorithm, while also having the potential to improve performance or streamline algorithm construction processes by identifying new relevant features for algorithm construction.

Using machine learning to improve anaphylaxis case identification in medical claims data.

Objective: Anaphylaxis is a severe life-threatening allergic reaction, and its accurate identification in healthcare databases can harness the potential of "Big Data" for healthcare or public health purposes. Methods: This study used claims data obtained between October 1, 2015 and February 28, 2019 from the CMS database to examine the utility of machine learning in identifying incident anaphylaxis cases. We created a feature selection pipeline to identify critical features between different datasets. Then a variety of unsupervised and supervised methods were used (eg, Sammon mapping and eXtreme Gradient Boosting) to train models on datasets of differing data quality, which reflects the varying availability and potential rarity of ground truth data in medical databases. Results: Resulting machine learning model accuracies ranged between 47.7% and 94.4% when tested on ground truth data. Finally, we found new features to help experts enhance existing case-finding algorithms. Discussion: Developing precise algorithms to detect medical outcomes in claims can be a laborious and expensive process, particularly for conditions presented and coded diversely. We found it beneficial to filter out highly potent codes used for data curation to identify underlying patterns and features. To improve rule-based algorithms where necessary, researchers could use model explainers to determine noteworthy features, which could then be shared with experts and included in the algorithm. Conclusion: Our work suggests machine learning models can perform at similar levels as a previously published expert case-finding algorithm, while also having the potential to improve performance or streamline algorithm construction processes by identifying new relevant features for algorithm construction.

Preclinical safety and biodistribution of CRISPR targeting SIV in non-human primates.

In this study, we demonstrate the safety and utility of CRISPR-Cas9 gene editing technology for in vivo editing of proviral DNA in ART-treated, virally controlled simian immunodeficiency virus (SIV) infected rhesus macaques, an established model for HIV infection. EBT-001 is an AAV9-based vector delivering SaCas9 and dual guide RNAs designed to target multiple regions of the SIV genome: the viral LTRs, and the Gag gene. The results presented here demonstrate that a single IV inoculation of EBT-001 at each of 3 dose levels (1.4 × 1012, 1.4 × 1013 and 1.4 × 1014 genome copies/kg) resulted in broad and functional biodistribution of AAV9-EBT-001 to known tissue reservoirs of SIV. No off-target effects or abnormal pathology were observed, and animals returned to their normal body weight after receiving EBT-001. Importantly, the macaques that received the 2 highest doses of EBT-001 showed improved absolute lymphocyte counts as compared to antiretroviral-treated controls. Taken together, these results demonstrate safety, biodistribution, and in vivo proviral DNA editing following IV administration of EBT-001, supporting the further development of CRISPR-based gene editing as a potential therapeutic approach for HIV in humans.

Deep learning models for histologic grading of breast cancer and association with disease prognosis.

Histologic grading of breast cancer involves review and scoring of three well-established morphologic features: mitotic count, nuclear pleomorphism, and tubule formation. Taken together, these features form the basis of the Nottingham Grading System which is used to inform breast cancer characterization and prognosis. In this study, we develop deep learning models to perform histologic scoring of all three components using digitized hematoxylin and eosin-stained slides containing invasive breast carcinoma. We first evaluate model performance using pathologist-based reference standards for each component. To complement this typical approach to evaluation, we further evaluate the deep learning models via prognostic analyses. The individual component models perform at or above published benchmarks for algorithm-based grading approaches, achieving high concordance rates with pathologist grading. Further, prognostic performance using deep learning-based grading is on par with that of pathologists performing review of matched slides. By providing scores for each component feature, the deep-learning based approach also provides the potential to identify the grading components contributing most to prognostic value. This may enable optimized prognostic models, opportunities to improve access to consistent grading, and approaches to better understand the links between histologic features and clinical outcomes in breast cancer.

Independent assessment of a deep learning system for lymph node metastasis detection on the Augmented Reality Microscope.

Several machine learning algorithms have demonstrated high predictive capability in the identification of cancer within digitized pathology slides. The Augmented Reality Microscope (ARM) has allowed these algorithms to be seamlessly integrated within the pathology workflow by overlaying their inferences onto its microscopic field of view in real time. We present an independent assessment of the LYmph Node Assistant (LYNA) models, state-of-the-art algorithms for the identification of breast cancer metastases in lymph node biopsies, optimized for usage on the ARM. We assessed the models on 40 whole slide images at the commonly used objective magnifications of 10×, 20×, and 40×. We analyzed their performance across clinically relevant subclasses of tissue, including breast cancer, lymphocytes, histiocytes, blood, and fat. Each model obtained overall AUC values of approximately 0.98, accuracy values of approximately 0.94, and sensitivity values above 0.88 at classifying small regions of a field of view as benign or cancerous. Across tissue subclasses, the models performed most accurately on fat and blood, and least accurately on histiocytes, germinal centers, and sinus. The models also struggled with the identification of isolated tumor cells, especially at lower magnifications. After testing, we reviewed the discrepancies between model predictions and ground truth to understand the causes of error. We introduce a distinction between proper and improper ground truth for analysis in cases of uncertain annotations. Taken together, these methods comprise a novel approach for exploratory model analysis over complex anatomic pathology data in which precise ground truth is difficult to establish.

Enabling precision medicine via standard communication of HTS provenance, analysis, and results.

A personalized approach based on a patient's or pathogen's unique genomic sequence is the foundation of precision medicine. Genomic findings must be robust and reproducible, and experimental data capture should adhere to findable, accessible, interoperable, and reusable (FAIR) guiding principles. Moreover, effective precision medicine requires standardized reporting that extends beyond wet-lab procedures to computational methods. The BioCompute framework (https://w3id.org/biocompute/1.3.0) enables standardized reporting of genomic sequence data provenance, including provenance domain, usability domain, execution domain, verification kit, and error domain. This framework facilitates communication and promotes interoperability. Bioinformatics computation instances that employ the BioCompute framework are easily relayed, repeated if needed, and compared by scientists, regulators, test developers, and clinicians. Easing the burden of performing the aforementioned tasks greatly extends the range of practical application. Large clinical trials, precision medicine, and regulatory submissions require a set of agreed upon standards that ensures efficient communication and documentation of genomic analyses. The BioCompute paradigm and the resulting BioCompute Objects (BCOs) offer that standard and are freely accessible as a GitHub organization (https://github.com/biocompute-objects) following the "Open-Stand.org principles for collaborative open standards development." With high-throughput sequencing (HTS) studies communicated using a BCO, regulatory agencies (e.g., Food and Drug Administration [FDA]), diagnostic test developers, researchers, and clinicians can expand collaboration to drive innovation in precision medicine, potentially decreasing the time and cost associated with next-generation sequencing workflow exchange, reporting, and regulatory reviews.

In vitro cytokine licensing induces persistent permissive chromatin at the Indoleamine 2,3-dioxygenase promoter.

BACKGROUND: Mesenchymal stromal cells (MSCs) are being investigated as therapies for inflammatory diseases due to their immunosuppressive capacity. Interferon (IFN)-γ treatment primes MSC immunosuppression partially through induction of Indoleamine 2,3-dioxygenase (IDO1), which depletes tryptophan necessary to support proliferation of activated T cells. We investigated the role of histone modifications in the timing and maintenance of induced IDO1 expression in MSCs under clinical manufacturing conditions, such as cryopreservation. METHODS: We used chromatin immunoprecipitation and quantitative polymerase chain reaction (PCR) to assay levels of transcriptionally permissive acetylated H3K9 and repressive trimethylated H3K9 histone modifications surrounding the transcriptional start site for IDO1, and reverse transcriptase PCR and immunoblotting to detect messenger RNA (mRNA) and protein. RESULTS: MSCs derived from three donors approached maximum IDO1 mRNA levels following 24 hours of in vitro cytokine treatment. Induction of IDO1 expression correlated with increased acetylation of H3K9 concomitant with reduction of trimethylated H3K9 modifications at the promoter. Examination of two additional donors confirmed this result. While induced IDO1 levels decreased within 2 days after cytokine removal and freeze thawing, the activated chromatin state was maintained. Upon re-exposure to cytokines, previously primed MSCs accumulated near-maximum IDO1 mRNA levels within 4-8 h. DISCUSSION: Our data indicate that in vitro priming of MSCs causes chromatin remodeling at the IDO1 promoter, that this alteration is maintained during processing commonly used to prepare MSCs for clinical use and that, once primed, MSCs are poised for IDO1 expression even in the absence of cytokines.

High-performance integrated virtual environment (HIVE): a robust infrastructure for next-generation sequence data analysis.

The High-performance Integrated Virtual Environment (HIVE) is a distributed storage and compute environment designed primarily to handle next-generation sequencing (NGS) data. This multicomponent cloud infrastructure provides secure web access for authorized users to deposit, retrieve, annotate and compute on NGS data, and to analyse the outcomes using web interface visual environments appropriately built in collaboration with research and regulatory scientists and other end users. Unlike many massively parallel computing environments, HIVE uses a cloud control server which virtualizes services, not processes. It is both very robust and flexible due to the abstraction layer introduced between computational requests and operating system processes. The novel paradigm of moving computations to the data, instead of moving data to computational nodes, has proven to be significantly less taxing for both hardware and network infrastructure.The honeycomb data model developed for HIVE integrates metadata into an object-oriented model. Its distinction from other object-oriented databases is in the additional implementation of a unified application program interface to search, view and manipulate data of all types. This model simplifies the introduction of new data types, thereby minimizing the need for database restructuring and streamlining the development of new integrated information systems. The honeycomb model employs a highly secure hierarchical access control and permission system, allowing determination of data access privileges in a finely granular manner without flooding the security subsystem with a multiplicity of rules. HIVE infrastructure will allow engineers and scientists to perform NGS analysis in a manner that is both efficient and secure. HIVE is actively supported in public and private domains, and project collaborations are welcomed. Database URL: https://hive.biochemistry.gwu.edu.

Chromosomal stability of mesenchymal stromal cells during in vitro culture.

BACKGROUND AIMS: Mesenchymal stromal cells (MSCs) are being investigated for use in cell therapy. The extensive in vitro expansion necessary to obtain sufficient cells for clinical use increases the risk that genetically abnormal cells will arise and be propagated during cell culture. Genetic abnormalities may lead to transformation and poor performance in clinical use, and are a critical safety concern for cell therapies using MSCs. METHODS: We used spectral karyotyping (SKY) to investigate the genetic stability of human MSCs from ten donors during passaging. RESULTS: Our data indicate that chromosomal abnormalities exist in MSCs at early passages and can be clonally propagated. The karyotypic abnormalities observed during our study diminished during passage. CONCLUSIONS: Karyotyping of MSCs reveals characteristics which may be valuable in deciding the suitability of cells for further use. Karyotypic analysis is useful for monitoring the genetic stability of MSCs during expansion.

Chromatin Changes at the PPAR-γ2 Promoter During Bone Marrow-Derived Multipotent Stromal Cell Culture Correlate With Loss of Gene Activation Potential.

Bone marrow-derived multipotent stromal cells (BM-MSCs) display a broad range of therapeutically valuable properties, including the capacity to form skeletal tissues and dampen immune system responses. However, to use BM-MSCs in a clinical setting, amplification is required, which may introduce epigenetic changes that affect biological properties. Here we used chromatin immunoprecipitation to compare post-translationally modified histones at a subset of gene promoters associated with developmental and environmental plasticity in BM-MSCs from multiple donors following culture expansion. At many locations, we observed localization of both transcriptionally permissive (H3K4me3) and repressive (H3K27me3) histone modifications. These chromatin signatures were consistent among BM-MSCs from multiple donors. Since promoter activity depends on the relative levels of H3K4me3 and H3K27me3, we examined the ratio of H3K4me3 to H3K27me3 (K4/K27) at promoters during culture expansion. The H3K4me3 to H3K27me3 ratios were maintained at most assayed promoters over time. The exception was the adipose-tissue specific promoter for the PPAR-γ2 isoform of PPAR-γ, which is a critical positive regulator of adipogenesis. At PPAR-γ2, we observed a change in K4/K27 levels favoring the repressed chromatin state during culture. This change correlated with diminished promoter activity in late passage cells exposed to adipogenic stimuli. In contrast to BM-MSCs and osteoblasts, lineage-restricted preadipocytes exhibited levels of H3K4me3 and H3K27me3 that favored the permissive chromatin state at PPAR-γ2. These results demonstrate that locus-specific changes in H3K4me3 and H3K27me3 levels can occur during BM-MSC culture that may affect their properties. Stem Cells 2015;33:2169-2181.

DOT2: Macromolecular docking with improved biophysical models.

Computational docking is a useful tool for predicting macromolecular complexes, which are often difficult to determine experimentally. Here, we present the DOT2 software suite, an updated version of the DOT intermolecular docking program. DOT2 provides straightforward, automated construction of improved biophysical models based on molecular coordinates, offering checkpoints that guide the user to include critical features. DOT has been updated to run more quickly, allow flexibility in grid size and spacing, and generate an infinitive complete list of favorable candidate configurations. Output can be filtered by experimental data and rescored by the sum of electrostatic and atomic desolvation energies. We show that this rescoring method improves the ranking of correct complexes for a wide range of macromolecular interactions and demonstrate that biologically relevant models are essential for biologically relevant results. The flexibility and versatility of DOT2 accommodate realistic models of complex biological systems, improving the likelihood of a successful docking outcome.

Comparative surface geometry of the protein kinase family.

Identifying conserved pockets on the surfaces of a family of proteins can provide insight into conserved geometric features and sites of protein-protein interaction. Here we describe mapping and comparison of the surfaces of aligned crystallographic structures, using the protein kinase family as a model. Pockets are rapidly computed using two computer programs, FADE and Crevasse. FADE uses gradients of atomic density to locate grooves and pockets on the molecular surface. Crevasse, a new piece of software, splits the FADE output into distinct pockets. The computation was run on 10 kinase catalytic cores aligned on the alphaF-helix, and the resulting pockets spatially clustered. The active site cleft appears as a large, contiguous site that can be subdivided into nucleotide and substrate docking sites. Substrate specificity determinants in the active site cleft between serine/threonine and tyrosine kinases are visible and distinct. The active site clefts cluster tightly, showing a conserved spatial relationship between the active site and alphaF-helix in the C-lobe. When the alphaC-helix is examined, there are multiple mechanisms for anchoring the helix using spatially conserved docking sites. A novel site at the top of the N-lobe is present in all the kinases, and there is a large conserved pocket over the hinge and the alphaC-beta4 loop. Other pockets on the kinase core are strongly conserved but have not yet been mapped to a protein-protein interaction. Sites identified by this algorithm have revealed structural and spatially conserved features of the kinase family and potential conserved intermolecular and intramolecular binding sites.