Diffeomorphic Counterfactuals With Generative Models.

Counterfactuals can explain classification decisions of neural networks in a human interpretable way. We propose a simple but effective method to generate such counterfactuals. More specifically, we perform a suitable diffeomorphic coordinate transformation and then perform gradient ascent in these coordinates to find counterfactuals which are classified with great confidence as a specified target class. We propose two methods to leverage generative models to construct such suitable coordinate systems that are either exactly or approximately diffeomorphic. We analyze the generation process theoretically using Riemannian differential geometry and validate the quality of the generated counterfactuals using various qualitative and quantitative measures.

#Covid-19: An exploratory investigation of hashtag usage on Twitter.

BACKGROUND: The literature highlights Twitter as a vital instrument tool for health policy-makers for health communication and promotion. Furthermore, Twitter is a tool allowing us to understand the focus of people regarding a topic of interest. OBJECTIVE: To provide health policy-makers with insights concerning key topics of interest in the Twitter community regarding Covid-19, and to support information search and health communication. METHOD: A total of 28.5M tweets have been retrieved, of which 6.9M tweets included hashtags. The data was analyzed using data science and natural language processing libraries. Qualitative analysis was performed using thematic analysis. RESULTS: 907k different hashtags were used. Of these, only 1192 hashtags were used more than 1000 times. The qualitative analysis resulted in 13 themes. The top three themes regarding the number of hashtags used were related to Covid-19, identifying information, interventions, and geographical tagging. We explored the relationship between themes and showed how health practitioners can understand the communication in relation to specific topics expressed as hashtags (e.g., #stayhome). CONCLUSIONS: The results provide first insights for policy-makers and health practitioners to identify relevant tweets and to choose appropriate hashtags for health communication. The results also show that only with a limited number of Tweets (10 per day) health organizations could have been among the top users.

Interprofessional Emergency Training Leads to Changes in the Workplace.

INTRODUCTION: Preventable mistakes occur frequently and can lead to patient harm and death. The emergency department (ED) is notoriously prone to such errors, and evidence suggests that improving teamwork is a key aspect to reduce the rate of error in acute care settings. Only a few strategies are in place to train team skills and communication in interprofessional situations. Our goal was to conceptualize, implement, and evaluate a training module for students of three professions involved in emergency care. The objective was to sensitize participants to barriers for their team skills and communication across professional borders. METHODS: We developed a longitudinal simulation-enhanced training format for interprofessional teams, consisting of final-year medical students, advanced trainees of emergency nursing and student paramedics. The training format consisted of several one-day training modules, which took place twice in 2016 and 2017. Each training module started with an introduction to share one's roles, professional self-concepts, common misconceptions, and communication barriers. Next, we conducted different simulated cases. Each case consisted of a prehospital section (for paramedics and medical students), a handover (everyone), and an ED section (medical students and emergency nurses). After each training module, we assessed participants' "Commitment to Change." In this questionnaire, students were anonymously asked to state up to three changes that they wished to implement as a result of the course, as well as the strength of their commitment to these changes. RESULTS: In total, 64 of 80 participants (80.0%) made at least one commitment to change after participating in the training modules. The total of 123 commitments was evenly distributed over four emerging categories: communication, behavior, knowledge and attitude. Roughly one third of behavior- and attitude-related commitments were directly related to interprofessional topics (e.g., "acknowledge other professions' work"), and these were equally distributed among professions. At the two-month follow-up, 32 participants (50%) provided written feedback on their original commitments: 57 of 62 (91.9%) commitments were at least partly realized at the follow-up, and only five (8.1%) commitments lacked realization entirely. CONCLUSION: A structured simulation-enhanced intervention was successful in promoting change to the practice of emergency care, while training teamwork and communication skills jointly.

The chronically inflamed central nervous system provides niches for long-lived plasma cells.

Although oligoclonal bands in the cerebrospinal fluid have been a hallmark of multiple sclerosis diagnosis for over three decades, the role of antibody-secreting cells in multiple sclerosis remains unclear. T and B cells are critical for multiple sclerosis pathogenesis, but increasing evidence suggests that plasma cells also contribute, through secretion of autoantibodies. Long-lived plasma cells are known to drive various chronic inflammatory conditions as e.g. systemic lupus erythematosus, however, to what extent they are present in autoimmune central nervous system inflammation has not yet been investigated. In brain biopsies from multiple sclerosis patients and other neurological diseases, we could detect non-proliferating plasma cells (CD138+Ki67-) in the parenchyma. Based on this finding, we hypothesized that long-lived plasma cells can persist in the central nervous system (CNS). In order to test this hypothesis, we adapted the multiple sclerosis mouse model experimental autoimmune encephalomyelitis to generate a B cell memory response. Plasma cells were found in the meninges and the parenchyma of the inflamed spinal cord, surrounded by tissue areas resembling survival niches for these cells, characterized by an up-regulation of chemokines (CXCL12), adhesion molecules (VCAM-1) and survival factors (APRIL and BAFF). In order to determine the lifetime of plasma cells in the chronically inflamed CNS, we labeled the DNA of proliferating cells with 5-ethynyl-2'-deoxyuridine (EdU). Up to five weeks later, we could detect EdU+ long-lived plasma cells in the murine CNS. To our knowledge, this is the first study describing non-proliferating plasma cells directly in the target tissue of a chronic inflammation in humans, as well as the first evidence demonstrating the ability of plasma cells to persist in the CNS, and the ability of the chronically inflamed CNS tissue to promote this persistence. Hence, our results suggest that the CNS provides survival niches for long-lived plasma cells, similar to the niches found in other organs. Targeting these cells in the CNS offers new perspectives for treatment of chronic autoimmune neuroinflammatory diseases, especially in patients who do not respond to conventional therapies.

SILVA tree viewer: interactive web browsing of the SILVA phylogenetic guide trees.

BACKGROUND: Phylogenetic trees are an important tool to study the evolutionary relationships among organisms. The huge amount of available taxa poses difficulties in their interactive visualization. This hampers the interaction with the users to provide feedback for the further improvement of the taxonomic framework. RESULTS: The SILVA Tree Viewer is a web application designed for visualizing large phylogenetic trees without requiring the download of any software tool or data files. The SILVA Tree Viewer is based on Web Geographic Information Systems (Web-GIS) technology with a PostgreSQL backend. It enables zoom and pan functionalities similar to Google Maps. The SILVA Tree Viewer enables access to two phylogenetic (guide) trees provided by the SILVA database: the SSU Ref NR99 inferred from high-quality, full-length small subunit sequences, clustered at 99% sequence identity and the LSU Ref inferred from high-quality, full-length large subunit sequences. CONCLUSIONS: The Tree Viewer provides tree navigation, search and browse tools as well as an interactive feedback system to collect any kinds of requests ranging from taxonomy to data curation and improving the tool itself.

25 years of serving the community with ribosomal RNA gene reference databases and tools.

SILVA (lat. forest) is a comprehensive web resource, providing services around up to date, high-quality datasets of aligned ribosomal RNA gene (rDNA) sequences from the Bacteria, Archaea, and Eukaryota domains. SILVA dates back to the year 1991 when Dr. Wolfgang Ludwig from the Technical University Munich started the integrated software workbench ARB (lat. tree) to support high-quality phylogenetic inference and taxonomy based on the SSU and LSU rDNA marker genes. At that time, the ARB project maintained both, the sequence reference datasets and the software package for data analysis. In 2005, with the massive increase of DNA sequence data, the maintenance of the software system ARB and the corresponding rRNA databases SILVA was split between Munich and the Microbial Genomics and Bioinformatics Research Group in Bremen. ARB has been continuously developed to include new features and improve the usability of the workbench. Thousands of users worldwide appreciate the seamless integration of common analysis tools under a central graphical user interface, in combination with its versatility. The first SILVA release was deployed in February 2007 based on the EMBL-EBI/ENA release 89. Since then, full SILVA releases offering the database content in various flavours are published at least annually, complemented by intermediate web-releases where only the SILVA web dataset is updated. SILVA is the only rDNA database project worldwide where special emphasis is given to the consistent naming of clades of uncultivated (environmental) sequences, where no validly described cultivated representatives are available. Also exclusive for SILVA is the maintenance of both comprehensive aligned 16S/18S rDNA and 23S/28S rDNA sequence datasets. Furthermore, the SILVA alignments and trees were designed to include Eukaryota, another unique feature among rDNA databases. With the termination of the European Ribosomal RNA Database Project in 2007, the SILVA database has become the authoritative rDNA database project for Europe. The application spectrum of ARB and SILVA ranges from biodiversity analysis, medical diagnostics, to biotechnology and quality control for academia and industry.

FastaValidator: an open-source Java library to parse and validate FASTA formatted sequences.

BACKGROUND: Advances in sequencing technologies challenge the efficient importing and validation of FASTA formatted sequence data which is still a prerequisite for most bioinformatic tools and pipelines. Comparative analysis of commonly used Bio*-frameworks (BioPerl, BioJava and Biopython) shows that their scalability and accuracy is hampered. FINDINGS: FastaValidator represents a platform-independent, standardized, light-weight software library written in the Java programming language. It targets computer scientists and bioinformaticians writing software which needs to parse quickly and accurately large amounts of sequence data. For end-users FastaValidator includes an interactive out-of-the-box validation of FASTA formatted files, as well as a non-interactive mode designed for high-throughput validation in software pipelines. CONCLUSIONS: The accuracy and performance of the FastaValidator library qualifies it for large data sets such as those commonly produced by massive parallel (NGS) technologies. It offers scientists a fast, accurate and standardized method for parsing and validating FASTA formatted sequence data.

The SILVA and "All-species Living Tree Project (LTP)" taxonomic frameworks.

SILVA (from Latin silva, forest, http://www.arb-silva.de) is a comprehensive resource for up-to-date quality-controlled databases of aligned ribosomal RNA (rRNA) gene sequences from the Bacteria, Archaea and Eukaryota domains and supplementary online services. SILVA provides a manually curated taxonomy for all three domains of life, based on representative phylogenetic trees for the small- and large-subunit rRNA genes. This article describes the improvements the SILVA taxonomy has undergone in the last 3 years. Specifically we are focusing on the curation process, the various resources used for curation and the comparison of the SILVA taxonomy with Greengenes and RDP-II taxonomies. Our comparisons not only revealed a reasonable overlap between the taxa names, but also points to significant differences in both names and numbers of taxa between the three resources.

The SILVA ribosomal RNA gene database project: improved data processing and web-based tools.

SILVA (from Latin silva, forest, http://www.arb-silva.de) is a comprehensive web resource for up to date, quality-controlled databases of aligned ribosomal RNA (rRNA) gene sequences from the Bacteria, Archaea and Eukaryota domains and supplementary online services. The referred database release 111 (July 2012) contains 3 194 778 small subunit and 288 717 large subunit rRNA gene sequences. Since the initial description of the project, substantial new features have been introduced, including advanced quality control procedures, an improved rRNA gene aligner, online tools for probe and primer evaluation and optimized browsing, searching and downloading on the website. Furthermore, the extensively curated SILVA taxonomy and the new non-redundant SILVA datasets provide an ideal reference for high-throughput classification of data from next-generation sequencing approaches.

CDinFusion--submission-ready, on-line integration of sequence and contextual data.

State of the art (DNA) sequencing methods applied in "Omics" studies grant insight into the 'blueprints' of organisms from all domains of life. Sequencing is carried out around the globe and the data is submitted to the public repositories of the International Nucleotide Sequence Database Collaboration. However, the context in which these studies are conducted often gets lost, because experimental data, as well as information about the environment are rarely submitted along with the sequence data. If these contextual or metadata are missing, key opportunities of comparison and analysis across studies and habitats are hampered or even impossible. To address this problem, the Genomic Standards Consortium (GSC) promotes checklists and standards to better describe our sequence data collection and to promote the capturing, exchange and integration of sequence data with contextual data. In a recent community effort the GSC has developed a series of recommendations for contextual data that should be submitted along with sequence data. To support the scientific community to significantly enhance the quality and quantity of contextual data in the public sequence data repositories, specialized software tools are needed. In this work we present CDinFusion, a web-based tool to integrate contextual and sequence data in (Multi)FASTA format prior to submission. The tool is open source and available under the Lesser GNU Public License 3. A public installation is hosted and maintained at the Max Planck Institute for Marine Microbiology at http://www.megx.net/cdinfusion. The tool may also be installed locally using the open source code available at http://code.google.com/p/cdinfusion.