2024 MCBK North American chapter meeting—Lightning talk and demonstration abstracts

Abstract

POSTERS

DEMONSTRATIONS

Saketh Boddapati, University of Michigan College of Literature, Science, and the Arts

[email protected]

Yongqun “Oliver” He, University of Michigan Medical School

[email protected]

Healthcare providers learn continuously as a core part of their work. However, as the rate of knowledge production in biomedicine increases, better support for providers' continuous learning is needed. Tools for learning from clinical data are widely available in the form of clinical quality dashboards and feedback reports. However, these tools seem to be frequently unused.

Making clinical data useful as feedback for learning appears to be a key challenge for health systems. Feedback can include coaching, evaluation, and appreciation, but systems developed for performance improvement do not adequately recognize these purposes in the context of provider learning. Moreover, providers have different information needs, motivational orientations, and workplace cultures, all of which affect the usefulness of data as feedback.

To increase the usefulness of data as feedback, we developed a Precision Feedback Knowledge Base (PFKB) for a precision feedback system. PFKB contains knowledge about how feedback influences motivation, to enable the precision feedback system to compute a motivational potential score for possible feedback messages. PFKB has four primary knowledge components: (1) causal pathway models, (2) message templates, (3) performance measures, and (4) annotations of motivating information in clinical data. We also developed vignettes about 7 diverse provider personas to illustrate how the precision feedback system uses PFKB in the context of anesthesia care. This ongoing research includes a pilot study that has demonstrated the technical feasibility of the precision feedback system, in preparation for a trial of precision feedback in an anesthesia quality improvement consortium.

Bruce Bray, University of Utah, on behalf of the HL7 Learning Health Systems Work Group

[email protected]

Data is the lifeblood of computable biomedical knowledge (CBK) and must adhere to standards to achieve the interoperability needed to generate virtuous learning cycles within a learning health system (LHS). The HL7 Learning Health System Work Group (HL7 LHS WG) conducted a scoping review to compile an initial list of standards that can support the LHS across “quadrants” of a virtuous learning cycle: (1) knowledge to action, (2) action to data, (3) data to evidence, and (4) evidence to knowledge. We found that few standards explicitly refer to an overarching framework that aligns interoperability and data standards across the phases of the LHS. We will describe our initial work to identify relevant gaps and overlaps in standards in this environment. Future work should address standards coordination and pilot testing within an LHS framework. These efforts can enhance collaboration among communities such as MCBK and HL7 to promote standards-based computable knowledge mobilization.

Allen Flynn, University of Michigan, on behalf of the Knowledge Systems Lab, Department of Learning Health Sciences

[email protected]

The Knowledge Object (KO) is a modular, extensible digital object, designed to allow computable biomedical knowledge (CBK) to be managed as a resource and implemented as a service. This poster describes how the KO model has evolved to better support the FAIR principles by:

Enabling multiple services to increase interoperability and reuse.

KOs were originally developed to run with an Activator, which loaded and deployed KOs on request, exposed the service, and routed responses as a RESTful API. While the intent was to facilitate low-friction KO implementation, the Activator could limit interoperability and the potential for reuse. Expanding the model to reduce reliance on the Activator and allow for multiple services means KOs can meet a wider range of stakeholder needs. Current work includes developing updated specifications and reference implementation for activation.

Enabling multiple implementations of knowledge and services to increase interoperability and reuse. The legacy KO model included one implementation of both the CBK payload and the service to activate it. Engineering a KO to contain multiple implementations of CBK and services means KOs can meet a wider range of stakeholder needs. We are updating our model and engineering KOs capable of carrying multiple implementations and services.

Improving findability, accessibility interoperability, and reuse through an updated model and metadata. Metadata for the legacy KO model primarily described the service. Using the new model, we are now developing standards-based extensible Linked Data metadata to describe the KO, the knowledge it contains, and the services that “activate” that knowledge.

Nicole Gauthreaux, NORC at the University of Chicago

[email protected]

Courtney Zott, NORC at the University of Chicago

[email protected]

Prashila Dullabh, NORC at the University of Chicago

[email protected]

This poster is relevant to clinicians, health system leaders, informaticians, and researchers interested in driving future mobilization of computable knowledge for patient-centered clinical decision support (PC CDS). We conducted a cross-cutting synthesis of real-world PC CDS projects to identify the types of measures used, measurement challenges and limitations, and action steps to advance PC CDS measurement. We reviewed research products from 20 PC CDS projects funded by the Agency for Healthcare Research and Quality (AHRQ) to gather information on their studies, and we conducted key informant interviews with Principal Investigators of nine projects to gather their experiences and challenges with PC CDS measurement.

Findings from the synthesis revealed a considerable focus on measuring the effectiveness of the PC CDS, primarily by collecting patient and clinician perspectives on the usability and acceptability of the tool and observing patient health outcomes from the intervention. Many projects incorporated patient perspectives in their study, yet there were more process measures (e.g., patient satisfaction with the design) than outcome measures (e.g., patient activation to manage their health due to the PC CDS). Few projects measured safety, or the technical performance and information presented by the PC CDS technology. Finally, equity measures rarely extended beyond descriptive analyses of participant socio-demographics. Key informants described other evaluation challenges related to patient recruitment, technical limitations, and imprecision in data collection specific to PC CDS interventions. These findings provide a basis for guiding future development of measures that promote the adoption and use of knowledge for patient-centered care.

Pawan Goyal, American College of Emergency Physicians

[email protected]

Data is driving the future of medicine. We've already seen the critical importance of real-time insights to new and emerging health threats during the COVID-19 pandemic, as well as the impact of health care trends and patterns of resource utilization. With the new Emergency Medicine Data Institute (EMDI), the American College of Emergency Physicians (ACEP) is rapidly moving emergency medicine to the forefront of data-driven quality and practice innovation. This new initiative is poised to become a central source of intelligence and knowledge generation across all emergency medicine stakeholders. Harnessing the power of information that physicians are already recording, ACEP is synthesizing and standardizing data across multiple billing and EHR environments, innovating new research, and pursuing national-level grants all while enhancing value for emergency physicians, patients, and the broader health care community.

Indika Kahanda, PhD, University of North Florida

[email protected]

The quest for mobilizing computable biomedical knowledge, inconsistencies, and contradictions in biomedical literature would be a significant barrier. Given the exponential growth of scientific information, researchers often face the daunting task of detecting contradictory statements on crucial health topics. This work proposes to develop a full-fledged, trustworthy automated pipeline for explainable contradiction detection, which will integrate an Information Retrieval (IR) system backed by a local data store, predictive models, and an Explainable AI (XAI) component. Users can input queries on medicine and health topics, and the system will identify top documents and sentences through syntactic analysis and refine results via semantic examination for relevant research claims. These sentences are forwarded to the predictive models backed by Large Language Models, which will classify each pair as contradictory. The XAI component will help output visual explanations based on these predictions. We have used ManConCorpus, a popular biomedical contradiction corpus on cardiovascular diseases, to develop and evaluate our predictive models. The preliminary results demonstrate that PubMedBERT, with an F1 score of 97%, can outperform BioBERT, Bioformer, and Distil-BERT in classifying a given pair of sentences relevant to cardiovascular disease as contradictory or not. Further investigation is necessary to ensure that the models are robust in performing similarly on any health and medical topic. In the future, these predictive models will be combined with the aforementioned IR/XAI components for developing the prototype pipeline. This study has implications for medical and healthcare practitioners, researchers, students, systematic review authors, and the biomedical text-mining community.

Zach Landis-Lewis, University of Michigan, Department of Learning Health Sciences

[email protected]

Healthcare providers learn continuously as a core part of their work. However, as the rate of knowledge production in biomedicine increases, better support for providers' continuous learning is needed. Tools for learning from clinical data are widely available in the form of clinical quality dashboards and feedback reports. However, these tools seem to be frequently unused.

Making clinical data useful as feedback for learning appears to be a key challenge for health systems. Feedback can include coaching, evaluation, and appreciation, but systems developed for performance improvement do not adequately recognize these purposes in the context of provider learning. Moreover, providers have different information needs, motivational orientations, and workplace cultures, all of which affect the usefulness of data as feedback.

To increase the usefulness of data as feedback, we developed a Precision Feedback Knowledge Base (PFKB) for a precision feedback system. PFKB contains knowledge about how feedback influences motivation, to enable the precision feedback system to compute a motivational potential score for possible feedback messages. PFKB has four primary knowledge components: (1) causal pathway models, (2) message templates, (3) performance measures, and (4) annotations of motivating information in clinical data. We also developed vignettes about 7 diverse provider personas to illustrate how the precision feedback system uses PFKB in the context of anesthesia care. This ongoing research includes a pilot study that has demonstrated the technical feasibility of the precision feedback system, in preparation for a trial of precision feedback in an anesthesia quality improvement consortium.

Siddharth Limaye, Carle Illinois College of Medicine

[email protected]

Understanding causation is an essential element of medical education. While a plethora of research exists regarding the use of knowledge graphs for topics such as drug discovery and personalized medicine, there is relatively less work regarding (1) the use of these graphs for medical education and (2) the use of causal graphs for these purposes. Pathology Graphs is a proof-of-concept application that uses a typed database to represent the causal model underlying several physiologic and pathophysiologic processes in the human body. Effectively, Pathology Graphs is an interactive, digital revision of the concept map.

While graph databases are suitable for building concept maps and causal graphs, Pathology Graphs instead uses a typed database to implement a higher-order graph structure. This change allows Pathology Graphs to represent certain types of relationships (e.g., an enzyme modifying a reaction) which would otherwise not be simply expressible in a graph database. This change also allows inference of the downstream effects of a change in the pathology graph, such as predicting that product will decrease if an enzyme inhibitor is added to the system. This capability could be useful for further functionality, such as automatically generating multiple-choice questions and answers for students or allowing students to generate differential diagnoses by examining all the potential causes of a finding. Challenges remaining to be addressed for this application include data input, user interface, and output visualization, as the goal of this project is to be accessible for those without pre-existing programming knowledge.

Aswini Misro, YouDiagnose Limited

[email protected]

Background—AI is being incorporated into healthcare by major tech companies, but public acceptance remains challenging. The study aims to understand resistance to AI despite its increasing accuracy and potential to improve patient waiting times.

Method—In partnership with the University of Hull and the Academic Health Science Network (AHSN) in the UK, a study was conducted that involved 111 adult patients or carers. 9 did not respond while 28 demonstrated poor digital literacy and were therefore excluded. An interactive user page was designed for the remaining users (n = 74) to engage with a medical chatbot. The participants were asked whether participants would use an AI-powered medical chatbot or nurse for triage at their nearest A&E. Unstructured open-ended interviews were conducted to understand the participants' reasoning behind their answers.

Finding – Participants ranged in age from 21 to 74, with a slight female majority (40 female vs. 34 male). The majority, 79.8% (n = 59) respondents, expressed their comfort with a nurse-led care while a mere 20.2% (n = 15) showed readiness to interact with a medical chatbot. Notably, the age group of 20–40 was most open to the idea of consulting a chatbot. The four primary objections were: 90.5% (n = 67) believed that the chatbot fails to justify its decisions, 59.5% (n = 44) doubt its accuracy, 78.4% (n = 58) felt that the chatbot was inflexible, and 79.7% (n = 59) found it unemotional and detached.

Conclusion—Effective treatment relies on trust, respect, and understanding crucial elements needed when incorporating AI in medicine. It should be introduced progressively, considering patients' emotions and societal circumstances.

Jerome Osheroff, TMIT Consulting LLC/University of Utah/VA

[email protected]

Dave Little (Epic) Stephanie Guarino (Nemours, ChristianaCare) Teresa Cullen (Pima County Health Dept) Rosie Bartel (Patient Partner) Joshua E. Richardson (RTI International) For the POLLC and SCD Learning Communities

Understanding gaps in care is essential for clinical quality improvement efforts. The Pain Management/Opioid Use LHS Learning Community (POLLC), an initiative that engages care delivery organizations (CDOs) and other stakeholders to collaboratively accelerate quality improvement efforts and results, has been working since 2022 to develop and implement a care gap report for this target. Participants identified 12 important potential care improvement opportunities to assess appropriate long-term opioid use for chronic pain, for example, high opioid doses, multiple emergency department visits, and no Prescription Drug Monitoring Program Check. In March 2023, Epic released a pilot Care Gap Query that uses SQL code to identify patients meeting these criteria; it requires an Epic analyst with SQL expertise to implement and modify the query. In November 2023, Epic released a Care Gap Report (CGR) using a Reporting Workbench Template that enables Epic users without special expertise to run and configure the report. It enables users to take bulk action on report results, for example, placing orders and triggering communications. POLLC participants are exploring opportunities to aggregate care gap results from individual CDOs in a region into “population CGRs” that can be used to inform and guide public health interventions. A parallel learning community on sickle cell disease (SCD) that grew out of POLLC is developing analogous care gap reports for SCD in Epic and Oracle EHRs. Approaches are being explored to leverage interoperable, computable biomedical knowledge to make creating and deploying CGRs across targets, EHRs, and CDOs faster and more efficient.

Henrique Santos, Rensselaer Polytechnic Institute

[email protected]

Paulo Pinheiro, Instituto Piaget

[email protected]

Deborah L. McGuinness, Rensselaer Polytechnic Institute

[email protected]

Many countries perform surveys to gather data from their population for supporting decision-making and development of public policies. Questionnaires are possibly the most used type of data acquisition instrument in surveys, although additional kinds may be employed (especially in health-related surveys). In the United States, the NHANES is a national health and nutrition examination survey conducted by the National Center for Health Statistics, designed to collect data on adults and children's health and nutritional status. Data is organized in several tables, each containing variables to a specific theme, such as demographics, and dietary information. In addition, data dictionaries are available to (sometimes partially) document the tables' contents. While data is mostly provided by survey participants, instruments might be collecting data related to other entities (e.g., from participants' households and families, as well as laboratory results from participants' provided blood and urine samples). All this complex knowledge can often only be elicited by humans when analyzing and understanding the data dictionaries in combination with the data. The representation of this knowledge in a machine-interpretable format could facilitate further use of the data. We detail how Semantic Data Dictionaries (SDDs) have been used to elicit knowledge about surveys, using the publicly available NHANES data and data dictionaries. In SDDs, we formalize the semantics of variables, including entities, attributes, and more, using terminology from relevant ontologies, and demonstrate how they are used in an automated process to generate a rich knowledge graph that enables downstream tasks in support of survey data analysis.

Deborah Swain, North Carolina Central University

[email protected]

Following MCBK pilot training in 2021–22 supported by an Institute of Museum and Library Services (IMLS) grant, we designed and developed an open educational resource (OER) platform to be accessible and sustainable for global users. Students and development partners included librarians in medical libraries and informatics graduate students in the United States and Canada. However, international applicants were unable to participate. The OER collection provides full access.

The concept of an open pedagogy for resources supports open commons and sharing in the larger MCBK community. Policies for OER allow open content and open education practices. In the future, new material and potential courses and textbooks can become part of this MCBK OER collection of documents and slides.

Currently, the MCBK resources are hosted in the North Carolina Digital Online Collection of Knowledge (NC Docks): https://libres.uncg.edu/ir/nccu/clist.aspx?id=41690. Technical support is provided by UNC-Greensboro and NCCU Research and Instructional Services Librarian, Danielle Colbert-Lewis.

[See references and special contributors.]

Deborah Swain, North Carolina Central University

[email protected]

Amrit Vastrala, North Carolina Central University

[email protected]

Bias in AI (artificial intelligence) and ML (machine learning) refers to the systematic errors that can occur in training data, algorithms, or models and lead to discriminatory or unfair outcomes for groups of people. This bias may be deliberate or inadvertent, and it may result from a number of factors including data collections, pre-existing social biases, improper algorithm design, and ethical motivation of XAI (explainable AI).

The effects of bias can range from maintaining social injustices to making unreliable or inaccurate predictions to patients. The methodologies and frameworks for practice that biomedical researchers and health practitioners have proposed include fairness metrics, pre-processing methods, post-processing strategies to detect and mitigate bias, and models. Both model-level explanations for providers and prediction-level explanations for users/patients have been researched. This poster summarizes evidence-based research and thoughtful recommendations for key stakeholders. Recognizing recent research and reference publications are the primary objective of our literature review and interviews.

Bias from AI is a difficult problem to solve due to complexity and nuance. There is still a lot of work to be done to ensure that biomedical systems are fair and transparent. Everyone involved can address bias in AI/ML and help develop best practices for building responsible and trustworthy systems. Ongoing research and collaboration among experts in ethics, social science, and healthcare will be essential.

Yujia Tian, University of Michigan

[email protected]

Although the etiology of most mental illnesses remains unclear, it is believed that they could be caused by a combination of genetic, social factors, and personal characteristics. The complex clinical manifestations of mental illnesses pose challenges in medical diagnosis. Many infrastructural models exist to support classification of mental illnesses. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) serves as the principal authority for psychiatric diagnoses by the American Psychiatric Association (APA). The International Statistical Classification of Diseases and Related Health Problems (ICD) also provides a taxonomic and diagnostic classification of mental illnesses. However, DSM-5 and ICD still struggle with accurate diagnosis with heterogeneity and comorbidity. In contrast, the Hierarchical Taxonomy Of Psychopathology (HiTOP) system offers a more continuous perspective of the disease spectrum. The Research Domain Criteria (RDoC) framework focuses on dimensions of behavioral/psychological functioning and their implementing neural circuits. Ontology brings a new direction to the future research of psychiatry. The Human Phenotype Ontology (HPO) standardizes various mental illnesses. HPO is also increasingly integrated with genomic data, enhancing its utility in clinical diagnostics and personalized medicine. We plan to integrate the advantages of existing infrastructure models in psychiatry into HPO and fill up missing aspects in the ontology. Ultimately, we can combine the concept of a learning health system (LHS) to create a medical environment that continuously learns and adapts, using big data and machine learning technology to optimize treatment strategies and improve the level of precision medicine.

Arlene Bierman, AHRQ

[email protected]

David Carlson, Clinical Cloud Solutions

[email protected]

Jenna Norton, NIDDK

[email protected]

Evelyn Gallego, EMI Advisors

[email protected]

Stanley Huff, MD, Graphite Health

[email protected]

Part 1: I will be demonstrating a new LOINC Ontology. The LOINC Ontology is being made available as a SNOMED CT extension. The LOINC team at Regenstrief Institute and SNOMED International have a new agreement to make all LOINC content available in a new SNOMED CT extension. The creation of the new ontology is proceeding in a step-wise fashion. The first content in the ontology is 24 000 quantitative laboratory tests. Information about the new ontology can be found at https://loincsnomed.org/ and a browser for the content can be found at https://browser.loincsnomed.org/.

Part 2: The new ontology allows sematic reasoning on LOINC content, a capability that has been lacking in previous releases of the LOINC terminology. For example, using the SNOMED Expression Constraint Language (ECL), you can easily identify the codes that represent fasting glucose levels regardless of the method used for the test. Future releases will allow deeper reasoning, for example, about what kinds of parenteral antibiotics are available for a particular kind of bacteria identified by culture.

Preston Lee, PhD, MBA, FAMIA – Skycapp

[email protected]

Adela Grando, PhD, FACMI, FAMIA – Arizona State University

[email protected]

Part 1: This Skycapp software delivery platform demo will show how a complex NIH-funded CDS Hooks service developed at Arizona State University (ASU) can be widely disseminated and deployed in an automated manner to worldwide evaluators and adopters. The system and approach are generalized to all “level 4” CBK types through strict adherence to FHIR (and other) data standards and infrastructural interoperability.

Part 2: Skycapp's implementation of post-publication CBK delivery and deployment is based on the balloted HL7/Logica Marketplace 2 STU 2 specification. We have purposefully designed the platform to provide a “publish → deploy → adopt” model of dissemination enabling adopters to evaluate artifacts in local context prior to deciding to pursue their adoption, thus encouraging CDS experimentation by deferring any sizable commitments of time or money.

Eric Mercer, Brigham Young University, Computer Science

[email protected]

Bryce Pierson, Brigham Young University, Computer Science

[email protected]

Keith A. Butler, University of Washington

[email protected]

Part-2: FAIR Principles Relevance

Findable: Complex HIT designs could be indexed for search engine discovery by the cognitive work problem they were proven and certified to solve.

Accessible: Our overall aim is automated translation back and forth between the concepts and languages of the design community and those of model checking, thereby making model checking far more accessible and usable for provider participation in HIT design.

Interoperable: The BPMN standard [1] is widely adopted and available in dozens of commercially supported modeling products. BPMN models can be exported as XML files, thereby carrying forward the conceptual design requirement onto implementation platforms.

Reusable: The finite state machine for the cognitive work problem that a workflow design must solve can be reused for model checking multiple HIT designs that purport to solve the same cognitive problem. Certification means they each can solve the identical problem, yet may differ widely in their qualities of usability, function allocation to human vs. computing, cost to develop/deploy, timeliness, etc.

The important, related principle of trustworthiness is also increased by model checking certificates that verify all sequences of an HIT design are correct.

Sabbir Rashid, Rensselaer Polytechnic Institute

[email protected]

Deborah McGuinness, Rensselaer Polytechnic Institute

[email protected]

A clinical decision support system (CDSS) can support physicians in making clinical decisions, such as differential diagnosis, therapy planning, or plan critiquing. To make such informed decisions, there may be a large amount of medical data and literature that a physician needs to keep track of such as new research articles, pharmacological therapies, and updates in Clinical Practice Guidelines. Therefore, a CDSS can be designed to assist physicians by providing relevant, evidence-based clinical recommendations, subsequently reducing the mental overhead required to keep up to date with an evolving body of literature. We designed a CDSS by leveraging Semantic Web technologies to create an AI system that reasons in a way similar to physicians. We base our abstraction of human reasoning on the Select and Test Model (ST-Model), which combines multiple forms of reasoning, such as abstraction, deduction, abduction, and induction, to arrive at and test hypotheses. Based on this framework, we perform ensemble reasoning, the integration and interaction of multiple types of reasoning. We apply our CDSS to the treatment of type 2 diabetes mellitus by designing a domain ontology, the Diabetes Pharmacology Ontology (DPO), that supports both deductive and abductive reasoning. DPO is additionally used to provide a schema for our knowledge representation of hypothetical patients, where each patient is encoded in RDF as a Personalized Health Knowledge Graph (PHKG). We build our system using the Whyis knowledge graph framework by writing software agents to perform custom deductive reasoning and integrate the performance of abduction using an existing reasoning engine, the AAA Abduction Solver. We apply our approach to perform therapy planning on the hypothetical patients, which we will showcase as a part of the demonstration of our system.

The use of semantic technologies allows us to leverage existing reasoning engines, both in terms of deductive and abductive reasoning, use formal knowledge representations, such as clinical ontologies and vocabularies, and incorporate existing techniques for capturing provenance, such as nanopublications. Additionally, our approach allows for the generation of justifications for the reasoning choices made. Furthermore, this work promotes the FAIR principles and guidelines that have been widely adopted and cited for publishing data and metadata on the web. For the ontology to be findable, globally unique and persistent identifiers are created for each resource in the ontology. Concepts are directly accessible from their URL and the ontology itself is directly accessible via the resource URL defined in the ontology. To promote interoperability, we link concepts in our ontology to other standard vocabularies, including LOINC, ChEBI, Symptom Ontology, and NCIT. Finally, to promote the reusability of our resource, we have published, made readily available, and adequately documented the ontology, PHKGs, and software that we use for our CDSS. The demonstration will show our hybrid reasoning clinical decision support system in action in a diabetes setting.

Farid Seifi, Knowledge Systems Lab, University of Michigan

[email protected]

Anurag Bangera, Knowledge Systems Lab, University of Michigan

[email protected]

Additionally, we have improved on our original metadata model and will discuss how standards-based Linked Data metadata improves the Findability, Accessibility, Interoperability, and Reusability of CBK packaged as KOs.

Legacy KOs could only be run as part of a RESTful web service that can be called by other systems. Additionally, the legacy KO model allowed for only one service and implementation. In contrast, the enhanced model allows for multiple services within a single KO, and multiple implementations of the same knowledge and/or services, each in a different programming language. In this demo, we will show how the files, metadata, code, and other information needed to a variety of technical paths can be packaged together inside a single compound digital Knowledge Object that has the potential to support reuse by a variety of people with different professional roles.

The enhanced KO model can package computable biomedical knowledge in a technically variform way so that a wider variety of stakeholders can more quickly and easily use it. By offering many technical paths to deploying and using the same computable knowledge, CBK artifacts can be used in multiple different contexts, increasing interoperability and reusability. Enabling various technical paths to using the same CBK provides different ways for application developers, system integrators, CBK evaluators, data analysts, and others to reuse CBK in ways that meet existing and emerging needs.

Mitchell Shiell, Ontario Institute for Cancer Research

[email protected]

Describe the computable biomedical knowledge (CBK) you will demonstrate:

Next-generation sequencing has made genomics datasets commonplace, posing new challenges for research groups who want to efficiently gather, store, and share their data, while maximizing its value and reuse. This creates a compelling case for new computational tools to mobilize these massive datasets at scale. Overture is a suite of open-source and free-to-use modular software that works in concert to build and deploy scalable genomics data platforms. These platforms streamline the gathering, organizing, and sharing of raw and interpreted data, making it accessible for both humans and machines to translate into knowledge.

Our MCBK demo will highlight how Overture creates data resources that broadly achieve FAIR data goals. We will demonstrate how our core microservices—Ego, Song, Score, Maestro, and Arranger—achieve these data goals with a presentation and practical demonstration of the Overture platform.

Describe how your CBK promotes the FAIR principles and/or trust:

Overture is comprised of five core components that each provide a foundation for mobilizing discoverable, FAIR (Findable, Accessible, Interoperable, and Reusable) genomics data. (1) Ego, Overture's identity and permission management service, enables accessibility with appropriate authentication and authorization procedures using standard and free protocols. (2) Song and (3) Score work together to support findability with data submission, management, and retrieval methods. These services significantly increase data quality, findability, and interoperability with automated tracking and custom metadata validations. (4) Maestro indexes data from a distributed network of Song metadata repositories into a unified Elasticsearch index, and (5) Arranger then uses this index to produce a graphQL search API that can be extended with a library of configurable search and portal UI components. Combining these services completes a comprehensive end-to-end data portal that broadly enables the secure, scalable reuse of genomics data. Overture aims to make large-scale genomics data FAIR and cost-effective for researchers worldwide, fostering mobilization and collaboration over data globally.