Research project IDEA-PRIO
The Modeling and Simulation Group at the University of Rostock and Limbus Medical Technologies GmbH working together to enhance the interpretation of genetic variants using artificial intelligence (AI) methods
Our Goal:
Design and implementation of an adaptive, scalable and efficient mechanism for variant prioritisation based on heterogeneous clinical data to dramatically improve diagnostic yield, greatly reduce the time required for the diagnostic process, and shorten test turnaround time.
It is generally expected that next-generation sequencing technology (NGS) will lead to more precise, faster, cheaper and overall more efficient and universal diagnostic methods for rare diseases. An essential step in the diagnosis of these diseases is the evaluation of genetic variants. This involves identifying the few variants that are relevant for medical diagnosis from the large number of genetic variants found. To support this process, variant prioritisation methods are used. They use curated gene-phenotype or gene-disease associations to evaluate the relevance of the variants. However, these methods have significant limitations with respect to the explainability of the prioritisation results and the use of additional data sources. Therefore, this project aims to develop a self-adaptive system that uses different medical and clinical data to incrementally improve the evaluation of gene variants. For this purpose, approaches from the area of the explanability of machine learning processes, workflows and provenance are evaluated, combined and, if necessary, adapted. In addition, methods from the field of modeling and simulation are used to develop and evaluate the system.
Missing explanability
The methods don't offer explanations neither for the prioritisation itself nor information on their data bases.
Focus on a single case
The methods may not consider similar cases and are based on databases with curated genotype-phenotype- and genotype-disease-associations.
Missing feedback from clinicians
The methods do not consider direct user feedback.
Static prioritization
The methods are initially trained and validated, changes need manual adaption.
The goal is to design and implement an enhanced variant prioritisation mechanism that:
delivers explainable prioritizations
The clinician is able to follow and understand the reasoning behind the AI-based variant interpretation.
is based on (daily) updated and comprehensive clinical data
The software plattform varvis® stores clinical cases including details on phenotype and variants.
is continuously improved by new diagnoses
and uses the best available machine learning method
The software plattform varvis® is cloud-based. The performance of the mechanism can be automatically monitored and replaced by an improved version.
Cooperation Partners
Modeling and Simulation Group,
University of Rostock
The modeling and simulation group at the University of Rostock, led by Prof. Dr. Adeline M. Uhrmacher, conducts methodological research on a broad range of topics (e.g., multi-agent modeling, multi-level modeling, model validation, spatial simulation) in various application areas (e.g., computational biology, autonomous software systems, ecology, and demography). She and her group have decades-long expertise in designing, developing, and maintaining complex open-source software systems, such as the modeling and simulation framework JAMES II, the modeling language ML-Rules, or the experiment language SESSL.
Limbus Medical Technologies
Limbus Medical Technologies, founded in 2015 in Rostock, Germany, is a medical device manufacturer and software development company. Its cloud-based software varvis® is a clinical decision support system to filter and evaluate genetic sequencing data. Focusing on clinical diagnostics, it is a complete solution for all clinical NGS applications, such as whole exome sequencing, cancer diagnostics or NGS panels, supporting NGS raw data processing, genomics data management, and variant interpretation. varvis® is a registered CE-IVD device according to directive 98/79/EC and as such made for use in clinical routine diagnostics.
team
Dr. Roland Ewald
Limbus Medical Technologies
Dr. Lena Hausdorf
Limbus Medical Technologies
Dr. Tobias Helms
Limbus Medical Technologies
Dr. Yvonne Kasmann
Limbus Medical Technologies
Dr. Stefan Leye
Limbus Medical Technologies
M.Sc. Tom Meyer
University Rostock
M.Sc. Andreas Ruscheinski
University Rostock
Dr. Stefan Rybacki
Limbus Medical Technologies
Publications
VPMBench: a test bench for variant prioritizing methods
Ruscheinski, A., Reimler, A.L., Ewald, R. et al. VPMBench: a test bench for variant prioritization methods. BMC Bioinformatics 22, 543 (2021).
Clinical diagnostics of whole-exome and whole-genome sequencing data requires geneticists to consider thousands of genetic variants for each patient. Various variant prioritization methods have been developed over the last years and each time a new method is developed, its effectiveness must be evaluated and compared to other approaches based on the most recently available evaluation data. Doing so in an unbiased, systematic, and replicable manner requires significant effort.
The open-source test bench “VPMBench” automates the evaluation of variant prioritization methods. VPMBench introduces a standardized interface for prioritization methods and provides a plugin system that makes it easy to evaluate new methods. It supports different input data formats and custom output data preparation. VPMBench exploits declaratively specified information about the methods, e.g., the variants supported by the methods. Plugins may also be provided in a technology-agnostic manner via containerization.