Enhancing clinical diagnostic analysis with a robust de novo mutation detection tool

PROJECT SUMMARY

This application proposes to supplement software development in our parent grant R01HG012286, entitled “Calypso: a web software system supporting team-based, longitudinal genomic diagnostic care”. We are developing Calypso to meet

diagnostic analysis needs in clinical settings where a large fraction of patients remain non-diagnostic for an extended period

of time, i.e. undiagnosed disease clinics, neonatal intensive care units, and pediatric subspecialty clinics. Our cloud-based

platform will provide the capacity for long-term storage and periodic automated reanalysis of the patient’s genomic data; a suite of intuitive IOBIO webtools will enable diagnostic analysis; and a case-focused communication and collaboration interface will coordinate diagnostic teamwork. However, even the best-orchestrated diagnostic variant analysis process

cannot succeed if the disease-causing variant remains undetected. Whereas established computational pipelines exist for

highly accurate and sensitive detection of inherited variations, current tools still underperform for detecting de novo disease- causing mutations, especially structural variant events. To address this bottleneck, we have developed a kmer-based

mutation detection software tool, RUFUS, and demonstrated its ability to substantially improve the detection of causative DNMs in a variety of diseases. In accordance with the aims of funding opportunity NOT-OD-22-068 “Enhancing Software

Tools for Open Science and the Cloud”, here we propose to enhance the impact of the currently research-grade RUFUS tool by improving its implementation and cloud-readiness to accelerate its adoption by the broader genomic medicine community. First, we will re-engineer the core RUFUS code base to produce a robust, production-ready, and easily

maintainable software package, without altering its already effective algorithmic behavior. We will replace RUFUS’s

currently ad hoc input/output handling with the de facto community standard HTSlib library; restructure logging to produce informative runtime messages; and implement automated code testing (both unit and integration testing) to ease future development. Second, we will enable cloud-native adoption of the RUFUS package which was originally designed to

operate in a Linux environment. We will improve scalability by adapting RUFUS for distributed computing, and thereby achieving a higher level of parallelization and execution speed than possible with the current, multi-threaded, implementation; and institute containerization to enable RUFUS’s incorporation into cloud-native runtime environments

and workflow language-base pipelines. Finally, third, we will enhance user and developer community engagement, by adopting standard versioning practices to provide the prerequisite software provenance for incorporation into clinical

diagnostic pipelines; enrolling our software into standard container registry services so users can easily find our tool; and expanding currently skeletal tool documentation to ease user adoption. Importantly, we will provide example nextflow workflows for RUFUS’s common use cases, together with representative datasets for each use case.