A major challenge in clinical genomics is the growing backlog of undiagnosed patients. While genomic sequencing has expanded rapidly, more than half of patients remain without a diagnosis after their initial test. Unlike many other diagnostic results, genomic data can be stored and reanalyzed as scientific knowledge evolves. However, manually reviewing thousands of previously analyzed genomes is labor-intensive, costly, and difficult for clinical laboratories to sustain.
Researchers from Australia and the US have developed an open-source tool called Talos to address this challenge. Published in Nature Medicine, the software automates genomic reanalysis by incorporating regularly updated gene-disease and variant evidence. Rather than requiring analysts to manually review every case, Talos prioritizes variants associated with newly established evidence, substantially reducing the amount of manual interpretation required.
Reducing the workload of genomic reanalysis
The researchers first validated Talos using genomic data from 1,089 individuals across Australia and the US. In trio analyses, the software identified approximately 90 percent of previously established diagnoses while returning a median of just 1.3 candidate variants per trio for manual review.
The team then applied Talos to a cohort of 4,735 children and adults who previously underwent genomic testing but did not receive a diagnosis. The automated reanalysis identified 241 additional diagnoses, representing a diagnostic yield of 5.1 percent. According to the researchers, more than one-third of these diagnoses resulted from newly established gene-disease associations that became available after the original testing.
Zornitza Stark, a clinical geneticist at Victorian Clinical Genetics Services, said the automated approach makes routine genomic reanalysis practical at scale by helping laboratories incorporate new scientific evidence into existing patient records. In the study, the median interval between new evidence becoming publicly available and a resulting diagnosis was 32 days, with some diagnoses made within a single day.
Open-source design lowers computational costs
The project was a collaboration among the Murdoch Childrens Research Institute, Victorian Clinical Genetics Services, the Centre for Population Genomics, the Broad Institute of MIT and Harvard, and Microsoft Research. The researchers designed Talos to run on standard computing infrastructure, allowing laboratories to deploy the software without specialized hardware.
Kaitlin Samocha, an associate member at the Broad Institute and assistant professor at Harvard Medical School, noted that reducing the number of variants requiring review helps clinical teams focus their efforts on the most relevant findings.
The computational costs reported in the study were relatively low. Running the initial workflow costs less than US$12 per 1,000 genomes, while ongoing monthly reanalysis costs less than US$2 annually per 1,000 genomes. These estimates reflect cloud computing expenses and do not include implementation, validation, integration, or personnel costs. Jeremiah Wander, a principal researcher at Microsoft Research, said the open-source software is fully auditable and compatible with standard computing environments.
Implications for laboratory operations
For laboratory leaders overseeing genomic testing programs, automated reanalysis offers a potential strategy for making better use of existing sequencing data without proportionally increasing manual review efforts. As new gene-disease relationships continue to be discovered, laboratories face growing pressure to revisit previously analyzed cases while maintaining current testing workloads.
The authors suggest that automated workflows such as Talos could help laboratories perform more consistent genomic reanalysis by incorporating updated scientific evidence into routine operations while substantially reducing the number of variants requiring manual interpretation. By streamlining retrospective analyses, laboratories may be able to devote more staff time to complex case review and active diagnostic testing while improving the long-term value of stored genomic data.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
