We have established and tested a genomic and bioinformatic pipeline (GRACE — Germline Recognition, Annotation, and Classification Engine) to identify likely pathogenic genetic variants among patients with paediatric cancer and other rare diseases. These genomic data, while research based, can be reviewed by genetic counselors and validated for potential utility in patient care.

Such pipelines exist in many instituitions however they often fail to properly account for ancestral differences and founder mutations, both of which may be important for proper interpretation in Asian and other non-European populations. Our pipeline provides a QC approach that increases the potential for discovery in these often overlooked populations instead of screening them out. This addresses the issue that a one size fits all screening protocol does not adequately address the needs to many unique genomic groups.

The goal of this session is to engage with other groups to discuss the current implementation of the pipeline and to seek input on demographic, ancestral, and genomic attributes of populations in geographic regions that are not yet well represented in order to expand the pipeline to more properly identify pathogenic variants within those communities. Discussion will extend to how best to address isolated communities and familial studies for which reference comparison of variant allele frequencies are often unknown. The GA4GH community is invited to join a working group to identify genomic data sources to facilitate better estimate of non-disease genetic landscapes which would inform the interpretation of results. Lastly, we have vetted the use of AI tools (eg. Biomni, Google Co-Scientist, MARRVEL-AI) for genetic variant interpretation and will discuss our benchmarking experience. This session could link to other AI governance sections to develop larger protocols for better management and security of genomic data when using AI tools.