A study published August 12, 2026 in Nature Communications by researchers at Dana-Farber/Boston Children’s Cancer and Blood Disorders Center and Mass General Brigham suggests that adding targeted genetic sequencing to routine newborn screening could identify infants at elevated risk for early-onset cancer years before symptoms appear, opening a window for surveillance that currently does not exist.
The Study Design
The research team, led by Lisa Diller, M.D., Richard B. Parad, M.D., M.P.H., and Arindam Bhattacharjee, Ph.D., analyzed archived dried blood spots collected at birth from 1,948 Michigan children who were later diagnosed with solid tumors or brain cancers by age 8. Using sequencing technology and computational variant-classification pipelines, the team screened each sample for pathogenic or likely pathogenic variants across 11 well-established cancer-predisposition genes, including RET and RB1.
What the Data Showed
The results were striking: 132 of the children, or roughly 7%, carried a pathogenic or likely pathogenic variant that had been present since birth but went undetected by standard newborn screening. Extrapolated across the general population, the authors estimate that approximately 1 in 27,000 newborns could develop early-onset cancer that would be detectable through this kind of screening. Children who carried these mutations were diagnosed with cancer at a median age of 14 months, compared with 32 months for children without an identified variant. The effect was sharpest for specific genes: all six children with RET mutations were later diagnosed with medullary thyroid carcinoma, and 40% of retinoblastoma cases in the cohort carried RB1 mutations, with those carriers diagnosed at a median of 9 months versus 23 months for non-carriers.
Where AI Fits In
Manually reviewing genomic sequencing data for actionable variants across a large newborn population is not practical without computational triage. Automated variant-classification software, increasingly built with machine-learning components, is used to sort the millions of genetic variants each newborn’s genome contains down to the small number that are both pathogenic and clinically actionable. Companies including 3billion have already launched commercial genomic newborn screening services, such as 3B-NEO, covering nearly 600 clinically actionable genetic conditions, signaling that the infrastructure for population-scale genomic screening is maturing in parallel with the research.
Background: A Slow-Moving Shift in Newborn Screening
Routine newborn screening in the U.S., delivered via the heel-stick blood test given to nearly every infant, has for decades relied on biochemical assays rather than DNA sequencing, and cancer risk has never been part of that panel because it would require genomic analysis rather than a metabolic test. The Dana-Farber and Mass General Brigham findings are among the first to make a population-scale case, using real archived samples, that sequencing-based screening could meaningfully shift the age at which certain childhood cancers are caught.
Two Perspectives on Expanding Screening
Advocates for expanded genomic newborn screening argue that catching mutation carriers years earlier allows clinicians to institute vetted surveillance protocols, such as regular thyroid ultrasounds for RET carriers, that can catch tumors while they are still small and highly treatable. Parad noted that with collaboration between screening programs, geneticists and oncologists, preventive care can be provided before symptoms ever emerge.
Critics, including some clinical geneticists and bioethicists, caution that expanding newborn genetic testing raises difficult questions about consent, insurance discrimination, and the psychological burden on families told their healthy-appearing infant carries a cancer-risk mutation. There are also unresolved questions about false positives, since carrying a pathogenic variant does not guarantee a child will develop cancer, and lifelong surveillance protocols carry their own costs and anxieties.
What’s Next
The research team says larger, prospective studies are needed before genomic cancer screening could be added to state newborn screening panels, which are currently governed by a slow-moving federal advisory process. In the meantime, the study adds to a growing body of evidence, alongside pilot programs in Belgium and elsewhere, that computational genomics is inching toward becoming a standard part of the first days of life rather than a specialty tool used only after a diagnosis is already suspected.
The Michigan cohort was chosen in part because the state has maintained an archive of newborn dried blood spots for decades, giving researchers a rare opportunity to retrospectively test a screening approach against known clinical outcomes rather than waiting years for a prospective trial to mature. Bhattacharjee’s team noted that the computational pipeline used to classify variants had to be tuned specifically to distinguish genuinely pathogenic mutations from the much larger number of benign genetic variants every newborn carries, a task that becomes more tractable as reference databases of confirmed cancer-linked mutations continue to grow. The authors say they are now working with state public health laboratories to design a prospective pilot that would sequence newborns in real time rather than relying solely on retrospective archives, a step that would be necessary before any state health department could consider adding the test to its standard panel.
