Dr. Francesca Rosamilia is a biologist with a PhD in Biostatistics from the University of Genoa (2024), specializing in human genomics and complex diseases. Her doctoral research combined WES, GWAS, and proteomic data to investigate the genetic basis of Hirschsprung disease. During her PhD, she was a Visiting PhD Student at Columbia University Medical Center in New York. She is currently a visiting researcher at FIMM, University of Helsinki, where she works on polygenic risk scores and large-scale biobank data.
Predictive utility of childhood polygenic scores for adult disease risk and inform early-life prevention
Rosamilia Francesca1,2, Fiorito Giovanni1, Vartiainen Pekka2, Tantari Giacomo3, Rubinacci Simone2, Uva Paolo1, Ceccherini Isabella4
1Clinical Bioinformatic Unit, IRCCS Giannina Gaslini Institute, Genoa, Italy; 2Institute for Molecular Medicine Finland, Helsinki, Finland, 3Pediatric Clinic, IRCCS Giannina Gaslini Institute, Genoa, Italy, 4UOSD Research Laboratories Aggregation Area, IRCCS Giannina Gaslini Institute, Genoa, Italy
Background: Many adult diseases trace their origins to childhood, yet whether pediatric genetic risk profiles can effectively predict adult-onset conditions remains poorly characterized.
Study question: We investigated if pediatric polygenic scores (PGS) can identify individuals at increased risk for adult-onset diseases, revealing shared genetic architecture throughout life.
Methods: We constructed 73 PGS for 30 pediatric traits using GWAS summary statistics and the PGS Catalog. We tested associations between pediatric PGS and 25 adult phenotypes in FinnGen and UK Biobank (replication phase). Linkage disequilibrium score (LDSC) regression and multi-trait conditional joint analysis (mtCOJO) were used to distinguish genetic pleiotropy from effects mediated by genetically related traits.
Main results: We identified 27 significant pediatric–adult associations in FinnGen, primarily linking childhood autoimmune genetic risk to adult cardiometabolic and autoimmune phenotypes. These associations robustly replicated in the UK Biobank. Notably, the pediatric type 1 diabetes PGS strongly predicted adult rheumatoid arthritis, with individuals in the highest PGS decile showing a 1.8-fold increased risk compared with those in the lowest decile (OR=1.88, 95% CI=1.76–2.01; P=1.09×10⁻⁷⁵). LDSC and mtCOJO analyses confirmed that these overlapping risks represent direct genetic links rather than mediation by confounding traits.
Limitations: Analyses were limited to individuals of European ancestry and to currently available pediatric GWAS studies, limiting the transferability of results.
Wider implications: Systematic evaluation of the ability of the polygenic genetic score (PGS) derived from childhood traits aims to improve the prediction of complex diseases in adulthood by adding information beyond the adult-derived PGS and clinical risk factors.