Teunis, GijsDallman, Timothy J.Zajac, MagdalenaSkarżyńska, MagdalenaPetrovska, LiljanaPista, ÂngelaSilveira, LeonorClemente, LurdesThépault, AmandineBonifait, LaetitiaKerouanton, AnnaelleChemaly, MarianneAlvarez, JulioSoderlund, RobertNielsen, Eva MollerChattaway, MarieBurgess, KayeByrne, WilliamZomer, Aldert L.van den Beld, MaaikeHendrickx, Antoni P.A.Franz, EelcoPires, SaraHald, TineMughini-Gras, Lapo2026-01-212026-01-212025-10-13J Infect. 2025 Nov;91(5):106632. doi: 10.1016/j.jinf.2025.106632. Epub 2025 Oct 130163-4453http://hdl.handle.net/10400.18/10733Non-typhoidal Salmonella is the second most frequently reported zoonotic pathogen in the European Union and European Economic Area. Most human infections are caused by serovars Enteritidis and Typhimurium. Genomic characterisation of Salmonella isolates from humans and animals has become a routine public health surveillance tool in many countries. In this study, the relative contributions of several potential sources of human infection of the five frequently reported Salmonella serovars were estimated using machine-learning methods based on a large, cross-sectional collection of genomes from human cases, and animal and environmental sources, across ten European countries. To define the population structure, core-genome Multilocus Sequence Typing was performed. A supervised machine-learning approach was applied for source attribution in the form of a Random Forest classifier. The source and country attribution models achieved moderate accuracy (F1=0.6–0.9), which is lower than in previous studies using machine-learning on Whole Genome Sequencing data. However, attributions of human clinical isolates to different sources were generally in line with previous findings for these five serovars. While the lack of clonality in some sources hindered their prediction, it is also likely that certain sources (e.g., pets) do not serve as major contributors to human infection. Therefore, in most cases attributing these sources to the livestock species they are typically associated with, is likely appropriate. Country attributions showed that substantial human cases are attributable to countries other than their own, indicating geographical interrelatedness of sources. This highlights the value of internationally harmonised Salmonella-control policies in the food production chain.Highlights: - Salmonella source attribution done in 10 EU/EEA countries. - Over 10,000 cgMLST genomes analyzed using Random Forest. - Five main serovars linked mainly to various livestock sources. - Cross-border cases support EU-wide harmonised control policies. - Pets likely play only a minor role in Salmonella transmission.engModellingSalmonellosisSource AttributionWhole-Genome SequencingEuropeInfecções Gastrointestinaisjournal article10.1016/j.jinf.2025.1066321532-274241093134