Longitudinal Clinical and Minimally Invasive Readouts Enable Early Assessment of Disease Trajectory in Preclinical...

Murine infection models are essential for investigating host-pathogen interactions, disease progression, vaccine-induced immune responses and protection. Early identification of disease trajectories is important for both animal monitoring and high-quality biological sample collection. Although clinical and microbiological readouts such as body-weight changes, clinical scores, and circulating bacterial burden are routinely used to monitor disease progression, individual parameters often fail to capture short-term disease trajectory when analysed separately. Here, four longitudinal clinical and minimally invasive readouts (cumulative weight loss, day-to-day weight change, clinical score and blood bacterial load) were integrated using Machine Learning approaches to assess short-term disease trajectory and the occurrence of a fatal outcome within the subsequent three days. Two independently conducted Salmonella enterica serovar Typhimurium murine infection studies were analysed, with the first one used for model development and internal testing and the second one for independent external validation. Four classifiers (Random Forest, Support Vector Machine, XGBoost, and penalized logistic regression) alongside three ensemble strategies (majority voting, weighted voting, and stacked meta-learner) were evaluated in this experimental context. In the internal test set, all models achieved high performance, while in the independent external validation dataset, the stacked ensemble outperformed all other models achieving an accuracy of 0.96 and a specificity of 0.75. Although evaluated in a relatively small cohort using a single mouse genetic background and one Salmonella enterica serovar Typhimurium strain, these results indicate that the proposed Machine Learning framework, based on readily obtainable and minimally invasive longitudinal readouts, can serve as a valuable tool to anticipate short-term disease trajectory, support disease monitoring and improve experimental planning.