Abstract
BACKGROUND: Experimental models using Plasmodium berghei ANKA (PbA)-infected mice have been essential for uncovering cerebral malaria (CM) pathogenesis. However, variability in experimental CM (ECM) incidence, onset, and mortality introduce challenges when analyses rely solely on infection day, which may reflect different disease stages among animals. METHODS: We applied machine learning to predict ECM risk and onset in a cohort of 153 C57BL/6, 164 CBA, and 53 Swiss Webster mice. First, we fitted a logistic regression model to estimate the risk of ECM at any day using parasitemia data from day 1 to day 4. Next, we developed and trained a Random Forest Regressor model to predict the exact day of symptom onset. RESULTS: A total of 64.5% of the cohort developed ECM, with onset ranging between 5 and 11 days. Early increases in parasitemia were strong predictors for the development of ECM, with an increase in parasitemia equal to or greater than 0.05 between day 1 and day 3 predicting the development of ECM with 97% sensitivity. The Random Forest model predicted the day of ECM onset with high precision (mean absolute error: 0.43, R(2): 0.64). CONCLUSION: Parasitemia dynamics can effectively identify mice at high risk of ECM, enabling more accurate modeling of early pathological processes and improving the consistency of experimental analyses.