Additional file 2 of Parasite associations predict infection risk: incorporating co-infections in predictive models for neglected tropical diseases

Additional file 2: Figure S1. Observed infection prevalence for the four studied helminth parasites across 177 schools in Rwanda. Figure S2. Parasite co-infections and conditional correlation coefficients. Figure S3. Out-of-sample classification accuracies of the multi-parasite conditional random fields (CRF) and single-parasite gradient boosted machine (GBM) models. Figure S4. Area under the curve of the receiver operating characteristics (AUCs) for school-wide infection prevalence predicted by the conditional random fields (CRF) and single-parasite gradient boosted machine (GBM) models. Figure S5. Sensitivities of the conditional random fields (CRF) and single-parasite gradient boosted machine (GBM) models for predicting individual-level co-infections between A. lumbricoides and T. trichiura. Figure S6. Pearsonʼs correlations between each school’s observed standardised A. lumbricoides + T. trichiura co-infection ratio (SCR) and predicted SCRs from the conditional random fields (CRF) and single-parasite gradient boosted machine (GBM) models. Figure S7. Map of Rwanda’s provencial districts. Figure S8. Predicted infection prevalence of hookworm in Rwandan schoolchildren. Figure S9. Predicted infection prevalence of Schistosoma mansoni in Rwandan schoolchildren.