Factors Controlling the Distribution of Intermediate Host Snails of Schistosoma in Crater Lakes in Uganda: A Machine Learning Approach

Abstract

Schistosomiasis affects over 700 million people globally. 90% of the infected live in sub-Saharan Africa, where the trematode species Schistosoma mansoni and S. haematobium transmitted by intermediate hosts (IH) of the gastropod genera Biomphalaria and Bulinus are the major cause of the human disease burden. Understanding the factors influencing the distribution of the IH is vital towards the control of human schistosomiasis. We explored the applicability of a machine learning algorithm, random forest, to determine significant predictors of IH distribution and their variation across different geographic scales in crater lakes in western Uganda. We found distinct variation in the potential controls of IH snail distribution among the two snail genera as well as across different geographic scales. On the larger scale, geography, diversity of the associated mollusk fauna and climate are important predictors for the presence of Biomphalaria, whereas mollusk diversity, water chemistry and geography mainly control the occurrence of Bulinus. Mollusk diversity and geography are relevant for the presence of both genera combined. On the scale of an individual crater lake field, Biomphalaria is solely controlled by geography, while mollusk diversity is most relevant for the presence of Bulinus. Our study demonstrates the importance of combining a comprehensive set of predictor variables, a method that allows for variable selection and a differentiated assessment of different host genera and geographic scale to reveal relevant predictors of distribution. The results of our study contribute to making realistic predictions of IH snail distribution and schistosomiasis prevalence and can help in supporting strategies towards controlling the disease.