Spatial distribution prediction of environmental microbial communities based on deep learning combined with MAXENT model

The study of soil microbial community distribution is crucial for understanding ecosystem functioning. However, predicting their spatial distribution remains challenging due to complex interactions between environmental factors such as soil pH, temperature and precipitation. Traditional distribution prediction models usually have difficulty in handling high-dimensional data and nonlinear relationships, which may affect the accuracy of predictions. To address these limitations, this study combines deep learning techniques with the MAXENT model to develop a robust microbial distribution prediction framework. Experimental results show that the proposed microbial distribution model for Latin American soils has a Kappa coefficient of 0.87 and a spatial consistency score of 0.85, which are significantly better than the independent MAXENT model. These findings highlight the potential of the approach to improve spatial resolution, provide valuable ecological insights, and offer a reliable tool for soil microbial research, ecological restoration, and agricultural management.