Construction of CNN Algorithm Model for Predicting the Health Status of Bombyx Moth Larvae Using the Gram Staining Pattern of Intestinal Microbes

Abstract

The present article confirms the crucial role of the microbial environment in the large intestine as a measure of individual health, as demonstrated through silkworm larvae feces. To further explore this, an artificial intelligence-based Convolutional Neural Network (CNN) model was constructed using Gram staining pattern data of intestinal microbes, and its usability was confirmed. In order to investigate the effect of intestinal inflammation on individual health, silkworm moth larvae were fed Dextran Sulfate Sodium (DSS) along with mulberry leaves, and the concentration of DSS was varied. The experimental group exhibited a low distribution ratio of Proteobacteria at the Phylum level, while the control group had a high distribution ratio. The opposite trend was observed in Firmicutes, which was confirmed by changes in the Gram staining pattern at the sub-classification stage. The experimental group had a lower species diversity index compared to the control group, suggesting a significant effect of the DSS solution treatment on individual health. As a result, cocoon formation occurred more rapidly in the experimental group, and the rate of molting increased with weight gain. The CNN algorithm model was developed using big data on Gram staining patterns of intestinal microbes from both the experimental and control groups. Verification using the validation group demonstrated high accuracy. These findings suggest that the health status of silkworm larvae can be predicted through intestinal microbes using the CNN algorithm model, provided that pre-processing procedures such as Gram staining are further elaborated in future studies.