StressGenePred

StressGenePred is a software tool to analyze time-series gene expression data from plants subjected to various environmental stresses. It aims to identify stress-specific biomarker genes and predict the type of stress a plant is experiencing based on its gene expression profile.

Key features of StressGenePred:

1. Integration of multiple stress types: It can analyze time-series transcriptome data from plants exposed to different stresses, such as heat, cold, salt, and drought.

2. Twin neural network model: The tool employs a twin neural network consisting of a biomarker gene discovery model and a stress type prediction model, which share the same logical layer to reduce training complexity.

3. Confident Multiple Choice Learning (CMCL) loss: This loss function ensures that the twin model selects biomarker genes that respond specifically to a single stress.

4. Simple feature embedding method: StressGenePred uses a simple feature embedding method to represent the gene expression data.

5. Improved accuracy and specificity: Experiments using Arabidopsis gene expression data showed that StressGenePred outperformed other methods, such as limma feature embedding, support vector machines, and random forests, in classifying stress types and identifying known stress-related genes with higher specificity.