Related Experiment Video
Updated: Aug 12, 2025
06:41
Imaging and Analysis for Quantifying Maize (Zea mays) Abiotic Stress Phenotypes
Published on: March 28, 2025
966
DeepAProt: Deep learning based abiotic stress protein sequence classification and identification tool in cereals
Bulbul Ahmed1, Md Ashraful Haque2, Mir Asif Iquebal1
1Division of Agricultural Bioinformatics, ICAR-Indian Agricultural Statistics Research Institute, New Delhi, India.
Frontiers in Plant Science
|January 30, 2023
Summary
Climate change impacts crop yields, especially in Poaceae family crops. A new Deep Learning model using the SIELU activation function accurately identifies abiotic stress proteins, aiding food security.
Area of Science:
- Agricultural Science
- Computational Biology
- Biotechnology
Background:
- Climate change poses significant threats to global crop production, particularly affecting staple crops from the Poaceae family.
- Extreme weather conditions induce abiotic stress in plants, leading to reduced crop yields and impacting food security.
- Artificial intelligence (AI) and computational methods are increasingly vital for predicting plant stress responses and interpreting complex biological data.
Purpose of the Study:
- To develop a novel Deep Learning (DL) model for classifying abiotic stress-responsive proteins in Poaceae crops.
- To introduce and evaluate a new activation function, Gaussian Error Linear Unit with Sigmoid (SIELU), within the DL model.
- To create a user-friendly bioinformatics tool for rapid and cost-effective identification of stress proteins in crops.
Main Methods:
- Data on cold, drought, heat, and salinity stress-responsive proteins from Poaceae crops were collected from public databases.
- A DL model was developed incorporating the novel SIELU activation function and other hyperparameters.
- The performance of the SIELU-based DL model was compared against the GeLU activation function, Support Vector Machine (SVM), and Random Forest (RF) algorithms.
Main Results:
- The DL model with the SIELU activation function achieved high accuracy rates: 95.11% for cold, 80.78% for drought, 94.97% for heat, and 81.69% for salinity stress.
- The proposed SIELU activation function demonstrated superior performance compared to the GeLU activation function, SVM, and RF methods.
- A web-based tool (DeepAProt) and a mobile application were developed for practical application of the model.
Conclusions:
- The novel SIELU activation function significantly enhances the accuracy of DL models for identifying abiotic stress proteins in Poaceae crops.
- The developed DeepAProt tool offers a rapid, economical, and convenient resource for researchers in crop stress management.
- This work contributes to improving crop production, enhancing food security, and supporting UN Sustainable Development Goal 2 (Zero Hunger).
