Related Experiment Video
Updated: Dec 12, 2025
11:38
Author Spotlight: Enhancing PSC-to-Functional Cell Differentiation Using ML Models Based on Live-Cell Bright-Field Imaging
Published on: October 4, 2024
946
DeepPred-SubMito: A Novel Submitochondrial Localization Predictor Based on Multi-Channel Convolutional Neural Network
Xiao Wang1, Yinping Jin1, Qiuwen Zhang1
1School of Computer and Communication Engineering, Zhengzhou University of Light Industry, Zhengzhou 450002, China.
International Journal of Molecular Sciences
|August 14, 2020
Summary
This study introduces DeepPred-SubMito, a novel deep learning predictor for accurately identifying mitochondrial protein locations. It addresses limitations in existing methods by considering all four compartments and handling imbalanced datasets for improved disease pathogenesis insights.
Area of Science:
- Biochemistry
- Bioinformatics
- Computational Biology
Background:
- Mitochondrial protein localization is crucial for cellular function; mislocalization contributes to human diseases.
- Existing prediction tools often overlook the intermembrane space and rely on manual feature engineering.
- Imbalanced datasets pose challenges for accurate machine learning model training.
Purpose of the Study:
- To develop a novel, end-to-end deep learning predictor, DeepPred-SubMito, for accurate protein submitochondrial localization.
- To address the limitations of existing methods, including the underrepresentation of the intermembrane space and dataset imbalance.
- To provide a data-driven approach for predicting protein locations within mitochondria.
Main Methods:
- Implemented random over-sampling to mitigate the impact of imbalanced datasets.
- Utilized a multi-channel bilayer convolutional neural network to learn high-level features from protein subsequences.
- Employed a fully connected layer for final prediction output.
Main Results:
- DeepPred-SubMito demonstrated superior performance compared to state-of-the-art predictors on benchmark datasets (SM424-18 and SubMitoPred).
- Validated effectiveness through 10-fold and 5-fold cross-validation.
- Confirmed predictive accuracy on the M983 dataset, highlighting its utility for real-world applications.
Conclusions:
- DeepPred-SubMito offers a significant advancement in predicting protein localization within all four mitochondrial compartments.
- The deep learning approach overcomes limitations of traditional methods, enabling more accurate disease mechanism and drug design insights.
- This tool provides a robust, data-driven solution for mitochondrial protein localization prediction.
