Related Experiment Video
Updated: Dec 24, 2025
10:00
In Vivo Forward Genetic Screen to Identify Novel Neuroprotective Genes in Drosophila melanogaster
Published on: July 11, 2019
9.9K
Essential gene prediction in Drosophila melanogaster using machine learning approaches based on sequence and
Olufemi Aromolaran1,2,3, Thomas Beder2, Marcus Oswald2
1Department of Computer & Information Sciences, Covenant University, Ota, Ogun State, Nigeria.
Computational and Structural Biotechnology Journal
|April 8, 2020
Summary
Predicting essential genes computationally saves resources compared to experimental screens. This study uses machine learning with diverse features in Drosophila melanogaster and humans, achieving high accuracy and outperforming simpler methods.
Area of Science:
- Computational Biology
- Genomics
- Machine Learning
Background:
- Essential genes are critical for organism viability.
- Experimental methods like RNAi screens are resource-intensive.
- Computational prediction offers a scalable alternative.
Purpose of the Study:
- To develop and evaluate a machine learning model for predicting essential genes in Drosophila melanogaster.
- To assess the contribution of various genomic and proteomic features to prediction accuracy.
- To validate the approach in human essential gene prediction.
Main Methods:
- Generated 27,340 features from nucleotide/protein sequences, gene networks, protein-protein interactions, conservation, and annotations.
- Employed machine learning with cross-validation for essential gene prediction.
- Compared performance against a benchmark method using only protein sequence features.
Main Results:
- Achieved high prediction performance in D. melanogaster (ROC-AUC 0.90, PR-AUC 0.30, F1 0.34).
- Significantly outperformed the benchmark method (P < 0.001).
- Demonstrated excellent results in human essential gene prediction (ROC-AUC 0.97, PR-AUC 0.73, F1 0.64).
Conclusions:
- A comprehensive feature set integrating diverse biological data improves essential gene prediction.
- Machine learning models can effectively predict gene essentiality across species.
- Network, functional, and sequence-based features are highly informative.
