Predicting essential genes of 41 prokaryotes by a semi-supervised method.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Essential genes are crucial for organism survival and reproduction.
• Supervised machine learning methods for essential gene prediction often require substantial labeled data.
• Insufficient labeled data can limit the effectiveness of traditional prediction models.

Purpose of the Study:

• To develop and evaluate a novel semi-supervised learning method for predicting essential genes.
• To address the challenge of insufficient labeled data in essential gene prediction.
• To assess the performance and universality of the proposed method across multiple prokaryotic species.

Main Methods:

• Proposed a graph-based semi-supervised learning method named Learning with Local and Global Consistency (LGC).
• Applied the LGC classifier to predict essential genes in 41 prokaryotes.
• Evaluated the classifier's performance using intra-organism prediction and leave-one-species-out cross-validation.

Main Results:

• The LGC method achieved an average AUC of 0.723 for intra-organism prediction with a 0.5 labeled sample ratio across 41 organisms.
• Demonstrated acceptable prediction performance even when labeled data is limited.
• Showcased good universality of the proposed method across different prokaryotic species.

Conclusions:

• The LGC method offers an effective solution for essential gene prediction with limited labeled data.
• The approach exhibits strong performance and broad applicability in prokaryotic essential gene identification.
• Semi-supervised learning provides a viable alternative when large-scale labeled datasets are unavailable.