Human essential gene identification based on feature fusion and feature screening

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Essential genes are vital for species survival and are key targets for novel antibacterial drug development.
• Identifying essential genes is challenging due to their environmental variability.
• Accurate identification of essential genes is critical for therapeutic strategies.

Purpose of the Study:

• To develop a robust prediction model for identifying essential human genes.
• To enhance the accuracy of essential gene identification through advanced computational methods.
• To provide an improved methodology for essential gene discovery.

Main Methods:

• Collected essential gene data from human cancer cell lines.
• Utilized seven feature encoding methods (e.g., Kmer, Z-curve) for gene sequence characterization.
• Employed feature fusion and optimization strategies.
• Applied machine learning algorithms (classical and deep learning) for model construction and evaluation.

Main Results:

• Single-feature models achieved an Area Under the Curve (AUC) of 0.830.
• Feature fusion and optimization improved model performance.
• Classical machine learning models reached an AUC of 0.823.
• The deep learning model achieved a high AUC of 0.860.

Conclusions:

• Feature fusion and optimization significantly enhance the performance of essential gene prediction models.
• The developed deep learning model offers an advantageous and accurate method for essential gene identification.
• This approach provides a valuable tool for future drug discovery and therapeutic development targeting essential genes.