Predicting gene essentiality in Caenorhabditis elegans by feature engineering and machine-learning

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Identifying essential genes is crucial for understanding fundamental biological processes.
• Predicting essential genes computationally from molecular and phenomic data remains challenging.
• Functional genomic tools have experimentally identified essential genes, but predictive models are needed.

Purpose of the Study:

• To develop and validate a machine-learning (ML)-based workflow for genome-wide essential gene prediction in Caenorhabditis elegans.
• To identify key molecular and phenomic predictors of essential genes.
• To explore the relationship between essential genes and their genomic context.

Main Methods:

• Utilized extensive datasets from Caenorhabditis elegans.
• Constructed a machine-learning workflow for essential gene prediction.
• Performed complementary analyses to associate essential genes with chromosomal location and other features.

Main Results:

• Achieved highly accurate classifications of essential genes using trained ML models.
• Identified strong predictors for essential genes, including chromosomal location, low single nucleotide polymorphism (SNP) density, and promoter epigenetic markers.
• Found essential genes are involved in protein/nucleotide processing, transcribed widely, enriched in reproductive tissues, or targeted by CSR-1-bound small RNAs.

Conclusions:

• Essential genes in C. elegans exhibit specific genomic and functional characteristics.
• Hypothesized an interplay between epigenetic markers and germline small RNA pathways, potentially involving transcription-based memory.
• The ML approach requires further evaluation for applicability to other metazoans.