Cross-species enhancer prediction using machine learning.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Cis-regulatory elements (CREs), like enhancers, control gene expression and influence complex traits.
• Knowledge of enhancers is limited in non-model mammalian species, hindering research.
• Comparative genomics has had limited success in identifying enhancers in these species.

Purpose of the Study:

• To investigate the efficacy of Machine Learning (ML) methods for identifying enhancers in non-model mammalian species.
• To leverage existing human and mouse enhancer data for cross-species prediction.
• To provide a valuable resource of predicted enhancers for cattle, pigs, and dogs.

Main Methods:

• Tested nine ML models using four DNA sequence representations.
• Utilized human and mouse enhancer data (VISTA and ChIP-seq) for cross-species prediction.
• Applied models to cattle, pig, and dog genomes.

Main Results:

• Identified 809,399 to 877,278 enhancer-like regions (ELRs) in the study species.
• Predicted ELRs constituted 11.6-13.7% of each genome, comparable to human genome proportions.
• Demonstrated ML's predictive capability for enhancer identification across mammalian species.

Conclusions:

• ML methods show strong potential for identifying enhancers in non-model mammals.
• The predicted enhancers offer a valuable resource for future research in cattle, pigs, and dogs.
• This study expands the understanding of CREs in understudied mammalian genomes.