Modeling the vertebrate regulatory sequence landscape by UUATAC-seq and deep learning

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Vertebrate genome regulatory sequences are not fully understood.
• Identifying and interpreting cis-regulatory elements (cCREs) is crucial for understanding gene regulation.

Purpose of the Study:

• To develop a rapid and sensitive method for mapping chromatin accessibility landscapes.
• To identify and characterize candidate cis-regulatory elements (cCREs) across diverse vertebrate species.
• To develop a deep-learning model for interpreting cis-regulatory grammar and predicting cCREs.

Main Methods:

• Developed an ultra-throughput, ultra-sensitive single-nucleus assay for transposase-accessible chromatin using sequencing (UUATAC-seq).
• Mapped cCREs across five representative vertebrate species using UUATAC-seq.
• Introduced Nvwa cis-regulatory element (NvwaCE), a deep-learning model for cis-regulatory sequence interpretation.

Main Results:

• UUATAC-seq enabled 1-day chromatin accessibility landscape construction.
• Genome size influences the number, but not the size, of cCREs across species.
• NvwaCE demonstrated conserved regulatory grammar, organizing cCREs into functional modules.
• NvwaCE accurately predicted mutation effects on cCRE function, consistent with QTL and genome editing data.

Conclusions:

• UUATAC-seq provides a valuable resource for rapid cCRE mapping.
• Regulatory grammar is more conserved than nucleotide sequences across vertebrates.
• NvwaCE advances the interpretation of cis-regulatory language and its functional organization.