The evolution, evolvability and engineering of gene regulatory DNA

Area of Science:

• Genomics
• Evolutionary Biology
• Systems Biology

Background:

• Mutations in non-coding DNA regulatory sequences significantly impact gene expression, organismal traits, and fitness.
• Mapping DNA sequences to fitness landscapes is crucial for understanding evolution but challenging due to vast sequence spaces.

Purpose of the Study:

• To develop sequence-to-expression models for capturing regulatory fitness landscapes.
• To decipher principles governing regulatory evolution.
• To enable the design of regulatory sequences for expression engineering.

Main Methods:

• Utilized millions of randomly sampled yeast promoter DNA sequences and their measured expression levels.
• Developed deep neural network models to predict gene expression from DNA sequence.
• Applied models to study expression divergence under different evolutionary regimes (drift, selection-mutation).

Main Results:

• Achieved excellent prediction performance with sequence-to-expression models, enabling generalization across vast sequence spaces.
• Regulatory evolution is rapid, subject to diminishing returns epistasis, and constrained by conflicting environmental objectives.
• Stabilizing selection moderates regulatory complexity, and regulatory mutation effect sizes follow a power law.

Conclusions:

• The developed models provide a general framework for designing regulatory sequences and engineering gene expression.
• The study offers insights into the dynamics of regulatory evolution, mutational robustness, and evolvability.
• An approach for detecting selection signatures and discovering convergent regulatory evolution was presented.