Deciphering DNA's Sequence-Dependent Structure and Deformability with Normalizing Flows.

Area of Science:

• Biophysics
• Computational Biology
• Molecular Biology

Background:

• DNA structure and deformability are crucial for cellular functions.
• Accurately modeling DNA's conformational behavior is a persistent challenge.
• Existing models rely on simplified elastic energy functions, neglecting complex stereochemical effects.

Purpose of the Study:

• To develop a novel AI-based method for deciphering sequence-dependent DNA structure and deformability.
• To accurately quantify deformation energies for any double-stranded DNA sequence.
• To overcome limitations of previous functional form assumptions in DNA mechanics.

Main Methods:

• Utilized normalizing flows, a type of AI model.
• Captured multimodal and correlation effects between DNA's internal coordinates.
• Developed a new approach to model DNA conformational flexibility.

Main Results:

• The AI method accurately describes sequence-dependent DNA structure and deformability.
• Normalizing flows effectively model complex correlations in DNA internal coordinates.
• Deformation energies for double-stranded DNA structures and sequences can be precisely quantified.

Conclusions:

• The proposed AI-based method offers a significant advancement in modeling DNA mechanics.
• This approach accurately quantifies DNA deformation energies, addressing limitations of prior methods.
• The method has broad future applications and can be extended to other complex biomolecules.