ThermalProGAN: A sequence-based thermally stable protein generator trained using unpaired data.

Area of Science:

• Biotechnology
• Protein Engineering
• Computational Biology

Background:

• Protein engineering for novel properties is challenging, often relying on inefficient trial-and-error methods.
• Existing predictive models require difficult-to-obtain paired data, limiting advancements in protein design.
• There is a significant need for efficient methods to generate proteins with enhanced stability for industrial and academic applications.

Purpose of the Study:

• To introduce a novel sequence-based approach for designing thermally stable proteins.
• To develop and validate a generative adversarial network (GAN) model, ThermalProGAN, for protein engineering.
• To demonstrate the feasibility of transferring desired protein properties using sequence information.

Main Methods:

• Development of the sequence-based unpaired-sample of novel protein inventor (SUNI) framework.
• Implementation of ThermalProGAN, a generative adversarial network, for protein sequence generation.
• Utilizing molecular dynamics simulations to assess the thermal stability of generated proteins.

Main Results:

• ThermalProGAN successfully generated thermally stable protein variants by mutating a median of 32 residues.
• A known protein (1RG0) was modified to a thermally stable form with 51 mutations, retaining structural similarity and likely function.
• Molecular dynamics simulations (840 ns total) confirmed increased thermal stability for engineered proteins, including COVID-19 vaccine candidates.

Conclusions:

• The study presents a successful proof of concept for transferring desired protein properties, specifically thermal stability, using sequence-based generative models.
• ThermalProGAN offers a viable alternative to traditional protein engineering methods, enabling the design of proteins with enhanced stability.
• The source code for ThermalProGAN is publicly available, facilitating further research and application in protein design.