Generating new protein sequences by using dense network and attention mechanism.

Area of Science:

• Biotechnology
• Computational Biology
• Protein Science

Background:

• Protein engineering aims to enhance protein properties and functions through gene sequence modification.
• De novo protein design is crucial for creating novel proteins with desired characteristics.

Purpose of the Study:

• To introduce the Dense-AutoGAN model for generating complex protein sequences.
• To improve the accuracy and effectiveness of protein sequence generation for research.

Main Methods:

• Developed Dense-AutoGAN, integrating Generative Adversarial Networks (GAN) with an Attention mechanism and a novel Dense convolutional neural network.
• Utilized an Encoder-decoder architecture within the GAN to refine sequence similarity and control variation range.
• Employed multi-layer dense network transmission to expand the training space and enhance sequence generation.

Main Results:

• Dense-AutoGAN successfully generated complex protein sequences mapped to specific protein functions.
• Generated protein sequences demonstrated high accuracy and effectiveness in chemical and physical properties.
• Comparative analysis confirmed superior performance of Dense-AutoGAN over existing models.

Conclusions:

• Dense-AutoGAN represents a significant advancement in de novo protein design.
• The model effectively generates novel proteins with enhanced properties, meeting diverse research needs.
• This technology holds promise for future applications in biotechnology and drug discovery.