Protein Structure Prediction, Structural Bioinformatics and Deep Learning
Year: 2024
Type: article
Abstract: Protein structure prediction is essential for understanding protein stability, and interactions. It holds immense potential for drug discovery and protein engineering. However, despite advancements in structural bioinformatics and artificial intelligence, a standardised model for structure prediction still needs to be worked out. Even prominent models like AlphaFold often undergo architectural changes. To address this gap, a comprehensive detail of recent progress and challenges in deep learning-based protein structure prediction has been presented. Additionally, a benchmark system for structure prediction and visualization, enabling analysis of user-provided protein sequences has been introduced. Looking to the need for efficient and accurate methods to decipher protein structures and their biological roles, DeepProtein has been introduced. This model leverages the potent representation learning capabilities of the Transformer architecture to directly predict secondary and tertiary structures from integer-encoded amino acid sequences. The results demonstrate DeepProtein’s effectiveness in secondary structure prediction. Further refinement is necessary to enhance its performance in predicting higher-order structures. The present document highlights the potential of Transformer-based architectures for structure prediction, paving the way for future research in structural bioinformatics and related fields.
Author: Tejas Agarwal
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W4405153727