AI-enabled exploration of the "dark matter" of the protein universe
Evan Zihe Song · Theoretical and Natural Science · 2024
The "dark matter" of the protein universe, consisting of proteins lacking structural information or functional annotations, represents a significant challenge in understanding the complexity of life. Recent breakthroughs in artificial intelligence (AI), particularly in protein structure prediction, have revolutionized our ability to illuminate this uncharted territory. AI-based methods such as AlphaFold and RoseTTAFold can predict protein structures with unprecedented accuracy and scale, while large-scale databases provide access to the predicted structural models for hundreds of millions of proteins. Leveraging these AI tools and databases, researchers can uncover novel protein families, folds, and functions, and even design new proteins, paving the way for advances in basic biology, biotechnology, and medicine. This review discusses the recent progress of AI-enabled exploration of the "dark matter" of the protein universe, highlights recent advancements, and outlines future challenges and opportunities in this field.