The Iconic α-Helix: From Pauling to the Present

Area of Science:

• Biochemistry
• Computational Biology
• Artificial Intelligence

Background:

• The protein folding problem has been a central challenge in science for nearly a century.
• Early attempts to predict protein structures from amino acid sequences yielded limited success.
• X-ray crystallography confirmed the Pauling-Corey-Brandson α-helix model in 1951.

Purpose of the Study:

• To review the historical context of protein structure prediction.
• To highlight the transformative impact of deep learning AI on protein structure prediction.
• To discuss the implications of AI-driven predictions in biochemistry.

Main Methods:

• Review of historical protein structure prediction efforts.
• Introduction of deep learning artificial intelligence, specifically AlphaFold2.
• Utilizing neural networks for macromolecular structure prediction.

Main Results:

• Deep learning AI, exemplified by AlphaFold2, can now predict most protein structures at near-atomic resolution.
• AlphaFold2 was recognized as Science magazine's 2021 "breakthrough of the year."
• ~200 million predicted protein structures are available via the AlphaFold2 Protein Structure Database.

Conclusions:

• Deep learning AI has revolutionized protein structure prediction, surpassing previous methodologies.
• The success of AI models like AlphaFold2, despite lacking physical chemistry underpinnings, presents a scientific conundrum.
• This advancement may signal a paradigm shift beyond traditional reductive scientific methods.