The AlphaFold revolution
In 2020, DeepMind's AlphaFold2 demonstrated that it could predict the three-dimensional structure of a protein from its amino acid sequence with remarkable accuracy. Protein structure determination had been one of biology's hardest problems for 50 years. By 2022, AlphaFold had predicted structures for virtually all proteins whose sequences are known — over 200 million. Researchers can now download the predicted structure of almost any protein in seconds rather than spending years determining it experimentally.
Drug discovery
Conventional drug discovery takes on average 12-15 years and over a billion dollars to bring a new drug to market. AI-driven drug discovery platforms can generate and screen candidate molecules computationally at speeds that would be impossible experimentally, and can predict which candidates are likely to bind to target proteins and avoid dangerous side effects.
The reproducibility challenge
AI-assisted research also raises new challenges. AI models can identify patterns in large datasets that human researchers would miss — but they can also identify spurious patterns. The reproducibility crisis in science may be amplified if AI tools find patterns that are specific to the training data rather than reflecting real phenomena.