For 50 years, predicting the final 3D structure of a protein from its amino acid sequence was one of biology’s grand challenges. The breakthrough of AI models like AlphaFold largely solved this problem, giving us an unprecedented look at the finished product. But this triumph raised an even more profound question: if we know the destination, can we now map the journey? The function of a protein is determined not just by its final shape, but by the complex, high-speed journey it takes to get there—its folding pathway.

Understanding these pathways has been incredibly difficult. The process involves a dizzying number of fleeting, transient intermediate shapes, and simulating every atomic movement is computationally prohibitive, even for supercomputers. This is where the next wave of AI is making its mark. Scientists are now combining the predictive power of AI with molecular dynamics (MD) simulations to chart these once-invisible routes. Instead of brute-forcing the simulation, AI models can learn the patterns of folding from shorter simulations and intelligently predict the most likely next steps.

This AI-accelerated approach allows researchers to model entire folding pathways, identify key intermediate states, and even predict how a genetic mutation might alter the folding route, potentially leading to disease. We are moving beyond a static picture of proteins and into the realm of protein dynamics. It’s the difference between having a photograph of a car and a GPS animation of its entire journey, including every turn and traffic jam. This deeper understanding is essential for designing drugs that can stabilize correctly folded proteins or correct misfolding pathways, opening up new strategies to combat diseases rooted in protein misfolding.