It might take time, but with your list of clues you would probably be able to piece together your Lego set eventually. But in the case of protein folding, no supercomputer is powerful enough to make any significant advancement on its own. That’s where quantum computers come into play.
Manya Bhargava returns to the IMSE blog to continue exploring the protein folding problem. In her previous blog, she explained how AlphaFold (AI system) uses known structures to predict the structures of unknown proteins. However, any AI system relies heavily on experimentally obtained data. On this new blog entry, Manya explains how quantum computers help to advance the problem by providing new solutions.
Superposition, being ON and OFF
Quantum computers utilize quantum mechanics to perform calculations in a different, and often more efficient, way to classical computers. Everyday computers use bits which exist in one of two states, taking the value of either 0 or 1 to store information. Just like a switch turning on or off. Quantum computers encode data using ‘qubits’ (quantum bits) which also have two states. These qubits could be electrons having ‘up’ and ‘down’ spins, or superconducting circuits with two energy levels. But unlike bits, they can exist in a combination of these states called a ‘superposition’ – they can exist in both states at the same time! This allows quantum computers to store much more information and perform more calculations in parallel compared to regular computers. It’s like having switches that can be both on and off at the same time.
Entanglement is another intrinsic property forming the basis of quantum computing. When two particles are entangled, changing the state of one will instantly affect the state of the other, no matter how far apart they are. It’s like flipping two coins and seeing that if one lands on heads, the other will always land on tails. Researchers can create groups of entangled qubits which can be used to encode information on quantum circuits, immensely increasing the processing power of quantum computers.
Applying quantum computing to the protein folding problem
Let’s return to the protein folding problem. Scientists have been able to model amino acid chains on quantum circuits, encoding the properties of the amino acids onto qubits. Entangled qubits can represent amino acids which are next to (and hence interacting with) each other in the protein chain and capture attributes like the orientation or hydrophobic character of amino acids. The quantum computer then encodes a superposition of many different protein configurations at the same time. Scientists can iterate through these, optimizing parameters until the lowest energy configuration is found – which corresponds to the native structure of the protein.
But this brings its own difficulties. Quantum computers with large numbers of qubits are incredibly difficult to build and maintain. As technology rapidly develops, researchers will be able to model increasingly complicated proteins. Quantum computers could provide us with a different solution to the PFP. Physical interactions between amino acids are simulated to reach a protein shape without the need for prior structure data. Used alongside programs like AlphaFold, this could be an important step forwards towards holistic, accurate and complete protein folding simulation. Building practical quantum computers is an immense task, but exciting new developments bring important applications ever closer.
The ability to precisely predict protein structures from only the amino acid sequence has been a holy grail of biology research for decades. Multidisciplinary collaboration between physics, biology and computer science makes it almost within reach. With rapid innovation and development, this is an exciting space to watch.