Proteins, often envisioned as static 3D structures, are far from being rigid sculptures. In reality, many proteins possess the remarkable ability to transform, altering their shapes to meet specific biological demands. Some configurations can transmit harmful signals during events like strokes or heart attacks, while others act as safeguards, limiting the molecular fallout. Essentially, proteins function as biological transistors, governing the body’s molecular “computer” that responds to internal and external cues. Scientists have long explored these shape-shifting proteins to unravel the mysteries of our physiological processes.
But what if we could engineer entirely novel biological “transistors” from scratch, independent of nature’s constraints? This question has led to the intersection of artificial intelligence and biology, marking a profound shift in our understanding of proteins and their potential applications.
AI’s Leap into Protein Science
Recent breakthroughs in deep learning have empowered AI to predict protein structures accurately, a milestone five decades in the making. These powerful algorithms have even ventured into the realm of imagination, conjuring up protein structures that defy the constraints of evolution. However, there’s a catch—these AI-generated structures, while intricate, remained static, akin to digital protein sculptures frozen in time.
A recent study in Science is challenging this norm by introducing flexibility into designer proteins. These innovative structures can transition between two distinct forms, akin to a hinge that can be open or closed, responding to external biological “locks.” Each state serves as a biological “0” or “1,” controlling cellular functions.
Dr. Florian Praetorius, a study author from the University of Washington, notes, “Before, we could only create proteins that had one stable configuration. Now, we can finally create proteins that move, which should open up an extraordinary range of applications.”
Lead author Dr. David Baker envisions these designer proteins being employed in various fields, from forming nanostructures that respond to environmental chemicals to revolutionizing drug delivery methods.
Deciphering the Protein Code
To understand this breakthrough, let’s revisit Biology 101. Proteins are the architects and builders of our bodies, originating from DNA. Genetic information is translated into amino acids, akin to beads on a string, which are then intricately folded into 3D shapes. Some protein parts cling to others, forming secondary structures that can resemble Twizzlers or intricate carpet-like sheets. These shapes further combine, culminating in highly complex protein architectures.
Understanding how proteins acquire their shapes enables us to potentially engineer entirely new proteins, expanding the biological toolkit and offering new avenues to combat diseases and infections.
In 2020, DeepMind’s AlphaFold and David Baker’s RoseTTAFold shook the structural biology field by accurately predicting protein structures solely from amino acid sequences. These AI models have since predicted the shapes of nearly every known and unknown protein, revolutionizing biological research.
Generative AI models like DALL-E and ChatGPT took this a step further, dreaming up entirely novel protein structures. The possibilities appeared boundless, from proteins that bind hormones to regulate calcium levels to artificial enzymes catalyzing bioluminescence.
However, a critical element was missing: flexibility. Many proteins change shape to convey vital biological messages, and replicating these shifts could unlock innovative medical solutions.
AI-Designed Biological Transistors
Designing a protein at the atomic level is complex; creating one with two configurations is a monumental challenge. The AI’s task is akin to creating snowflakes, each with unique structures, using the same amino acid “ice crystals.” These structures must rapidly shift between configurations, akin to an “on” or “off” switch, while remaining compatible within living cells.
The team developed rules for the design: the structures should appear dramatically different between states, change swiftly, function well within bodily fluids, and serve as reliable switches based on inputs and outputs.
The final design resembles a hinge with two rigid parts capable of relative movement, alongside a folded component. The protein starts in a closed state, triggered by a small peptide that binds to the hinges, forcing a change in configuration. This “effector peptide” was thoughtfully designed for specificity, reducing the risk of off-target binding.
AI can now engineer proteins with two distinct states, essentially creating biological transistors for synthetic biology. While custom-designed effector peptides are currently used, the potential extends to natural peptides, offering a revolutionary paradigm shift in biotechnology.
Study author Dr. Philip Leung aptly states, “This could revolutionize biotechnology in the same way transistors transformed electronics.”
By Impact Lab
