In 2018, when DeepMind unveiled AlphaFold to the public, a pivotal discovery in medical technology that could predict single-chain protein structures, it disrupted the way biology is done, giving new directions in the field of structural biology.
Two years later, DeepMind announced AlphaFold 2, which cracked the half-century puzzle of protein folding, a defining moment for computational science and AI in the realm of life sciences. Since then, we have had AlphaFold-Multimer, AlphaFill and more.
While the original AlphaFold was pivotal for predicting single-chain protein structures, the newest version, AlphaFold-latest, released this week, is even bigger and better. It can now anticipate structures from nearly all molecules in the Protein Data Bank (PDB)—a comprehensive database for 3D biological molecule structures—and has extended its capabilities to include small molecules, proteins, nucleic acids, and molecules with post-translational modifications.
The AlphaFold-latest has been developed by Isomorphic Labs, a London-based company under Alphabet’s umbrella, along with DeepMind.
Isomorphic Behind AlphaFold-latest Revolution
Founded in 2021 by Demis Hassabis, who also had a significant role in DeepMind’s inception, Isomorphic endeavors to use AI in drug discovery and research on severe human diseases.
The brains behind Isomorphic include tech veteran Miles Congreve, serving as chief scientific officer, who contributed to the design of 20 clinical-stage drugs and co-invented Kisqali (Ribociclib), a marketed breast cancer treatment. Sergei Yakneen is the chief technology officer with over two decades of expertise spanning engineering, machine learning, product development, and research in life sciences and medicine.
The company’s name reflects its philosophy: the idea that strategies from AI can be mapped onto pharmacology to solve complex biological challenges more efficiently.
At the core of Isomorphic Labs’ philosophy is the belief in an interdisciplinary approach, combining insights from AI with biosciences to innovate and accelerate the development of new medicines. They prioritise AI and machine learning, suggesting a transformative potential beyond the mere augmentation of existing methods.
The goal of Isomorphic Labs is to make the drug discovery process more scalable, reducing the traditionally long timelines and high costs of bringing new treatments to market.
As they continue to navigate the complex landscape of AI-driven drug discovery, Isomorphic Labs’ philosophy is likely to evolve, adapting to new challenges and scientific advancements in their pursuit of revolutionising how we discover and develop new medicines.
What AlphaFold-latest will Achieve
The new AlphaFold model surpasses traditional methods like AutoDock Vina in ligand docking accuracy by starting from scratch with only protein sequences and ligands. It also betters its predecessor, AlphaFold 2.3, especially in predicting protein-protein interactions and excels in modelling antibody binding.
Additionally, it leads in protein-nucleic acid interface prediction and RNA structure forecasting, albeit slightly behind the best manual methods from CASP15. Moreover, it now predicts structures of complex components, including bonded ligands and various molecular modifications.
By accurately modelling proteins, ligands, nucleic acids, and post-translational modifications together, it facilitates a deeper understanding of complex biological mechanisms. A prime illustration is the structure of CasLambda—a variant in the CRISPR system, known for genome editing—bound to crRNA and DNA. The model’s prediction of CasLambda, notable for its compact size, hints at potential for more efficient genome editing applications.
AlphaFold Through the Years
When OpenAI debuted ChatGPT, it broke the internet. However, OpenAI owes it to Google Brain, the brain behind Transformer, the neural network architecture that powers ChatGPT’s language models GPT 3.5, GPT-4 and more, which was introduced in a seminal 2017 paper.
The AlphaFold model is also built with transformers. Transformers have taken the world of machine learning by storm since being introduced by Google Brain researchers in a seminal 2017 paper. The AlphaFold team created a new type of transformer designed specifically to work with three-dimensional structures, which they call Invariant Point Attention (IPA).
Since then, AlphaFold has found a variety of real-life applications including finding vaccines for malaria, liver cancer, COVID-19, delivering gene therapy and more. Not just drug discovery, AlphaFold finds a range of applications beyond it.
For example, DeepMind partnered with the Centre for Enzyme Innovation at the University of Portsmouth to engineer faster enzymes for recycling some of the world’s most polluting single-use plastics. It has been used to design new types of proteins that more efficiently break down plastic waste. The researchers have engineered an enzyme that can break down entire plastic containers. Enzymatic depolymerization is a promising method for recycling plastics, as enzymes can be much more specific than chemical catalysts and can degrade a much more diverse waste stream.
With applications reaching far beyond medicine, AlphaFold has captured the global scientific community’s attention as the most-read paper of 2022. At a time when big techs like Meta is laying off their protein folding team, Google open-sources the codes of AlphaFold, allowing countless companies to innovate and devise extraordinary solutions with the potential to transform human lives.
The post Will Alphabet Achieve With Isomorphic Labs What It Couldn’t With DeepMind? appeared first on Analytics India Magazine.
