An AI that can design new proteins could help unlock new cures and materials 

An AI that can design new proteins could help unlock new cures and materials 

A new AI tool could allow researchers to discover previously unknown proteins and create new ones. It could lead to new materials, faster research, and the development of more efficient vaccines.

Alphabet-owned AI lab DeepMind took the world by surprise in 2020 when it announced AlphaFold, an AI tool that used deep learning to solve one of the “grand challenges” of biology: accurately predicting the shapes of proteins. Understanding the shape of proteins is crucial to our lives. Earlier this summer DeepMind announced that AlphaFold could now predict the shapes of all proteins known to science.

The new tool, ProteinMPNN, described by a group of researchers from the University of Washington in two papers published in Science today (available here and here), offers a powerful complement to that technology.

These papers show how deep learning is revolutionizing the design of protein by giving scientists new research tools. ProteinMPNN will allow researchers to create new proteins from scratch. Instead of tweaking proteins that already exist in nature, researchers have traditionally used deep learning to engineer proteins.

” Proteins are the solution to all problems in nature, from harvesting sunlight to creating molecules. David Baker, one of the scientists behind this paper and director of The Institute for Protein Design at Washington University, says that proteins are the basis of biology.

” They evolved during evolution to solve the problems organisms faced during evolution. Today, we face new challenges like covid.

Proteins are composed of hundreds of thousands amino acids, which are linked up in long chains that then fold into three-dimensional shapes. AlphaFold allows researchers to predict the structure and provide insight into their behavior.

ProteinMPNN will assist researchers with the reverse problem. If researchers already know the exact structure of a protein, they can use it to find the amino acids sequence that will fold into that structure. The system relies on a neural network that has been trained on a large number of examples, which then folds into three-dimensional structures.

But researchers also have to solve another problem. Researchers must also solve another problem in order to design proteins that address real-world problems. For example, a new enzyme that can digest plastic.

To do that, researchers in Baker’s lab use two machine-learning methods, detailed in an article in Science last July, that the team calls “constrained hallucination” and “in painting.”

“Constrained hallucination” lets users do a random search among all possible protein sequences and favor sequences with certain functions. Machine learning’s ability of crunching large data sets makes it possible to explore all possible structures. This “hallucination”, however, allows you to explore the space for all possible proteins. There are 20 amino acids, which can be combined into a massive number of possible sequences.

Nature has only sampled… a small fraction. Baker states that limiting the search to sequences found in nature would not get you anywhere. In painting” works in the same way as autocomplete in a word process, but for protein sequences and structures. These methods allow researchers to create a new protein that isn’t known in nature, such as a huge ring-like structure.

Baker’s group is exploring whether these ring-like structures can be used in tiny machines that operate at nanoscale. These nanomachines could one day be used to unclog arteries.

The ability to use machine-learning to design proteins in such a way is “a very large deal,” Lynne Regan, professor of Biochemistry and Biotechnology at the University of Edinburgh.

Machine learning will speed up the process and allow researchers to create new proteins and structures at a larger scale. The software is more than 200 times faster than the previous best tool and requires minimal user input, potentially lowering the barriers to entry for protein design. These contributions, and others, are changing the field of biomolecular design and prediction,” says Jeffrey Gray of Johns Hopkins University’s department of chemical and biomolecular engineers. Gray believes the implications are significant in understanding biology, health and designing new molecules to alleviate human suffering.

Gray says his lab will combine deep-learning tools they developed with ones from the Baker lab to better understand the immune system and immune-related diseases, and use AI to design therapeutics.

AlphaFold ushered biology into a new age by solving the protein structure prediction problem and demonstrating how AI and [machine-learning] can transform biology,” says Pushmeet Sharma, head of DeepMind’s AI for Science team. “ProteinMPNN is another proof of this paradigm shift, designing proteins for specific tasks.”

ProteinMPNN, which is now available free on the open-source software repository GitHub, will give researchers the tools to make unlimited new designs. Baker states, “The challenge is obviously… what are you going in design?”

Read More