Developers
July 28, 2020

Using AI For Scientific Discovery To Solve Protein Folding Problem

Scientists turn to AI to research 3D models of protein structures.
Source: Pixabay

How artificial intelligence can accelerate and impact scientific discoveries? Today we will talk about a system named AlphaFold. It counts with a dedicated team working on the use of AI to push research forward. It brings together experts from biology, physics, and machine learning. What do they do? They apply their knowledge to predict 3D structures of proteins.

AlphaFold has been developed over years of work.  It is built on decades of research based on datasets that can predict protein structures. The 3D models of proteins generated by AlphaFold are the most accurate yet developed.

Alphafold counts with open source code and an entire community that helps the development move forward. There are also independent implementations including developed models.  

What is the protein folding problem?

Proteins are large molecules that can be found in all essential life. Almost all functions that our body performs rely on proteins.  Contracting muscles, sensing light, turning food into energy.

What proteins can do depend on their 3D structures? Antibody proteins used by our immune systems are y-shaped, forming hooks. The antibodies latch onto viruses and bacteria, being able to detect disease-causing microorganisms so it can be eliminated.  

There are many types of proteins. Collagen proteins are shaped like cords, transmitting tension between cartilages ligaments bones and the skin.

The proteins have recipes, called genes. The genes are encoded in our DNA. If there's a mistake in the genes, it is translated as a malformed protein, which could result in disease or death.

What is this telling us? That many diseases are linked to malformed proteins. By knowing the recipe of a protein sometimes you can know its shape. Sometimes because sometimes the protein is so big that is hard to know its shape. Proteins are made of chains of amino acids.  

There's a study, called the Levinthal's paradox, that shows that it would take longer than the age of the universe to randomly enlist the possible configurations of a protein before reaching the 3D structure.

Proteins fold spontaneously, before one second. The prediction of how these chains fold into the 3D structure is known as the protein folding problem. Something that scientists work and have been working for decades.  

Scientists are interested in determining the structures of proteins because of protein's form. The form is thought to dictate its function. If scientists know the form and shape, the role of the cell can be guessed and a step further Is taken towards knowing it. Scientists can then develop the needed drugs that work with the protein's shape

In the past decades, researchers have been working determining shapes of proteins in labs using experimental techniques. These techniques include cryo-electron microscopy, nuclear magnetic resonance and X-ray crystallography.

Each method depends on trial and error. This procedure can take years of work and cost up to hundreds of thousands of dollars per each protein structure. For this exact reason, biologists are looking into AI to help the process. AI brings the ability to predict a protein's shape computationally from its genetic code, instead of determining it long and costly processes. AI can accelerate the research process.

Nowadays, the genomics is rich in data thanks to the reduction in the cost of genetic sequencing. Deep learning appears in the scene to help in the prediction problem of genomic data.

There's a global competition called the Critical Assessment of Techniques for Protein Structure Prediction that works to catalyze research and measure progress on the newest methods working to improve the accuracy of predictions.  

The CASP organizers have named "unprecedented progress in the ability of computational methods to predict protein structure".  The team of DeepMind has been classified as first in the ranking among the teams that entered.

The DeepMind team focused on the problem of modeling target shapes from scratch. A high degree of accuracy has been achieved when predicting the properties of protein structure. 

Both of these methods are based on deep neural networks that are trained to predict the properties of the protein. There are two properties that the networks predict. These are the distances between pairs of amino acids and the angles between chemical bonds that connect the amino acids.

In conclusion, scientists are turning into AI to do protein-based research. AI can accelerate the process in which a protein is analyzed and this can be life-changing for the time that it takes to develop a drug. We have seen how diseases are sometimes related to protein malfunction or deficiency. Based on the Levinthal's paradox, which states that "it would take longer than the age of the universe to randomly enlist the possible configurations of a protein before reaching the 3D structure", scientists have decided to use AI and ML.

Tags3DAIAlphaFoldProtein Folding
Lucas Bonder
Technical Writer
Lucas is an Entrepreneur, Web Developer, and Article Writer about Technology.

Related Articles

Back
DevelopersJuly 28, 2020
Using AI For Scientific Discovery To Solve Protein Folding Problem
Scientists turn to AI to research 3D models of protein structures.

How artificial intelligence can accelerate and impact scientific discoveries? Today we will talk about a system named AlphaFold. It counts with a dedicated team working on the use of AI to push research forward. It brings together experts from biology, physics, and machine learning. What do they do? They apply their knowledge to predict 3D structures of proteins.

AlphaFold has been developed over years of work.  It is built on decades of research based on datasets that can predict protein structures. The 3D models of proteins generated by AlphaFold are the most accurate yet developed.

Alphafold counts with open source code and an entire community that helps the development move forward. There are also independent implementations including developed models.  

What is the protein folding problem?

Proteins are large molecules that can be found in all essential life. Almost all functions that our body performs rely on proteins.  Contracting muscles, sensing light, turning food into energy.

What proteins can do depend on their 3D structures? Antibody proteins used by our immune systems are y-shaped, forming hooks. The antibodies latch onto viruses and bacteria, being able to detect disease-causing microorganisms so it can be eliminated.  

There are many types of proteins. Collagen proteins are shaped like cords, transmitting tension between cartilages ligaments bones and the skin.

The proteins have recipes, called genes. The genes are encoded in our DNA. If there's a mistake in the genes, it is translated as a malformed protein, which could result in disease or death.

What is this telling us? That many diseases are linked to malformed proteins. By knowing the recipe of a protein sometimes you can know its shape. Sometimes because sometimes the protein is so big that is hard to know its shape. Proteins are made of chains of amino acids.  

There's a study, called the Levinthal's paradox, that shows that it would take longer than the age of the universe to randomly enlist the possible configurations of a protein before reaching the 3D structure.

Proteins fold spontaneously, before one second. The prediction of how these chains fold into the 3D structure is known as the protein folding problem. Something that scientists work and have been working for decades.  

Scientists are interested in determining the structures of proteins because of protein's form. The form is thought to dictate its function. If scientists know the form and shape, the role of the cell can be guessed and a step further Is taken towards knowing it. Scientists can then develop the needed drugs that work with the protein's shape

In the past decades, researchers have been working determining shapes of proteins in labs using experimental techniques. These techniques include cryo-electron microscopy, nuclear magnetic resonance and X-ray crystallography.

Each method depends on trial and error. This procedure can take years of work and cost up to hundreds of thousands of dollars per each protein structure. For this exact reason, biologists are looking into AI to help the process. AI brings the ability to predict a protein's shape computationally from its genetic code, instead of determining it long and costly processes. AI can accelerate the research process.

Nowadays, the genomics is rich in data thanks to the reduction in the cost of genetic sequencing. Deep learning appears in the scene to help in the prediction problem of genomic data.

There's a global competition called the Critical Assessment of Techniques for Protein Structure Prediction that works to catalyze research and measure progress on the newest methods working to improve the accuracy of predictions.  

The CASP organizers have named "unprecedented progress in the ability of computational methods to predict protein structure".  The team of DeepMind has been classified as first in the ranking among the teams that entered.

The DeepMind team focused on the problem of modeling target shapes from scratch. A high degree of accuracy has been achieved when predicting the properties of protein structure. 

Both of these methods are based on deep neural networks that are trained to predict the properties of the protein. There are two properties that the networks predict. These are the distances between pairs of amino acids and the angles between chemical bonds that connect the amino acids.

In conclusion, scientists are turning into AI to do protein-based research. AI can accelerate the process in which a protein is analyzed and this can be life-changing for the time that it takes to develop a drug. We have seen how diseases are sometimes related to protein malfunction or deficiency. Based on the Levinthal's paradox, which states that "it would take longer than the age of the universe to randomly enlist the possible configurations of a protein before reaching the 3D structure", scientists have decided to use AI and ML.

3D
AI
AlphaFold
Protein Folding
About the author
Lucas Bonder -Technical Writer
Lucas is an Entrepreneur, Web Developer, and Article Writer about Technology.

Related Articles