12-16-2020, 10:10 AM
Imagine a world where plastic waste doesn’t clog up our life-giving rivers but is instead broken down into its constituent parts and returned to the earth.
Imagine a world where carbon dioxide is captured from the atmosphere, reducing global temperatures and helping mitigate climate change.
Imagine a world where new diseases are deciphered and drugs devised in a matter of weeks, not years or decades.
These are some of the many ways the world could transform for the better, thanks to the artificial intelligence (AI) breakthrough that took place slightly more than two weeks ago when an AI program designed by Google DeepMind solved one of science’s grand challenges: the protein folding problem.
It was such a profound advance that it moved Andrei Lupas, an evolutionary biologist at the Max Planck Institute for Developmental Biology to call it a game-changer. He added: “This will change medicine. It will change research. It will change bioengineering. It will change everything.”
Mohammed AlQuraishi, a computational biologist at Columbia University reiterated this, gushing: “It’s a breakthrough of the first order, certainly one of the most significant scientific results of my lifetime.”
For some context, a grand challenge is a difficult and highly consequential but unsolved scientific or technological problem. One of the grand challenges in physics is the search for a grand unified theory – one that reconciles quantum mechanics with general relativity; while a grand challenge in climate change is the development of feasible carbon capture technology.
Similarly, the protein folding problem is a grand challenge in biology – one that has eluded scientists for half a century. Until now that is.
But before we delve into the details of the breakthrough, we need to understand what a protein is. I know what you’re thinking. You know what protein is – it’s what you’re told to consume copious amounts of when you’re hitting the gym right? Well yes, but it’s much more than that. So much more.
Proteins are the building blocks of life. They are a chain of amino acids, each of which is made up of carbon, oxygen, hydrogen and nitrogen atoms, although some also have sulfur atoms and one even has a selenium atom. In total, there are 20 types of amino acids – which in different combinations make up all the proteins that are a part of you and me.
Proteins are vital to a host of things, including digesting food (amylase, pepsin), transporting substances in the body (haemoglobin), defending the body against pathogens (immunoglobulin), and making muscles contract (actin, myosin).
Rest assured, we would not exist if not for these phenomenal molecular wonders that animate us.
The function of a specific protein is dictated by its complex 3D structure – something that at a quick glance looks like confetti. An oft-quoted maxim in molecular biology is “structure is function”. Traditionally, the shape of a protein was determined by expensive and laborious techniques like X-ray crystallography and nuclear magnetic resonance. This is why even though we know of around 200 million proteins across many lifeforms, we’ve only determined the 3D structure of around 170,000 of them – a minuscule fraction.
However, a long-standing hypothesis in the field holds that the structure of a protein should be able to be determined using only its amino acid sequence. This prophetic postulation propelled scientists to pore through the data, looking to determine the structure and by consequence, the function, of a host of proteins using only their amino acid sequences. However, their decades-long effort only resulted in middling results.
But all that changed when AlphaFold 2 – an AI system designed by DeepMind – entered the fray. It’s the same company that created AlphaGo, the AI system that bested Go grandmaster Lee Sedol in 2016, sending shockwaves through many in the AI world who thought that such a monumental feat was at least a decade away. For context, the game of Go is many times more complex and nuanced than chess and was thought by some to require human intuition to master, unlike chess which is bound by strict logic.
In this year’s Critical Assessment of Protein Structure Prediction (CASP) competition – the Olympics of protein folding if you will – DeepMind’s AlphaFold 2 algorithm came in head and shoulders above the competition and essentially solved the problem of protein folding.
It achieved a median score of 92.4 GDT (Global Distance Test) out of 100. Anything above 90 is considered solved. It is so mind-bogglingly accurate that the predictions only have an average error of the width of an atom (0.1 nanometres).
This has numerous potential uses, including two major ones. It could herald a revolution in the drug discovery process, accelerating the synthesis of life-saving drugs for a host of diseases, including Covid-19. Even more excitingly, it could aid in the design of novel proteins that could degrade plastic, produce biofuels, and capture carbon in the atmosphere – addressing two of the biggest environmental problems plaguing us today – plastic pollution and climate change.
Its incredible potential impelled Arthur Levinson, Founder and CEO of Calico to proclaim: “AlphaFold is a once in a generation advance, predicting protein structures with incredible speed and precision. This leap forward demonstrates how computational methods are poised to transform research in biology and hold much promise for accelerating the drug discovery process.”
The stunning result achieved by DeepMind is a fantastic example of the wide-ranging and consequential ways AI is and will continue to transform the world we live in. It’s a potent tool that will not only help computing advance by leaps and bounds but will also be the next great tool in scientific inquiry and technological advancement.
In a decade, specialised AI programmes will become just as indispensable as Microsoft Excel, AutoCad, and Adobe Photoshop are to professionals in their respective fields. It will bleed into every industry, reducing our workload, increasing productivity but also invariably increasing unemployment.
Just as a telescope opens up new never-before-seen vistas of space and a microscope opens up an almost infinitesimally small universe, AI will open up portals through which we can view the world in a different, more nuanced light. It’ll reveal things in better, clearer detail, and stumble into solutions to problems we didn’t even know we had.
I expect AI to herald a new age of scientific discovery, becoming the preeminent sandbox for scientists. In addition, it’ll help create mechanistic entities that blur the line between life and non-life, as Boston Dynamics and some other companies have already started to demonstrate.
And it’s best we Malaysians get on board this high-speed train into the future as soon as possible. It waits for no one.
The writer can be contacted at kathirgugan@gmail.com.
Source: FMT
Imagine a world where carbon dioxide is captured from the atmosphere, reducing global temperatures and helping mitigate climate change.
Imagine a world where new diseases are deciphered and drugs devised in a matter of weeks, not years or decades.
These are some of the many ways the world could transform for the better, thanks to the artificial intelligence (AI) breakthrough that took place slightly more than two weeks ago when an AI program designed by Google DeepMind solved one of science’s grand challenges: the protein folding problem.
It was such a profound advance that it moved Andrei Lupas, an evolutionary biologist at the Max Planck Institute for Developmental Biology to call it a game-changer. He added: “This will change medicine. It will change research. It will change bioengineering. It will change everything.”
Mohammed AlQuraishi, a computational biologist at Columbia University reiterated this, gushing: “It’s a breakthrough of the first order, certainly one of the most significant scientific results of my lifetime.”
For some context, a grand challenge is a difficult and highly consequential but unsolved scientific or technological problem. One of the grand challenges in physics is the search for a grand unified theory – one that reconciles quantum mechanics with general relativity; while a grand challenge in climate change is the development of feasible carbon capture technology.
Similarly, the protein folding problem is a grand challenge in biology – one that has eluded scientists for half a century. Until now that is.
But before we delve into the details of the breakthrough, we need to understand what a protein is. I know what you’re thinking. You know what protein is – it’s what you’re told to consume copious amounts of when you’re hitting the gym right? Well yes, but it’s much more than that. So much more.
Proteins are the building blocks of life. They are a chain of amino acids, each of which is made up of carbon, oxygen, hydrogen and nitrogen atoms, although some also have sulfur atoms and one even has a selenium atom. In total, there are 20 types of amino acids – which in different combinations make up all the proteins that are a part of you and me.
Proteins are vital to a host of things, including digesting food (amylase, pepsin), transporting substances in the body (haemoglobin), defending the body against pathogens (immunoglobulin), and making muscles contract (actin, myosin).
Rest assured, we would not exist if not for these phenomenal molecular wonders that animate us.
The function of a specific protein is dictated by its complex 3D structure – something that at a quick glance looks like confetti. An oft-quoted maxim in molecular biology is “structure is function”. Traditionally, the shape of a protein was determined by expensive and laborious techniques like X-ray crystallography and nuclear magnetic resonance. This is why even though we know of around 200 million proteins across many lifeforms, we’ve only determined the 3D structure of around 170,000 of them – a minuscule fraction.
However, a long-standing hypothesis in the field holds that the structure of a protein should be able to be determined using only its amino acid sequence. This prophetic postulation propelled scientists to pore through the data, looking to determine the structure and by consequence, the function, of a host of proteins using only their amino acid sequences. However, their decades-long effort only resulted in middling results.
But all that changed when AlphaFold 2 – an AI system designed by DeepMind – entered the fray. It’s the same company that created AlphaGo, the AI system that bested Go grandmaster Lee Sedol in 2016, sending shockwaves through many in the AI world who thought that such a monumental feat was at least a decade away. For context, the game of Go is many times more complex and nuanced than chess and was thought by some to require human intuition to master, unlike chess which is bound by strict logic.
In this year’s Critical Assessment of Protein Structure Prediction (CASP) competition – the Olympics of protein folding if you will – DeepMind’s AlphaFold 2 algorithm came in head and shoulders above the competition and essentially solved the problem of protein folding.
It achieved a median score of 92.4 GDT (Global Distance Test) out of 100. Anything above 90 is considered solved. It is so mind-bogglingly accurate that the predictions only have an average error of the width of an atom (0.1 nanometres).
This has numerous potential uses, including two major ones. It could herald a revolution in the drug discovery process, accelerating the synthesis of life-saving drugs for a host of diseases, including Covid-19. Even more excitingly, it could aid in the design of novel proteins that could degrade plastic, produce biofuels, and capture carbon in the atmosphere – addressing two of the biggest environmental problems plaguing us today – plastic pollution and climate change.
Its incredible potential impelled Arthur Levinson, Founder and CEO of Calico to proclaim: “AlphaFold is a once in a generation advance, predicting protein structures with incredible speed and precision. This leap forward demonstrates how computational methods are poised to transform research in biology and hold much promise for accelerating the drug discovery process.”
The stunning result achieved by DeepMind is a fantastic example of the wide-ranging and consequential ways AI is and will continue to transform the world we live in. It’s a potent tool that will not only help computing advance by leaps and bounds but will also be the next great tool in scientific inquiry and technological advancement.
In a decade, specialised AI programmes will become just as indispensable as Microsoft Excel, AutoCad, and Adobe Photoshop are to professionals in their respective fields. It will bleed into every industry, reducing our workload, increasing productivity but also invariably increasing unemployment.
Just as a telescope opens up new never-before-seen vistas of space and a microscope opens up an almost infinitesimally small universe, AI will open up portals through which we can view the world in a different, more nuanced light. It’ll reveal things in better, clearer detail, and stumble into solutions to problems we didn’t even know we had.
I expect AI to herald a new age of scientific discovery, becoming the preeminent sandbox for scientists. In addition, it’ll help create mechanistic entities that blur the line between life and non-life, as Boston Dynamics and some other companies have already started to demonstrate.
And it’s best we Malaysians get on board this high-speed train into the future as soon as possible. It waits for no one.
The writer can be contacted at kathirgugan@gmail.com.
Source: FMT