How AI Is Drastically Changing the Medical World
Photo Source: Pixabay
Our world keeps changing, but so much of it seems to happen without us even noticing it. The Coronavirus hit the world like a sledgehammer this year, but a profound change that will affect everyone occurred on the last day of November 2020 with little fanfare. Artificial intelligence developed by Google’s DeepMind called AlphaFold2 convincingly bested nearly 100 other teams of scientists and computer engineers at solving a problem that has challenged researchers for over seventy years. With artificial intelligence, scientists have stunning new tools to be able to predict the 3D shape of proteins through their building blocks, amino acids. This will help further along scientific research and speed up drug development.
Protein are often called the building blocks or machinery of the body. They perform many different functions. The body produces thousands of different proteins, all with various jobs from nerve function to immune system response, and to do all these jobs, proteins come in all different shapes and sizes. Proteins appear in many different sizes and shapes depending on the function they perform in our bodies. Collagen protein adds structure to bones and tendons, hemoglobin helps red blood cells transport oxygen throughout our bodies, and keratin makes up hair, skin, and nails. The instructions to make these different kinds of proteins come from specific instructions in our genetic code--DNA. Little protein machines in our cells translate genetic code into long strings of chemicals called amino acids. These long strings of amino acids then fold into proteins. Sometimes the proteins don’t work correctly and can cause disease.
Scientists have studied protein shapes to understand how they work, and healthcare researchers need to understand how proteins work to make medicines to treat diseases. Scientists have developed complicated techniques to predict the 3D structure of proteins using protein crystals, x-rays, and modifying tiny pieces of a protein to see how its function changes. Such techniques require a lot of time, money, and particular expertise. “Each protein behaves quite differently in this respect, and protein crystals can be generated only through exhaustive trial-and-error methods that often take many years to succeed—if they succeed at all.”(Molecular Biology of the Cell. 4th edition)
As noted earlier, proteins start as strings of amino acids built from specific instructions in DNA. The string then folds into a 3D shape to form a mature protein. DNA gives us the first clue of protein structure from the sequence of amino acids it describes. Still, going from two dimensions to three with hundreds, even thousands, of amino acids remains very difficult. Imagine if a 3D structure like a house was delivered to the construction site in a long string of bricks, glass, wood, metal, and ceramic, predicting how to fold that long string to form a house would be almost impossible without some careful instructions.
For many years, scientists have tried to speed up the prediction of protein structures with computers with minimal success until recently. On November 30th, the results of the Critical Assessment of Structure Prediction (CASP) competition revealed that AlphFold2 from Google’s DeepMind deployed artificial intelligence in the contest to beat the other competitors convincingly. Ewen Callaway reports in Nature News that AlphaFold2 achieved nearly the same results in days in accurately predicting protein structures as a laboratory using traditional techniques could achieve in months, even years. Such a dramatic reduction in time to predict protein structure coupled with the significant increase in accuracy shown by AlphaFold2 opens the door to scientists and medical researchers to more quickly make discoveries and develop lifesaving therapies faster than ever before.
Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.
You can buy his book on Amazon in paperback and in kindle format here.