AI analyzing NBA player movement may help develop cancer treatments


pharmacy hours interviewed Dong Xu, PhD, MS, Distinguished Professor of Curators, Department of Electrical Engineering and Computer Science, University of Missouri College of Engineering, about research evaluating the application of a form of artificial intelligence (AI) to assist scientists in the development of new drug therapies medical treatments for cancer and other diseases.

Alana Hippensteele: Hello, I’m Alana Hippensteele with pharmacy hours. I am joined by Dong Xu, PhD, MS, a Distinguished Professor of Trustees in the Department of Electrical Engineering and Computer Science at the University of Missouri College of Engineering, who is here to discuss research intended to apply some form of artificial intelligence . previously used to analyze how NBA players move their bodies to help scientists develop new drug therapies for medical treatments for cancer and other diseases.

So, dr Xu, what kind of AI is the University of Missouri using in this drug development research and how can it help scientists with this research?

Dong Xu: Yes, so it’s a special kind of AI. So AI is a very big umbrella, and then within that big umbrella there’s a branch called machine learning, I think most people have probably heard of it. Then, in the field of machine learning, there is actually a new type of machine learning called deep learning. Then, within deep learning, there is a methodology called the Graph Neural Network. So we use the Graph Neural Network.

So, essentially, the feature of the neural graph network is to accommodate different identities and relationships. So in our case we look at different atoms or amino acids, how they interact with each other, how the dynamics change, and then we build the graph. Basically in the graph we have the node representing an amino acid and then between 2 nodes the interaction between those 2 amino acids. So that’s kind of a special method that we use.

Alana Hippensteele: That’s fascinating. How has this AI been previously used to analyze how NBA players move their bodies, and what was the impact of this research?

Dong Xu: Yes, so we actually borrowed that methodology. For other areas they use this method for example to study NBA players. They do it by finding some representative points on a player and then seeing between those points what kind of original patterns they have [there are]and then they can deduce a lot of information based on the dynamics of that game.

For example, they can see with which hand they are actually carrying a ball, but also the performance and how different players coordinate with each other. So it’s actually very informative to infer from a video without human intervention, and then you input a video and the output is an analysis of that piece.

Then this was used for another analysis, for example in astronomy or in the world of atoms, [looking at] how they interact with each other. But nobody has used this method for biological studies. So we’re the first to use it to study proteins.

Alana Hippensteele: That’s very, very interesting. How can this AI help to develop new drug therapies for medical treatments against cancer?

Dong Xu: Yes, well. As we do, our input is the trajectory of the molecular dynamics simulation of proteins. So this cancer simulation has been in science for many years. The fact is I did it as a PhD student about 30 years ago to basically simulate how protein moves, but it’s actually not easy to analyze search results. So you see a lot of movement, but you don’t really see much insight into the dynamics, and then our method can really give insight into the protein interaction mode. That is, between these amino acids, how they interact with each other, how they coordinate with each other.

In fact, in this article we have examined a very special type of interaction called an allosteric effect that occurs in a protein – this type of interaction is very interesting. So you have a site that can interact with a drug or another protein, but that interaction is being affected by another site on a protein that’s quite distant. Therefore, it is very difficult to understand how these remote locations influence this allosteric effect.

So people actually ran a simulation to try and figure it out, but it’s very difficult to figure out how this interaction works [sic] [impacts the] protein, and this method that we developed can really project how that distant site affects that binding site. We can tell which amino acids are involved. Therefore, it is very useful to study this way. Based on that we can actually design an intervention strategy, for example people used to design drugs pretty much at the binding site, but since we can find a pass it means we can potentially look at the intermediate sites at those sites for alternative strategies for the drug development.

Alana Hippensteele: That’s fascinating. What other treatment goals can this AI help with by developing therapies?

Dong Xu: Yes, so a by-product of methods that allow us to predict changes in amino acids fairly accurately in terms of impact. So let’s say you have a protein with amino acids, and then some amino acids could be changed, we call that a mutation. Indeed, it is not easy to predict the effects of mutations. But this method can predict energy changes for them fairly accurately, much more accurately than our previous methods.

With that we can first look at the effects of mutations in diseases, for example cancer, in a lot of these cancers the proteins have mutations, some mutations play an important role, some may not, and with this method we can really see which ones Changes that can alter protein stability or other things. So that can potentially give us a drug target where we should target the protein. So it can provide some kind of insight, although it’s not a direct drug development, it’s a general method. So I have to say that this is not a drug development method, but it will support part of the drug development process.


Comments are closed.