By Lauren Hines | Bond LSC
Artificial intelligence (AI) can do more than just write a book given a few words. It can help make cancer treatments more effective and predict the presence of disease in cells, which doctoral student Clement Essien did through his recent project.
“It’s exciting because for several years, I was a software engineer, and then I felt like I wanted to do something more with that,” Essien said. “I want to make some contribution to understanding disease and also in the diagnosis and treatment of diseases. So, I had to look for ways that I could apply computing to understand and possibly solve many biological problems.”
Essien — who works with Bond LSC principal investigator Dong Xu — is trying to predict the binding sites of metal-binding proteins called metalloproteins using AI technology, specifically deep learning.
“Metal binding can play a very functional role,” said Dong Xu, Shumaker Endowed Professor in Bioinformatics, Director of Information Technology Program. “That is why it’s important to know whether a protein binds to a particular metal, and also if you really could, you’d want more details on where it binds.”
Even though Essien’s work is still underway, it has big implications.
“[My work] helps to advance the research geared towards improving the prediction capability of machine learning modules that work on this problem and also provide an important step towards understanding protein functions, and their implications for gene product characterization, drug design for certain diseases and enzyme engineering.
Not only could predicting the binding sites of metal proteins help create drug targets and advance other research, but it could also possibly help identify the presence of disease.
By predicting the binding site, researchers can figure out the protein’s structure and therefore infer the function.
“[The function is] what tells you what role the protein plays in the body,” Essien said. “Also, the presence or absence or mutation of a particular protein sequence can cause diseases.”
Essien’s previous project had the same goal of predicting zinc binding sites, but now he is expanding his research to the study of many other metals found in the human body. Essien is also using an AI technique called Natural Language Processing (NLP).
One use for NLP would be to give an AI all the words in the dictionary and then ask it to write a book. In Essien’s work, he is trying to model the protein sequences as a text because they consist of a sequence of letters, and then get the AI to learn representations from it.
“So, we are trying to model the problem as a natural language problem in the sense that those series of letters you see in the protein sequences, they may be represented as words,” Essien said. “So, if you’re able to break that code you might be able to learn some important things.”
Essien published one paper in a conference and expects to have another one out in a few months. Although, he has much more to discover until then.
“It’s one thing to see I’m able to predict these to a certain accuracy, but it’s also important to learn what is going on inside,” Essien said. “Is there any new significance to learn?”