NCSA Affiliate Halil Kilicoglu Awarded NIH Grant to Improve Randomized Clinical Trials

Randomized clinical trials are a core component to developing medical treatments. However, NCSA affiliate, Halil Kilicoglu, sees a problem with these trials. Without consistent methodology and testing protocols on how the trials are conducted, the results aren’t always the high quality needed to treat patients effectively. In fact, these expensive trials can backfire if done incorrectly, wasting money and possibly harming patients. Kilicoglu, a professor in the School of Information Sciences, has been awarded a $1,328,502 grant from the National Institutes of Health to investigate ways to solve this problem. The project, “Computational Methods, Resources, and Tools to Assess Transparency and Rigor of Randomized Clinical Trials,” will continue for four years.

I am leading a team of computer/information scientists and clinical research methodologists in developing datasets, natural language processing (NLP) methods, and ultimately software tools that will help various stakeholders of biomedical communication assess and improve the reporting quality in randomized clinical trial protocols and result publications.

Halil Kilicoglu, associate professor, School of Information Sciences

NLP is where NCSA’s supercomputing resources will come into play. When fed linguistic data, computers use algorithms that model natural human language so they can understand it. A supercomputer like Delta can use NLP to read vast amounts of linguistic data and analyze it, but it still needs a model to know what it needs to look for. Researchers like Kilicoglu and his team perform that part of the work. With the help of testing protocols developed by Kilicoglu’s team, Delta takes all that data and identifies language within the trials that the team needs to analyze further. Delta is a tool that speeds up the process. Reading through all those trials would take huge teams months of time, perhaps even years. 

The team will also leverage the Radiant OpenStack cluster for interactive inference and the Clowder data management framework for developing the data pipelines and special-purpose clients that will allow users to interact meaningfully with the model outputs.

Being able to leverage Delta for training, Radiant for inference, and Clowder for data management should bootstrap the effort and allow us to focus on the more challenging parts of the project, including the accuracy of the results and the development of intuitive interfaces to interact with the outputs of the model.

Luigi Marini, lead research programmer NCSA

Kilicoglu is pleased to have a resource like NCSA on the campus for a project like this. “The NCSA team will help us turn the computational models we build in our lab to web-based tools that can be used by authors and journal editors and other stakeholders,” he said. NCSA is glad to be a part of such important work. The ultimate goal of this project is to improve clinical care and health policy, which will have a lasting and far-reaching impact.

Read the full release of the story here.

NCSA’s Luigi Marini, lead research programmer, contributed to this story.