Project Spotlight: Integrating Deep Learning with High Throughput Materials Engineering for Detecting Noroviruses

The Norovirus Sensor team, led by Carnegie Mellon’s B. Reeja Jayan and Amir Barati Farimani, are combining various classes of materials with deep learning approaches to develop sensitive and specific sensors. Their unique approach allows them to investigate applications of highly energy-efficient, environmentally benign, and scalable materials in the production of such sensors. Current use cases for these novel sensors include the detection of noroviruses and other pathological viruses that exist in healthcare and other highly trafficked areas.

Please share a little about your background.

Reeja Jayan is an assistant professor at CMU whose work lies at the intersection of materials, chemistry, and engineering. Dr. Jayan was previously a Postdoctoral Associate in Chemical Engineering at the Massachusetts Institute of Technology (MIT) working under the supervision of Professor Karen Gleason. She received her MS in Electrical Engineering and PhD in Materials Science and Engineering from The University of Texas at Austin (UT-Austin), working with Professor Arumugam Manthiram. Amir Barati Farimani is also an assistant professor at CMU whose work focuses on the application of machine learning, data science, and molecular dynamics simulations to health and bio-engineering problems. Dr. Farimani was previously a postdoctoral fellow at Stanford University working under the supervision of Professor Vijay Pande. He received his PhD from the University of Illinois at Urbana-Champaign, receiving the Stanly Wise best thesis award.

What led you to the PHDA?

The Norovirus Sensor project was connected to the PHDA through the CMLH. At the time, the team was looking at developing sensors for gluten – a protein found in foods, primarily in wheat and several grains within that family, with is a common cause of food sensitivity and intolerance. The team realized that they would need to find a way to identify a single protein of interest, in this case gluten, from a complex sample of food that contained a mixture of many proteins. The solution was to develop and apply deep learning methods to distinguish “gluten signals” from other sensor readings. This work led the team to the CMLH and PHDA as it aligned well with the mission to improve health by leveraging big data. Since the original conception, the team has found that some of the technologies that were created for the gluten sensing could be adapted to other molecules – and in fact, to dangerous pathogens like viruses.

Walk us through your project.

Norovirus results in about 685 million cases of disease and 200,000 deaths globally each year and causes about half of the food-borne disease outbreaks that occur in the United States each year. Because norovirus is highly contagious, it is important that settings where patients may be receiving treatment, such as hospitals, or where people may be in close contact, such as childcare centers and cruise ships, are able to prevent contamination and transmission.

In the current project, the team has two thrusts. The first thrust is the data analytics thrust and it involves algorithms created by Dr. Farimani. For this project, the team develops polymers that can be put on substrates like fabrics – which makes the process scalable and also cost-effective. These sensors will develop an electrical output in the presence of specific pathogens of interest – in the original case, the norovirus. The data is then analyzed using Dr. Farimani’s algorithm. The algorithm also helps to track mutations in viruses like the norovirus so that the binding between the virus and certain surface features on the sensory materials can be dynamically tuned and tailored in order to have an updated sensor platform as viruses change. This same approach is being tailored toward COVID-19.

 What are your project’s next steps?

The team is investigating sensing in the environment, on everything from everyday objects in the home to objects that come into the home from outside such as groceries, packages, and more. With COVID-19 detection now a high priority, Dr. Farimani’s team is trying to identify ways in which this virus can be attached to some of the materials in the sensor platform. It’s a new virus, so many of these mechanisms for binding are just newly being discovered.

In what ways has UPMC played a role in lending clinical expertise and sharing data?

UPMC’s clinical expertise has been hugely beneficial in this project. The team has discussed their approach with Dr. Graham Snyder, Medical Director of Infection Prevention for UPMC, and received valuable feedback based on his experiences in a hospital setting and expertise on existing methods for pathogen detection.

In addition, the team is working with a CMU faculty member in robotics, Carmel Majidi, PhD, who is helping integrate these sensors inside a robot that can interact with a room or some kind of environmental setting, inside or outside of a hospital, and collect samples and study them without the need of a more time-sensitive test. The need for quick diagnosis, especially in an emergency situation like a pandemic, has helped the way the team has framed this work. The team was also connected to a UPMC Mercy’s Dr. Mohamed Hamdy Yassin, and has begun discussions on topics related to environmental sampling, ideal materials, and cost and time effectiveness.

How do you and your project partners’ strengths complement each other?

Dr. Jayan and Dr. Farimani have two distinct skill sets. Dr. Jayan focuses on hardware and Dr. Farimani focuses on software and algorithms. These two skill sets are very complimentary and the two depend on each other and their individual expertise to complete the project successfully.