Diabetes Treatment Support

[© Diabetes Treatment Support Developers]


The number of people living with diabetes mellitus is constantly on the rise, from 108 million in 1980 to 422 million in 2014. And it’ll be the seventh leading cause of death in 2030. The need to rapidly and effectively diagnose and treat the illness is crucial. Recommendations in authoritative literature, Clinical Practice Guidelines (CPG)’s are often tailored to a diverse populations, but as patients get more complex they might not align well with supporting evidence. As a result, one of the challenges that physicians face in using CPG recommendations is determining how well the study evidence applies to a general clinical population. We want to assist the physician in identifying gaps within guidelines to make a informed tailored treatment recommendation. Once a patient alignment has been ascertained, we want to help the physician group the patient into different therapies (initial, dual and triple therapy), taking into consideration the patient comorbidities, previous patient history and drug complications. We aim for our system to make the physician aware of these adverse effects of clinically tested drugs - particularly to cardiotoxicity, and also we want to account for drug conflicts with comorbid conditions. Finally, the system will assign provenance (evidence) to the discovered treatment recommendations for the patient, and will be able to monitor the patient’s condition (both prescribed medicine and the prescribed diet) over time



  1. Integrate heterogeneous knowledge sources
  2. Assistance in decision making for complicated patients

Drug recommendation: With a lot of information available on antidiabetic drugs, and a very comprehensive and detailed ADA guidelines to follow, it is hard for a physician that has little time and a wide range of patients, to find a suggested drug tailored to a patient. This system makes this information easy to access.

Toxicity: There is a lot of toxicity information for drugs available on many different sources, Pubchem, DrugBank, ToxBank, however for a user that is not aware of all of the resources or has the time to go through all of the resources, having all of this information combined by semantics, lets the user easily access this information.

Cohort: It is important to know whether a patient that we are suggesting a drug to is represented in the studies that we are getting this suggestion from. It provides provenance and trust to the user that a suggested drug would suit a particular patient.

Integration of the above parts: having all of this information connected and easy to use is very helpful. A physician not only needs to know a suggested drug for a particular patient profile, they also need to know possible side effects of it, and whether the patient they are recommending to is well represented in the studies they are using. All of this information combined in one ontology, makes our ontology very powerful and useful. Combining the parts together, there are also various portions that are interconnected. For instance, toxicity of a drug can also be based on particular clinical trials on humans - our system can connect this to a cohort and compare our patient to this cohort as well.


  • Bhanushee Sharma : sharmb3 at rpi dot edu
  • Neha Keshan : keshan at rpi dot edu
  • Nkechinyere Nneka Agu : agun at rpi dot edu
  • Shruthi Chari : charis at rpi dot edu
  • Shweta Narkar : narkas at rpi dot edu
  • We would be delighted to have people use our ontologies, review the results and contribute to our work. Several components of this project reflect ongoing research, and although the course has ended we expect the work to be reused and extended over the next several years. If you are interested in getting involved, please review our Getting Involved and Maintenance Policy page and feel free to reach out to us with any feedback or to let us know you would like to work with us on any aspect of the project.


    We would like to thank Prof. Deborah McGuiness, Ms Elisa Kendall, Dr Jim McCusker for their continuing support throughout the course of this project. We would also like to thank Dr Oshani Seneviratne for her inputs and Dr Amar Das from IBM Research, Prof. Jonathan S. Dordick and Dr. Keith Fraser of Center for Biotechnology and Interdisciplinary Studies at Rensselaer Polytechnic Institute.

    This was developed as a part of the Ontology Engineering course supervised by Prof. Deborah McGuinness and Ms. Elisa Kendall at RPI in Fall'18