An Artificial Intelligence Program to Determine the Nutritional Phase of PWS

Funding Summary

 Individuals with Prader-Willi syndrome (PWS) go through 6 nutritional phases (NPs), but investigators need a standard method of determining the correct NP. Dr. Driscoll has developed an NP questionnaire and an Artificial Intelligence (AI) model to accurately classify the NP in people with PWS, which may lead to a better understanding of PWS and help guide future research and therapies. 

Dr. Theresa Strong, Director of Research Programs, shares details on this project in this short video clip. 

Lay Abstract

Individuals with Prader-Willi syndrome (PWS) go through 6 postnatal nutritional phases (NPs) beginning with poor appetite and feeding as newborns and gradually progressing in childhood to a seemingly unquenchable appetite and lack of feeling full.  Phase 1a is characterized by poor appetite, hypotonia and difficulty feeding.  This phase is followed by phase 1b when the infant’s appetite and feeding have significantly improved, and weight is increasing at a normal rate.  In phase 2a the weight increases without a significant change in appetite or caloric intake, while in phase 2b the weight gain is associated with a concomitant increased interest in food, but the child can feel full.  Phase 3 is characterized by an unquenchable appetite typically accompanied by aggressive food-seeking behavior and lack of feeling full.  Not all individuals with PWS go through all the stages described above, but the vast majority do.  In addition, some adults progress to phase 4 where they no longer have a voracious appetite.
In our original manuscript describing the NPs we listed common characteristics of each phase.  However, each person with PWS is an individual with a different genetic background, family environment, and age of diagnosis.  For these reasons many individuals may not have all the characteristics listed for each NP.  To address these issues investigators need a standard method of determining the correct NP so that we are all speaking the “same language”.  Thus, we have developed a NP questionnaire and an Artificial Intelligence (AI) model to address these issues which will be particularly supportive of research to develop and evaluate new therapeutic approaches and interventions for PWS.  Evaluating individuals by their NP is superior to evaluation by age in order to understand the metabolic, endocrine, and neurophysiologic triggers of each NP.
The first key objective of this project will be validating an AI model we have created that is capable of accurately classifying the NP in people with PWS and then storing that data.  Next is to ensure global accessibility to healthcare professionals, researchers, and clinical trials to ensure that all are using the same criteria to determine the NP of an individual.  The final objectives are to enable the data collected to advance the understanding of PWS and help guide future research and therapies.
We expect that our AI program will lead to a better understanding of the causes of the nutritional phases by enabling researchers doing metabolic, hormonal, biomarker, and neurophysiological studies to use the same criteria to identify the appropriate NP for their studies and thus enable pooling of data among investigators.  This will lead to more informed treatments for PWS.  It will also lead to better education of health care professionals, providers, and families.  The AI program can also be used in clinical trials to help guide researchers in collecting appropriate candidates for research and then evaluate the efficacy of the treatment using our NP scoring system.  It can also be useful in the future for longitudinal registries (e.g., Global PWS Registry), as well as biorepositories like COMBINEDBrain.