PROJECT SUMMARY/ABSTRACT
The incidence of type 2 diabetes (T2D) is rapidly increasing in youth, particularly in youth of color, and if
uncontrolled can lead to devastating complications as early as 10 years after diagnosis. Yet, there are limited
medication options for youth, who have worse response to available therapies such as metformin compared to
adults. There is therefore a critical need for the development of targeted and effective treatment approaches that
utilize the most appropriate therapeutic option right from the onset of disease. T2D is a heterogenous disease
with variations in mechanistic pathways related to insulin sensitivity, insulin deficiency, obesity and fat distribution
that contribute to disease progression. Methods to subtype individuals with T2D have been developed using
clinical and genomic machine learning based clustering approaches in adults. However, these clustering
approaches have not been evaluated in youth of diverse racial and ethnic backgrounds and using clinical
variables that are routinely measured in clinical practice. In K23 funded work, Principal Investigator Dr.
Srinivasan is evaluating the genetic and pharmacological determinants of metformin response in youth with T2D.
The proposed R03 work will broaden the scope of this work by evaluating the pathophysiological patterns
associated with the development of complications and metformin response in youth, a framework that can be
applied to other T2D medications beyond metformin. In this study, we propose to leverage existing pediatric T2D
datasets and utilize complementary clinical and genetic machine-learning clustering techniques to identify groups
of youth with T2D at highest risk for microvascular complications and most likely to fail metformin treatment,
based on underlying biological mechanisms. In Aim 1, we will categorize 974 youth with T2D from the Treatment
Options for Type 2 diabetes in Adolescents and Youth (TODAY) and SEARCH for Diabetes in Youth (SEARCH)
studies into pathophysiological subgroups based on clinical clusters developed in adults and evaluate the
association of cluster membership with T2D progression, development of microvascular complications and
metformin response. Additionally, we will develop and evaluate novel youth clusters based on routine clinical
variables using an unsupervised machine learning technique and will compare the performance with adult
clusters. In Aim 2, we will construct individual level polygenic scores derived from genetic clustering of T2D loci
and based on mechanistic pathways in TODAY and SEARCH to evaluate the association of genetic scores with
the same outcomes proposed in Aim 1, both alone and in combination with clinical clusters. In Aim 3, we will
validate youth-derived clinical and genomic clusters in a real-world electronic medical record-based youth
dataset from the Boston Children’s Hospital Precision Link Biobank for Health Discovery. This proposal will
generate a prediction model that leverages both routine clinical and genomic data to inform risk stratification and
to improve metformin prescribing. These results, in combination with Dr. Srinivasan’s K23 work will directly inform
a future R01 proposal evaluating precision dosing algorithms for metformin and other drug classes in youth.
