Machine Learning Guided Precision Genetic Testing for Identification of Monogenic Cardiovascular Disorders

NIH RePORTER · NIH · R01 · $756,895 · view on reporter.nih.gov ↗

Abstract

Monogenic disorders individually are rare but in aggregate are common, with estimates of up to 1 in 40 individuals afflicted. In monogenic cardiovascular (CV) disease, early identification of at-risk individuals has proven utility for prevention. Unfortunately, there is substantial clinical variability not only in identification of monogenic CV diseases but also in therapeutic interventions for these patients, in part because of their underdiagnosis. As such, there is a need for development of more targeted approaches for identification of monogenic CV disorders, but that go beyond traditional diagnostic models that result in underdiagnosis. There is an opportunity and need to integrate cutting-edge, high-throughput phenotyping approaches with genetics to identify patients at high risk, or already showing evidence of monogenic CV disorders, and determine if this approach can improve the care of these patients. The overall objective for this proposal is to develop, validate and determine clinical utility of a precision genetic testing approach guided by machine learning (ML)-based models. We hypothesize that ML electronic health record (EHR), genetic and imaging algorithms coupled with precision genetic testing will lead to enhanced diagnosis of hypertrophic cardiomyopathy (HCM) and transthyretin amyloidosis (ATTR-CM). We will accomplish this objective through the following aims: Aim 1. Refine, validate and determine generalizability of imaging-based deep learning (DL) algorithms for HCM and ATTR-CM using echocardiograms across four health systems caring for diverse patients (Duke, MUSC, Mayo Clinic, Cedars-Sinai); Aim 2. Aim 2. Develop, validate and determine generalizability of an EHR-based computable phenotype for HCM and ATTR-CM in four health systems; Aim 3. Identify HCM and ATTR-CM genetic VUS with high evidence for pathogenicity using a ML- algorithm; and Aim 4. Determine feasibility and clinical utility of a precision genetic testing approach integrating a DL-imaging algorithm, an EHR computable phenotype and genetic testing. This proposal holds great potential for demonstrating clinical utility of this approach, with an output of a coordinated system that could be scaled within other health systems.

Key facts

NIH application ID
10978288
Project number
1R01HL168940-01A1
Recipient
DUKE UNIVERSITY
Principal Investigator
Svati H. Shah
Activity code
R01
Funding institute
NIH
Fiscal year
2024
Award amount
$756,895
Award type
1
Project period
2024-07-01 → 2028-05-31