ABSTRACT
Novel approaches for early and accurate diagnosis of COVID-19 associated syndromes and
evaluation of clinical severity and outcomes of COVID-19 disease in children are urgently needed.
The overarching goal of this grant proposal is to develop clinical assays that can evaluate and predict
severity of pediatric COVID-19 disease, ranging from asymptomatic or mildly symptomatic to severe
manifestations such as multisystem inflammatory syndrome (MIS-C). To date, we have collected and
biobanked clinical samples from more than 400 patients across 3 academic hospitals, including
approximately 100 patients with MIS-C. In the first R61 phase of this project, we will continue to enroll
patients with pediatric COVID-19 and MIS-C for sample collection and longitudinal chart review and
testing (Aim 1), leverage machine learning to identify diagnostic and prognostic “omics” host
biomarkers based on RNA transcriptome profiling from nasal swab and whole blood samples (Aim 2)
and cell-free DNA analysis from plasma (Aim 3), and generate predictive models of clinical severity
and outcomes by incorporating longitudinal clinical, laboratory, viral, and omics data (Aim 4). Our
rationale for including these samples is that they are routinely obtained in hospitals and clinics and
permit easy and noninvasive collection without any special processing or handling requirements,
which will accelerate the development of omics-based clinical assays. Our Go/No-Go transition
milestones for transition to the R33 phase after 2 years include: (1) collection of longitudinal samples
from a minimum of 120 patients for each identified presentation (mildly symptomatic outpatient,
severely ill in the ICU, and MIS-C) and a comparable number of matched controls, (2) generation of
panels of candidate of severity and confirmation of a subset of biomarkers by qPCR, (3) development
of classifier models using machine learning using the biomarkers alone (for clinical assay
development), and (4) combining these omics biomarkers with additional clinical, viral, and laboratory
biomarkers into combined classifier models using machine learning. For the classifier models, the
minimum/goal performance requirements would be 70%/>80% sensitivity and 80%/>90% specificity.
In the second R33 phase, we propose to develop host-based clinical assays for diagnosis and
severity prediction of COVID-19-associated syndromes, including MIS-C, in children from nasal
swabs and blood (Aim 5) and validate these biomarker panels as a Laboratory Developed Test (LDT)
in a CLIA (Clinical Laboratory Improvement Amendments) diagnostic laboratory (Aim 6). These
assays will be evaluated for accuracy, precision, reproducibility, limits of detection (LOD), matrix
effect, interference, among other performance characteristics. We will work closely with the RADx-rad
Data Coordination Center (DCC) on assay development, testing, and validation for submission to the
FDA for Emergency Use Authorization (EUA) and timely deployment of these assays for clinical use.