DMS/NIGMS 1: Improving Modern Diagnostics: Detecting and Genotyping of Evolving Viral Pathogens

About This Grant

Infectious disease medicine is poised at the cusp of a revolutionary change in diagnostics, from simple binary diagnostics (presence/absence) directed at individual pathogens to universal tests for all pathogens along with instantaneous genomic characterization. Not only will there be benefits to sick individuals, but the entire population will benefit from better disease monitoring and knowledge of circulating strains. Nevertheless, there is still much work required to achieve genome-level diagnostic information. To achieve high sensitivity at low cost, we assume amplification of input RNA/DNA is necessary. In this context, we are faced with serious computational challenges to design sequence specific oligonucleotides, statistical challenges to account for the biases of sampling when perfect oligonucleotides are inevitably elusive, and diagnostic failures as tests lose sensitivity in the face of evolving pathogens. Recent protocol advances help to separate the confounded biases in traditional sequencing based assays, providing an opportunity for mathematical models to aid in solving the remaining challenges. We propose a novel branching process model exquisitely matched to the emerging sequencing protocols, along with novel statistical estimation methods to link the model to data (Aim 1). Aim 2 proposes to use this system to iteratively optimize a protocol for RNA virus diagnostics. Once initialized, the protocol is expected to be able to solve the oligonucleotide design problem in a data-driven manner. We also propose statistical methods for viral characterization (genotyping) in the face of inevitable biased sampling due to amplification; current methods fail to account for this bias. Aim 3 tackles longer term goals include (1) exploring the generalizability of our findings to related, but distinct protocols, (2) developing methods to monitor and update an active diagnostic protocol in the face of pathogen evolution, and (3) scaling our methods to large diagnostic datasets, such as those produced in diagnostic laboratories, by considering flexible neural networks to model the relationship between sequence and diagnostic detection. RELEVANCE (See instructions): Achieving the goals of this proposal will facilitate the development of many kinds of modern diagnostics for human and animal health. The proposed methodology can assist diagnostic test development, improvement, and maintenance in the face of evolving pathogen threats. Modern diagnostics that monitor infections and the genomic sequence of the infecting agent will facilitate pathogen-personalized treatments and better intervention strategies to protect a vulnerable population.