OmniScreen
EMSL Project ID
51660
Abstract
We propose OmniScreen, a four stage end-to-end platform for discovery of pathogenic signaturesand early detection of novel pathogens in complex environmental samples. First, OmniScreen will
integrate state-of-the-art cell separation, staining, and sorting techniques to achieve an extraction and
isolation throughput of 106 cells/hr and recovery of at least 80% live cells. Second, OmniScreen will
integrate automated acoustic liquid handling with high-throughput phenotypic screening to create a
direct pipeline for discovery of ideal maintenance and growth conditions for cells of unknown identity.
Third, OmniScreen will interrogate isolated cells with a battery of proteomics, transcriptomics, and
mammalian cell culture adhesion and confluence screens to link physical signatures of self-preservation,
niche-finding, and toxicity with distributional ‘omics measurements. Fourth, OmniScreen will generate
labeled data to train deep neural network classifiers, which will serve to discover nonlinear relationships
between features in heterogeneous data and signatures of pathogenicity. The outcome will be a highthroughput platform for discovery and genome-to-function multi-modal characterization of novel pathogens that includes proteomics and transcriptomic signatures of toxicity, niche-finding, and self preservation.
OmniScreen advances the state of the art in four distinct ways: 1) utilization of automated acoustic
liquid handling to perform multiplexed phenotypic screening, 2) integration of heterogeneous sources
of data coupled with data provenance tracking, 3) leveraging noise-robust and expressive deep
convolutional and multi-layer feedforward networks to construct feature-exposed classifiers, 4) advances
in high-throughput library design in synthetic biology to construct direct reporters and morphological
readouts of the different traits of pathogenicity. Current methods rely solely on the genotyping organisms
to predict pathogenicity. In addition, past approaches assume a reliable extraction, isolation, and
maintenance strategy for bacteria derived from complex matrices or environmental samples.
To date, an end-to-end platform for high-throughput pathogen characterization has been elusive due
to the requirement of a unified but diverse team of subject matter experts, e.g. mammalian biology,
toxicology, soil science, isolation and maintenance, proteomics, metagenomics and metaproteomics,
bioinformatics, algorithms, machine learning, and data engineering. Our team is led by a world expert in
soil microbiome science and supported by a strong interdisciplinary team of chemical biologist and
synthetic biologists, with expertise in mammalian and prokaryotic biology, toxicology, genome science,
activity-based protein profiling, proteomics, transcriptomics, machine learning, deep learning, and
computational biology.
If successful, OmniScreen will provide the first end-to-end platform for rapid, high-throughput
triage and identification of novel pathogens. It will transform the landscape of early detection systems
in disease spread monitoring and pathology. Moreover, it has the potential to preempt pandemics, saving the lives of warfighters and the general public, improving operational readiness of soldiers and providing tremendous savings in health care for disease-ravaged regions of the world.
Project Details
Start Date
2020-10-09
End Date
2021-09-30
Status
Closed
Released Data Link
Team
Principal Investigator
Co-Investigator(s)