SUMMARY This project endeavors to build a nanosensor array platform technology to detect whole disease fingerprints from patient biofluids to facilitate diagnosis, screening, and biomarker discovery efforts. Serum biomarker measure- ments are widely used as diagnostic indicators, but many markers are not sufficient for assessments of disease state. Major factors limiting diagnosis and screening using most biomarkers include their low specificity for dis- eases and the overall dearth of established molecular markers. Innovative approaches are needed to identify new biomarkers and/or improve screening and diagnostic efforts in the absence of validated biomarkers. We believe that the differentiation of diseased from normal biofluids may be achieved by the detection of a “disease fingerprint” through the collection of large data sets of molecular binding interactions to a diverse set of moder- ately-selective sensors, which are used to train machine learning algorithms. We will build a sensor array com- prising organic color centers (OCCs, covalently-modified carbon nanotubes) to transduce subtle differences in physicochemical properties of molecules in biofluids. With sufficient diversity, the sensors can differentiate bi- ofluids by disease status with the aid of machine learning processes. In preliminary experiments, we found that a library of OCC-DNA nanosensors exhibited sensitive and differentiated spectral variation to probe an ensemble of molecular binding events. Via machine learning algorithms, we built a prediction model of nanosensor re- sponses that reliably identified high-grade serous ovarian cancer (HGSC) substantially better than the estab- lished, FDA-approved biomarker, CA125, using an initial set of 264 patient serum samples (Nat Biomed Eng, 2022). Despite advances in the understanding and management of HGSC, survival is currently poor when diag- nosed at later stages, and detection is uncommon at early stages. Surgery is the first-line treatment, and cancer recurs in 70% of patients in remission. Secondary surgery can prolong survival but only if performed early enough to enable complete resection. Improved detection of early-recurrent and early-stage HGSC would therefore markedly increase survival rates. We plan to develop a robust diagnostic sensor platform to improve early de- tection of ovarian cancer and recurrence, and to accelerate biomarker discovery processes. Additionally, quan- titative analysis of proteins bound to the sensors can determine the unique pattern of protein adsorption respon- sible for the disease-specific spectral responses, thereby potentially facilitating biomarker discovery. We propose to investigate: 1) the diversity of molecular sensitivities of OCC-DNA nanosensor elements required to differen- tiate patient samples, 2) machine learning-based classification of disease, focusing on early-recurrence and early-stage HGSC, 3) the molecular mechanism of the sensor response, and 4) the potential of the array to facilitate identification of novel biomarkers. Successful completion of this work will result in a validated platform to enable concomitant identification of disease and acceleration of biomarker discovery processes in HGSC, with applicability to many potential indications.

NARRATIVE Serum biomarker measurements have been widely used as diagnostic indicators, but many markers are not sufficiently sensitive or specific assessments of disease state. This work builds a platform technology to detect whole disease fingerprints from patient biofluids to improve diagnosis, screening, and biomarker discovery processes. We will initially focus this effort on the outstanding need for improved detection of high-grade serous ovarian cancer.