The Salivary Raman COVID-19 Fingerprint

BACKGROUND/RATIONALE: The outbreak of coronavirus disease 2019 (COVID-19), caused by
infection of SARS-CoV-2, has rapidly spread to become a worldwide pandemic. Global research
focused on the understanding of the biochemical infective mechanism and on the discovery of a
fast, sensitive and cheap diagnostic tool, able to discriminate the current and past
SARS-CoV-2 infections from a minimal invasive biofluid. The fast diagnosis of COVID-19 is
fundamental in order to limit and isolate the positive cases, decreasing with a prompt
intervention the infection spreading. Moreover, the prediction of the respiratory infection
severity could be of crucial importance for the fast identification and discrimination
between mild clinical course, severe illness, and Acute Respiratory Distress Syndrome (ARDS).
One of the first infection sites of SARS-CoV-2 is the oral cavity where the virus is able to
bind and penetrate through the ACE2 receptors present on the epithelial cells of the salivary
glands. Thus, a high concentration of virus particles could be found in saliva in the
preliminary phases of the infection. Saliva is a complex biofluid composed of bioactive
molecules that can be collected with a really minimal-invasive procedure. Raman spectroscopy
is a non-invasive, fast and label-free vibrational technique, able to provide information
regarding presence, concentration, environment, modifications and interactions of all the
biochemical species present in a specific biofluid. Using the Raman spectroscopy, the
investiators will analyze saliva collected from healthy subjects, patients affected by
COVID-19 and subjects with a past infection by COVID-19. The data collected will be analyzed
and used to create a Raman database able to provide a classification model based on machine
learning. The possibility to monitor and characterize a potential salivary COVID-19
fingerprint could be of crucial importance for the monitoring and discrimination of COVID-19
subjects with a current and past infection from the healthy subjects.

OBJECTIVES: The aim of the project is to characterize and validate the salivary Raman
fingerprint of COVID-19, understanding the principal biomolecules involved in the differences
between the three experimental groups: 1) healthy subjects, 2) COVID-19 patients and 3)
subjects with a past infection by COVID-19. The large amount of Raman data will be used to
create a salivary Raman database, associating each data with the relative clinical data
collected. The Raman database will be used for the creation of a classification model through
the application of multivariate analysis in terms of principal component analysis and linear
discriminant analysis. This classification model will provide a fast tool for the
discrimination of the COVID-19 condition, potentially providing also information on the
respiratory clinical course of the patient. The model will be translated for the application
to a portable Raman spectrometer, leading to the creation of a Raman Point of Care METHODS:
Starting from the preliminary results and protocols of the Laboratory of Nanomedicine and
Clinical Biophotonics (LABION) - IRCCS Fondazione Don Gnocchi Milano, the saliva collected
from each experimental group will be analysed using Raman spectroscopy. All the data will be
processed for the baseline, shift and normalization in order to homogenize the signals
collected and creating in this way the Raman database. The average spectrum calculated from
each group will be characterized, identifying the principal families of biological molecules
responsible for the spectral differences. Consecutively, all the spectra will be processed
through multivariate analysis (principal component analysis and linear discriminant analysis)
obtaining in this way the classification model. LOOCV will be used for the training of the
classification model, which will be questioned using the subset validation analysis. The
partial correlation coefficient (Pearson's and Spearman's correlation) will be used for the
Raman correlation with the clinical parameter (e.g. COVID-19 clinical course) using as
control covariates the age and sex of the subjects. The classification model will be then
translated and used as point of care using a portable Raman equipped with a laser emitting at
785 nm, with a comparable spectral resolution.

- SAMPLE COLLECTION: Saliva will be collected with Salivette (SARSTEDT, Germany),
following the manufacturer's instructions. The cotton swab will be inserted in the
subject mouth and chewed for 60 seconds. The saliva collection will be achieved through
centrifugation of the swab (1000 g x 2 min), recording all the related parameters
(storage temperature and the time between collection and analysis). All the collection
procedures will be performed at least two hours after the last meal and teeth brushing.

- SAMPLE PROCESSING: Before the analysis, saliva (3 ul) will be deposited on aluminum foil
and dried overnight. The aluminum foil is fundamental to achieve the Surface Enhanced
Raman Scattering, increasing the saliva Raman signal.

- DATA COLLECTION: Raman spectra will be acquired using an Aramis Raman microscope (Horiba
Jobin-Yvon, France) equipped with a laser light source operating at 785 nm with 100%
(512mW) laser power. Acquisition time will be set at 30 seconds with double acquisition
and 2 seconds delay time to prevent the formation of artifact spectra. Before each
analysis, the instrument will be calibrated using the reference band of silicon. All the
signals will be acquired in the region between 400 and 1600 cm-1 with a resolution of
0.8cm-1, acquiring at least 25 spectra following a square-map. The software package
LabSpec 6 (Horiba Jobin-Yvon) will be used for map design and the acquisition of
spectra.

- DATA ANALYSIS: All the data will be fit using a fourth-degree polynomial curve to set
the baseline and consecutively normalized using unit vector. The contribution of
aluminum will be removed from each spectrum. The statistical analysis will be performed
using the multivariate approach. Briefly, principal component analysis and linear
discriminant analysis will be applied to extract the principal components and the
canonical variables. These features will be used for the leave one out cross-validation
(LOOCV), subset validation and correlation with the clinical parameters. Mann-Whitney
will be performed on PCs scores to verify the differences statistically relevant between
the analysed groups. The analysis will be performed using Origin software (OriginLab,
USA)

- CORRELATION: Partial correlation with Pearson's and Spearman's coefficients will be
performed on the variables extracted and the clinical parameters, using as control
covariates the age and sex of the subjects. Only values with p < 0.001 will be
considered as statistically relevant.

- TRANSLATION: The data and the classification model will be applied with a portable Raman
equipped with a laser emitting at 785 nm and with a spectral resolution comparable with
the one used for the previous analysis.