A non-invasive ultrasensitive diagnostic approach for COVID-19 infection using salivary label-free SERS fingerprinting and artificial intelligence

• Published in: Journal of photochemistry and photobiology. B, Biology Vol. 234; p. 112545
• Main Authors: Karunakaran, Varsha, Joseph, Manu M., Yadev, Induprabha, Sharma, Himanshu, Shamna, Kottarathil, Saurav, Sumeet, Sreejith, Remanan Pushpa, Anand, Veena, Beegum, Rosenara, Regi David, S., Iype, Thomas, Sarada Devi, K.L., Nizarudheen, A., Sharmad, M.S., Sharma, Rishi, Mukhiya, Ravindra, Thouti, Eshwar, Yoosaf, Karuvath, Joseph, Joshy, Sujatha Devi, P., Savithri, S., Agarwal, Ajay, Singh, Sanjay, Maiti, Kaustabh Kumar
• Format: Journal Article
• Language: English
• Published: Switzerland Elsevier B.V 01.09.2022; Elsevier BV
• Subjects: Algorithms; Artificial Intelligence; Coronaviruses; COVID-19; COVID-19 - diagnosis; COVID-19 Testing; Delivery of Health Care; Diagnosis; Diagnostic systems; Fingerprinting; Free surfaces; Genetic diversity; Humans; Infections; Label-free; Machine learning; Mutation; Nucleic Acids; Patients; Raman spectra; Reagents; Recovery; Recovery (Medical); Saliva; SARS-CoV-2; Severe acute respiratory syndrome coronavirus 2; Spike protein; Support vector machines; Surface enhanced Raman spectroscopy; Test procedures; Viral diseases; Viruses; COVID-19; SERS; Surface enhanced Raman spectroscopy; AuNPs; Diagnosis; SVM; Label-free; Saliva; Artificial intelligence; PCA
• ISSN: 1011-1344; 1873-2682
• DOI: 10.1016/j.jphotobiol.2022.112545

Clinical diagnostics for SARS-CoV-2 infection usually comprises the sampling of throat or nasopharyngeal swabs that are invasive and create patient discomfort. Hence, saliva is attempted as a sample of choice for the management of COVID-19 outbreaks that cripples the global healthcare system. Although limited by the risk of eliciting false-negative and positive results, tedious test procedures, requirement of specialized laboratories, and expensive reagents, nucleic acid-based tests remain the gold standard for COVID-19 diagnostics. However, genetic diversity of the virus due to rapid mutations limits the efficiency of nucleic acid-based tests. Herein, we have demonstrated the simplest screening modality based on label-free surface enhanced Raman scattering (LF-SERS) for scrutinizing the SARS-CoV-2-mediated molecular-level changes of the saliva samples among healthy, COVID-19 infected and COVID-19 recovered subjects. Moreover, our LF-SERS technique enabled to differentiate the three classes of corona virus spike protein derived from SARS-CoV-2, SARS-CoV and MERS-CoV. Raman spectral data was further decoded, segregated and effectively managed with the aid of machine learning algorithms. The classification models built upon biochemical signature-based discrimination method of the COVID-19 condition from the patient saliva ensured high accuracy, specificity, and sensitivity. The trained support vector machine (SVM) classifier achieved a prediction accuracy of 95% and F1-score of 94.73%, and 95.28% for healthy and COVID-19 infected patients respectively. The current approach not only differentiate SARS-CoV-2 infection with healthy controls but also predicted a distinct fingerprint for different stages of patient recovery. Employing portable hand-held Raman spectrophotometer as the instrument and saliva as the sample of choice will guarantee a rapid and non-invasive diagnostic strategy to warrant or assure patient comfort and large-scale population screening for SARS-CoV-2 infection and monitoring the recovery process. Label free detection of healthy, COVID-19 and recovered saliva using SERS guided Raman fingerprinting. [Display omitted] •Label-free SERS technique has been developed for the discrimination of healthy and COVID-19 infected subjects using saliva•A proof-of-concept model study with saliva spiked different types of corona virus spike proteins was performed•The trained support vector machine classifier achieved a prediction accuracy of 95% for COVID-19 diagnosis•Illustrated a database for different stages of patient recovery with varying Raman signatures.•Higher prediction accuracy could be due to this small sample size employed