S. Koya1, M. Brusatori1, S. Yurgelevic1, C. Huang1, L. N. Diebel2, G. W. Auner2 1Wayne State University,Smart Sensors And Integrated Microsystems, Michael And Marian Ilitch Department Of Surgery,Detroit, MI, USA 2Wayne State University,Michael And Marian Ilitch Department Of Surgery,Detroit, MI, USA
Introduction:
Clinical practice guidelines define Clostridium difficile infections (CDI) as diarrhea (≥3 unformed stools in 24 hrs.) with either a positive C. difficile stool test or detection of pseudomembranous colitis. Pathogenicity of CDI is due to toxins, toxin A (TcdA) and toxin B (TcdB). While presence of toxin is necessary for disease, detection of toxins using immunoassays is complex and lacks sensitivity. For this reason, positive C. difficile in stool by toxigenic culture (TC) and nucleic acid amplification testing (NAAT) are used. These tests are confounded by the presence of asymptomatic colonization of toxigenic C. difficile and leads to overdiagnosis of CDI. Raman spectroscopy (RS) is a novel technology that is used to detect bacteria and their toxins. RS doesn’t require antibodies for detection of toxins. We hypothesis that RS may be a sensitive method to detect C. difficile toxins in stool and solve overdiagnosis of CDI.
Methods:
CDI negative stool samples were spiked with varying concentrations of TcdA and TcdB. RS was performed on smear of stool on a mirror polished stainless-steel slide. RS of feces is difficult due to confounding background material and autofluorescence. The samples were photo-bleached to reduce autofluorescence. Raman spectra were obtained, background corrected, vector normalized and analyzed by machine learning methods including Support Vector Machine (SVM), Random Forest (RF) and Partial Least Square Linear Discriminant Analysis (PLS-LDA). The best model was chosen, and its accuracy was measured by train-test, cross-validation and bootstrap methods.
Results:
At 1ng/ml concentration, TcdA and TcdB spiked stool were distinguished from control stool by all models with various accuracies. SVM with linear kernel performed best for TcdA with an accuracy of 85% and and PLS-LDA performed best for TcdB with 75% accuracy.
Conclusion:
RS of feces is difficult due to autofluorescence. Despite the difficulty, RS successfully detected TcdA and TcdB in stool samples. The autofluorescence could be further decreased by using diluted samples of stool and the accuracy of the separation could be increased by deep learning algorithms. Thus, RS has the potential to rapidly detect C. difficile toxins in stool at clinically relevant concentrations, be a point of care diagnostic tool and solve overdiagnosis of CDI.
