The home of high impact research in analytical, bioanalytical and detection science.
Subscribers
Non-subscribers
- Purchase article PDF [£27 + taxes]
- Purchase article PDF member offer [£5 + taxes]
Free access
Paper
Analyst, 2009, 134, 1462 - 1471, DOI: 10.1039/b902154j
Viral infection of cells in culture detected using infrared microscopy
Gary Hastings, Peter Krug, Ruili Wang, Jing Guo, Hari P. Lamichhane, Tian Tang, Yu-sheng Hsu, John Ward, David Katz and Julia Hilliard
FTIR microscopy has been used to collect spectra for uninfected (mock) Vero cells, and cells that have been infected with herpes simplex virus type 1 (HSV-1) and human adenovirus type 5 (Ad-5). Cells were infected at a multiplicity of infection of 10, and studied at 24 hours post exposure. The spectra for infected samples display many differences compared to the spectra for uninfected samples. To estimate how well the spectra for uninfected and infected samples could be discriminated, we used logistic and partial least squares regression methods. We show that the spectra for HSV-1 and mock infected samples are well differentiated and, for a sensitivity of 95%, we calculate a specificity of 0.999 using partial least squares regression. Spectra for Ad-5 and mock infected samples are also well differentiated. We find that applying our regression models constructed with one data set to a new validating data set still gives very high levels of specificity for a given sensitivity. Spectra for Ad-5 and HSV-1 infected samples are also differentiable. Applying our constructed regression models to new validating data, however, leads to a decrease in the discrimination capability in this instance. If one is simply interested in differentiating spectra associated with uninfected and infected cells, without distinguishing the type of infection, then we show that logistic regression models can break down whereas partial least squares regression models perform well.
RSS feed