2.8. Statistical Analysis
The bioactivity assay data were analyzed by GraphPad Prism 7.0 (GraphPad Prism, San Diego, CA, USA). We fitted a four-parameter equation (Y = (a − d)/[1 +(x/c)b] + d) curve to calculate the IC50 of mAbs. The relative potency of the samples was calculated by the IC50 ratios with reference to the sample.
Solo software equipped with the Model Exporter add-on (Solo+Model_Exporter version 8.8; Eigenvector Research, Inc.; Manson, WA, USA) was used to develop Raman spectral models based on principal component analysis (PCA). The spectra used to build the models were collected at three replicate scans. Prior to incorporation into the model, the spectral range was reduced to exclude detector noise at > 1700 cm−1. Autoscale was applied followed by a 15-pt first derivative Savitzky–Golay polynomial smooth (second order). Finally, the models were refined by cross validation, using a Venetian blinds procedure. The PCA model was then exported from SOLO and used to follow the method in the TruScan RM Raman analyzers, along with the Raman spectral acquisition parameters. The acceptance criteria for each method were based on the threshold values for two statistics, reduced Hotelling’s T2 (Tr2) and Q-residuals (Qr), which were normalized by dividing the original values by the corresponding confidence level, respectively. Unless otherwise specified, the confidence level was set to 95%.
3. Results and Discussion