Multivariate calibration of covalent aggregate fraction to the raman spectrum of regular human insulin.

Insulin aggregates were prepared by exposing samples of formulated regular human insulin to agitation at 60 degrees C. Aliquots were drawn from the samples periodically over a time range spanning 192 h, and their aggregate compositions were determined with size exclusion chromatography. The complete data set was composed of 39 separate aliquots. The Raman spectra of three separate 10 microL volumes from each aliquot were measured using the drop-coat deposition Raman (DCDR) method. The spectra were calibrated to aggregate composition by partial least squares regression (PLS), resulting in linear calibration (R(2) = 0.997) with a root mean squared error of calibration (RMSEC) of 1.3% and a root mean squared error of cross validation (RMSECV) of 5.1% in aggregate composition. Though the time required for aggregates to form under stressed conditions showed substantial sample-to-sample variation, the correlation between aggregate composition and Raman spectrum was remarkably consistent, indicating that Raman spectroscopy may be a viable screening method for aggregation of protein drugs.