Application Notes

candidates according to their overall similarity to the reference molecule. A perfect match yields an overlap index value of 0. This analysis showed that sample #11 was the best match by a large margin (Figure 3A and B), while samples #1 and #5 had the lowest similarity scores. Figure 3: Quantitative similarity profiling of nanoDSF data. (A) Comparison of the unfolding signals and corresponding first derivative traces of the best match (sample #11) and the worst match (sample #01). (B) 1 st derivative overlap indices and resulting ranking of all candidates in respect to the reference molecule. In order to evaluate whether identification of biosimilar candidates correlates with complementary techniques, we compared the nanoDSF thermal profiling data to the data from HPLC methods used at UGA Biopharma to assess critical quality attributes of the Fc-fusion protein in question. Charge variants were profiled using a strong anion exchange column with a salt gradient. N-linked glycan profiling was done by HILIC with fluorescent glycan detection. Total sialic acid was quantified by fluorescent detection of released sialic acid. First, all biosimilar candidates were analyzed in charge variant profiling. Biosimilar candidates which showed dissimilar profiles (e.g. multiple peaks in place of a single peak) were not considered for further evaluation. For the remaining samples, similarity indexes were calculated using a proprietary UGA Biopharma algorithm, based on how close the match between the reference product and biosimilar candidate was (Table 1). Table 1. Samples ranked according to the charge variant similarity index. Rank SampleID 1 09 1 11 2 03 3 15 4 07 5 05 6 06b Next, biosimilar candidates were analyzed for their sialic acid content. Again, similarity to the reference product was used to calculate a similarity index (Table 2). Table 2. Samples ranked according to the sialic acid content similarity index. Rank SampleID 1 11 2 09 3 03 4 06b 5 15 6 07 7 05

• Cover