Coarse-grained models of monoclonal antibodies are often used for $\textit{in silico}$ description of statics and dynamics. In this thesis, we investigate two approaches to the construction of coarse-grained models - the k-means machine learning algorithm and the essential dynamics method. We reproduce coarse-grained models of proteins that are qualitatively similar to established 3-, 6-, and 12-particle coarse-grained models using these methods, demonstrating the validity of their application. In the second part of the thesis, we correlate the normal modes computed using the essential dynamics method with the aggregation rate, a critical quality attribute of proteins and with the sum of amplitudes in CDR regions. The results of the linear correlation indicate a likely correlation between the normal modes and the aggregation mechanisms. In this context, we identify two oscillatory modes where more significant correlations are observed.
|
