N-linked Glycosylation Process Controllability Analysis And Operation Space Design Based On PLS

The share of monoclonal antibody(mAbs)drugs with high degree of homogeneity,high targeting and strong specificity and many other advantages has been continued to grow rapidly in the domestic and international pharmaceutical market.As a ctitical part in the preparation of monoclonal antibodies,N-linked glycosylation has a crucial effect on the production of glycans.The N-linked glycosylation reaction mainly completes the post-translational process of proteins in Golgi which will affect the immunological specificity of monoclonal antibodies.But at present,due to theoretical difficulties,complex reaction mechanism,lack of technical means and other reasons,the on-line control of glycans production is still a huge challenge.This paper is based on the guidance and implementation of glycan production control strategy and the quality design of glycans,mainly completed the following work about the N-linked glycosylation.Firstly,based on the data-driven theory,the multivariate statistical regression technique was used to research the controllability of the glycosylation process of the N-linked glycosylation by partial least squares(PLS).The input and output data were obtained by N-linked glycosylation network model and experimental design method(DOE),and a positive model of PLS was established.The mapping relationship between input variable and output sugar type is constructed by singular value matrix.Through the process gain matrix K acquisition and singular value decomposition,the effects of different input glycosyltransferases and ribose donor concentrations on the polysaccharide products were discussed qualitatively by the singular value.From the simulation results,the degree of the controllability of different glycans is shown in detail.Secondly,the quantitative analysis of the quality of the glycan of N-linked glycosylation process was studied,a method of manipulated variable space was proposed.The method uses the partial least squares(PLS)inverse model of multivariate projection technique to design the operation variable space corresponding to the expected glycan quality,which has achieved good prediction.The method uses the acquired input and output data matrices to establish a PLS reverse model,and for each desired output of glycan,corresponding different input variables were found.At the same time,the system forecast uncertainty was considered,the input variables corresponding to each glycan are expanded within the confidence interval,to ensure the rationality and physical viability.The successful application of this method can provide some guidance for the on-line control and quality improvement of glycans.