A multiple linear regression model for structure of η-linked oligosaccharides

It is well-known that carbohydrate sugar chains, or glycans, play various well in cellular processes, including cancer, but the elucidation of glycans is difficult because of their complex structure. Both computational methods and mathematical models are necessary to integrate and analyze the information of glycomics data so as to efficiently detect glycan structures. In this paper, we propose a new model to predict the structure of N-glycans, which are the most common type of glycans. Our proposed prediction method is based on a Multiple Linear Regression (MLR) model. The coefficients of our proposed model are solved by using experimental data. We obtain our data from High Performance Liquid Chromatography (HPLC) experiments. Three sources of our data are adopted and they are divided into two parts: elution value on an Amide column and elution value on an OctaDecylSilane (ODS) column. After pre-processing the data, we then construct our proposed MLR model. The obtained correlation coefficients are 0.9680 for the Amide data and 0.9263 for the ODS data. We have also tested the correctness of the model statistically. The model test and correlation coefficients demonstrate both the accuracy and efficiency of our proposed model.