Fig. 1
From: Hierarchical and programmable one-pot synthesis of oligosaccharides
The workflow of Auto-CHO program. a The basic concept of Auto-CHO. b The detailed workflow. Auto-CHO allows users to input a desired glycan structure and the program returns with one-pot glycan synthesis options. To facilitate the ability of Auto-CHO, we have not only expanded the thioglycoside BBLs but also constructed a virtual BBL library by enumerating monosaccharide structures with different protecting group combinations and theoretical RRVs that are estimated by our RRV predictor. RRV predictors are trained by SVM regression models, and features from experimental BBLs with known RRVs. The predictor with the best performance by leave-one-out cross-validation (LOOCV) has been chosen as the final model. Independent test has also been applied to validate the result. Since the use of virtual BBL in glycosylation may not have been validated and it is uncertain if it can be successfully used in the one-pot synthesis process, text mining could be used to identify those virtual BBL candidates that have been reported in literature and thus, have a good chance of participating in the one-pot synthetic reaction. The synthetic methods provided by Auto-CHO have been further validated by four synthetic experiments in this study
