Hierachical and programmable one-pot method brings oligosaccharide synthesis to a whole new level.
One of the most important classes of biomolecules, oligosaccharides are involved in a variety of biochemical processes, including intercellular recognition, cell differentiation, cancer proliferation, inflammation and immune responses.
In humans, carbohydrates are often attached to glycoproteins and glycolipids, being formed from about 9 to 10 monosaccharides through glycosidic linkage to generate enormous structural diversity. For example, about 15 million tetra-saccharides could be formed from nine common monosaccharides.
However, it is very difficult or even impossible to isolate any oligosaccharide in its pure form from natural sources, and as a result synthesis has been utilised in order to furnish sufficient amounts of pure oligosaccharide samples for biological studies.
As part of an effort to improve classical solution phase synthesis, a time-consuming task that often requires a special strategy for each molecule, scientists have been developing effective methods to produce synthetic glycans, in order to speed up the pace of carbohydrate research.
Among the high-speed methods developed for oligosaccharide synthesis, automated solid phase synthesis, which was first reported by Peter Seeberger in 2001, was performed successfully in a modified ABI peptide synthesizer with features adapted for carbohydrate chemistry, dramatically reducing the time required yet proving impractical for obtaining complex and biologically relevant carbohydrate structures.
Today, this automated concept continues to be studied and improved upon, one notable development being the first enzymatic one-pot method developed in 1982 by Wong and Whitesides for the large-scale synthesis of oligosaccharides, as well as the first programmable automated one-pot synthesis method developed by the Wong group in 1999.
The programmable one-pot method was designed to enable the rapid synthesis of a large number of oligosaccharides, using the software Optimer to search Building Blocks (BBLs) with defined relative reactivity values (RRVs) to be used sequentially in one-pot reaction. Two problems existed, however: First, there were only about 50 BBLs with measured RRVs in the original library; second, this method could only synthesize small oligosaccharides due to the nature of RRV ordering requirement.
Now, the research team led by Dr. Chi-Huey Wong and Dr. Chung-Yi Wu of Academia Sinica’s Genomics Research Center, and Dr. Wen-Lian Hsu of its Institute of Information Science, have successfully used artificial intelligence (AI) and algorithm to design a new software program called Auto-CHO with the ability to select the desired building blocks for one-pot synthesis of oligosaccharides. This new technology will play a major role in facilitating the automation of glycan synthesis.
In undertaking this project, the research team expanded its database to include 154 validated BBLs and 50,000 virtual BBLs with predicted RRVs. The resulting virtual library utilized a machine learning approach to build an RRV predictor trained by validated BBLs with feature engineering and optimized processes. The final predictor has achieved outstanding performances based on cross-validation and independent tests, with many virtual BBLs further confirmed by subsequent experiments.
In addition, the research team has designed a new algorithm that supports hierarchical one-pot synthesis using fragments as BBLs generated by such synthesis. They also developed a new software known as Auto-CHO, free and open source GUI software that can be operated in Windows, macOS, and Ubuntu. This advanced programmable one-pot method provides potential synthetic solutions for numerous complex glycans. Moreover, users can give feedback on the use of virtual BBLs through Auto-CHO, which can help the scientific community fine-tune and improve virtual building blocks while removing all ineffective ones. Thanks to the Auto-CHO program, four biologically important glycans were successfully synthesized, as demonstrated in this research.
The research was published in Nature Communications on 6 December 2018. [APBN]