Carbohydrates are one of the four main building blocks of life, and are categorized as monosaccharides (sugars), oligosaccharides and polysaccharides. Each sugar can exist in two alternative anomers (in which a hydroxy group at C-1 takes different orientations) and each pair of sugars can form different epimers (isomers around the stereocentres connecting the sugars). This leads to a vast combinatorial complexity, intractable to mass spectrometry and requiring large amounts of sample for NMR characterization. Combining measurements of collision cross section with mass spectrometry (IM–MS) helps, but many isomers are still difficult to separate. Here, we show that recognition tunnelling (RT) can classify many anomers and epimers via the current fluctuations they produce when captured in a tunnel junction functionalized with recognition molecules. Most importantly, RT is a nanoscale technique utilizing sub-picomole quantities of analyte. If integrated into a nanopore, RT would provide a unique approach to sequencing linear polysaccharides.

Richard Feynman said “There’s plenty of room at the bottom”. This inspired the techniques to improve the single molecule measurements. Since the first single molecule study was in 1961, it has been developed in various field and evolved into powerful tools to understand chemical and biological property of molecules. This thesis demonstrates electronic single molecule measurement with Scanning Tunneling Microscopy (STM) and two of applications of STM; Break Junction (BJ) and Recognition Tunneling (RT). First, the two series of carotenoid molecules with four different substituents were investigated to show how substituents relate to the conductance and molecular structure. The measured conductance by STM-BJ shows that Nitrogen induces molecular twist of phenyl distal substituents and conductivity increasing rather than Carbon. Also, the conductivity is adjustable by replacing the sort of residues at phenyl substituents. Next, amino acids and peptides were identified through STM-RT. The distribution of the intuitive features (such as amplitude or width) are mostly overlapped and gives only a little bit higher separation probability than random separation. By generating some features in frequency and cepstrum domain, the classification accuracy was dramatically increased. Because of large data size and many features, supporting vector machine (machine learning algorithm for big data) was used to identify the analyte from a data pool of all analytes RT data. The STM-RT opens a possibility of molecular sequencing in single molecule level. Similarly, carbohydrates were studied by STM-RT. Carbohydrates are difficult to read the sequence, due to their huge number of possible isomeric configurations. This study shows that STM-RT can identify not only isomers of mono-saccharides and disaccharides, but also various mono-saccharides from a data pool of eleven analytes. In addition, the binding affinity between recognition molecule and analyte was investigated by comparing with surface plasmon resonance. In present, the RT technique is applying to chip type sequencing device onto solid-state nanopore to read out glycosaminoglycans which is ubiquitous to all mammalian cells and controls biological activities.