Research

Research in Bristol Chemometrics Group

Powerpoint presentations of group research (as zip files)

The Centre for Chemometrics has expanded significantly over the last few years. Since October 2001, there have been 61 members of the group, mainly research students, as listed below, from 16 different countries worldwide [Pakistan (2), Turkey (1) , China (6), France (4), UK (26),
Malaysia (3), Portugal (2), Italy (1), Thailand (3), Spain (1), Poland (2), Iran (3), Slovenia (1), Norway (1), Egypt (1), India (1)].

This makes for a very varied and international group.

We have developed substantial external collaborations with both industry and Universities. Our work is divided into five main themes, namely Forensics, Biological Pattern Recognition , Pharmaceuticals, Materials Analysis and Chromatography.

Forensic Work

A new and very promising avenue is forensic work, primarily in collaboration with Rich Sleeman and Jim Carter of Mass Spec Analytical, a company located in the Bristol area.

The aim is to use a technique called tandem mass spectrometry to look at the amount of illicit drugs, cocaine, cannabis, heroin and ecstasy in banknotes. In the UK almost all banknotes contain small amounts of cocaine. This is because the substance "rubs off" onto banknotes, from highly contaminated samples. Other substances such as heroin are less "sticky" but what is likely to happen is that huge levels of contamination come from notes handled freshly from drug dealers, which then enter general circulation and mix with non contaminated notes.

This means that the mass spectrometric technique can pick up patterns of contamination in banknotes. The fact that a person may have a small number of contaminated banknotes in their pocket does not mean they are a drug dealer or that the notes were recently in contact with drugs, however, the higher the frequency and intensity of contamination the more likely this is. The project aims to look at these patterns, a very substantial database is available for us, consisting of the analysis of several tens of thousands of banknotes, to establish statistical patterns. The results have been used in evidence in several court cases.

Gavin Lloyd, Lauren Baba and Sarah Dixon have contributed to this exciting and growing area.

Biological Pattern Recognition

This has been an area of substantial growth over the past couple of years. A major project supported by the US Army Research Office has taken us into contact with several groups recently including the Konrad Lorenz Institute in Vienna, the University of Indiana, the University of Newcastle, the University of Vienna, Vermicon Ltd in Munich, and the Medical University of Vienna.

The aim has been to look at signals in human secretions mainly sweat, but also saliva and urine in order to determine whether chemicals (as monitored by GCMS) can be related to genetics or human individuality. We also have microbial information as it is hypothesised that many compounds in sweat are the result of bacterial action, genotyping and in collaboration with human biologists have studied a large population in Southern Austria whose family histories are all known. In addition we have extensive survey results, providing us information on a wide range of material such as personal habits and emotional states, and also bioassays. This exciting project allows us to put GCMS, bioinformatics, microbiology and human biology together, and we have set up a major database which will pose us challenging problems for the future. This work also relates to some initial studies on mouse scent marks and urine. There is a big challenge developing methods which include Support Vector Machines, Principal Co-ordinates Analysis, Peak Picking and Aligning algorithms, which are necessary to meet the requirements of these new types of data.

In addition several other satellite biological and medical projects have been undertaken successfully, including performing pattern recognition on capillary electropherograms in rat toxicity studies in co-operation with Kings College London, and looking at pattern recognition in fluorescence spectroscopy for rapid genetic screening in co-operation with Sciona. Work on interpreting the GCMS of badger urine using pattern recognition started this area off.

Several people have worked in this fast developing area, including Sarah Dixon, Yun Xu, Simeone Zomer, Fan Gong, Hejun Duan, Nick Embden, Clare Catterall and Thomas Parfrey.

Pharmaceuticals

Pharmaceutical Reaction Monitoring is an important growth area. The aim is to be able to look at reactions as they progress, on-line. We have a specially strong collaboration with Richard Escott and colleagues in GlaxoSmithKline.

An important aspect of this work is to establish that on-line methods for analysis, such as using a uv/vis or mid infrared spectrometer, dipped in the reaction mixture, together with chemometrics for handling the resultant data, provide similar results to off-line approaches such as HPLC. Especially interesting to us involve methods for handling the resultant spectroscopic information obtained by probes to obtain concentration profiles. A number of approaches, many based on alternating least squares, that combine chemometric (or multivariate) modelling of spectra with kinetics modelling have been investigated. These have been extended to provide flexible kinetics for any reaction stoichiometry, to look at how common errors, e.g. in calibration, influence the performance of these methods, compare PLS with external calibration to kinetic approaches that do not require calibrants and worked on self modelling methods. We have worked on the comparison of predictions between a Spectraprobe (mid Infrared) probe and a Zeiss (uv/vis) probe, and obtained very promising results. The multiprobe area is likely to be one of significant growth over the next couple of years. We are also looking into the use of HPLC for monitoring reactions in real time, combined with chemometrics approaches for process monitoring such as Q-charts and D-charts. Especially challenging is to develop software that can be employed in real-time and on-line to obtain predicted concentration profiles from reactions. Most sophisticated chemometrics methods are performed post-processing, and often in environments such as Matlab that require specialist skills. The Boris software package which tackles these issues has been pioneered and is used in-house in GSK.

In addition we also have an interest the use of coupled mass spectrometry (LCMS and pyrolysis GCMS) for looking at the origins of tablets. This is important in the area of patent protection work. Different manufacturing routes often result in very subtle and small impurities in tablets, which can be picked up using coupled chromatogram and chemometrics.

People who have worked recently in this area include Lifeng Zhu, Simeone Zomer, Tom Thurston, Antonio Carvalho, Jirut Wattoom, Steve Matthews, Miguel Sanchez and Thomas Eriksen.

Coupled Chromatography

This has been an important application area for the group over many years, partly due to access to excellent instrumentation within the School of Chemistry.

LCNMR has been an area which is especially new to chemometrics and poses challenges but has significant potential. This involves recording both NMR and HPLC information simultaneously. Although used for some years by organic chemists wishing to look at impurities, it is not generally employed by mainstream analytical chemists, and chemometricians have not had much chance to examine the data, which contains specific features not present, for example in diode array HPLC. This project has involved both experimental and theoretical work. We have evaluated a vast number (over 40) methods for determining the number of components, including cross-validation, Malinowski type functions, SIMPLISMA, orthogonal project approach and so on.

We have had several studies using GCMS and LCMS and chemometrics. During the next year there will be a major upgrade to mass spectrometric facilities within the chemistry department, including new LCMS facilities. This is likely to be an important growth point over the next two years. We continue our interest in diode array HPLC, and several projects use our in-house equipment. We have applied a wide variety of methods for resolving these multivariate techniques, and compared the information content of DAD-HPLC to LCMS and LCNMR.

Hailin Shen, Christian Airiau, Mohammed Wasim and Hassan Sukri have contributed recently to this area, building on the success of several students over the past few years.

Plastics analysis using thermomechanical methods

A new area in collaboration with Triton Technology involves using thermomechanical methods to characterise plastics (polymers). The idea is that when a plastic is heated, its physical state changes as it progresses from solid to glass to liquid, and these properties are highly characteristic of a specific material.

The idea is to develop an instrument that can take an unknown and by heating it one can automatically determine the type of material. This can also be used for quality control purposes, e.g. to determine whether a plasticís specifications remain constant, to compare suppliers, and to compare grades. Most chemometrics has focussed on spectroscopy and chromatography, but this project allows us to apply pattern recognition methods to a unique new area. The initial results are very promising, and we are pioneering this new an exciting area. It is expected that the methods we develop will become incorporated into new cost effective instrumentation and allow suppliers and users of plastics with no knowledge of chemometrics or of polymer chemistry to automatically identify materials.

Bozena Lukasiak, Rita Faria and Dale Chapman are working on this, in co-operation with John Duncan and colleagues.