Projets du Fond National Suisse (FNS)

How to make rational decisions: Bayesian networks and decision theory in forensic science applications

Responsable : Franco Taroni

The present research program thus aims at conducting fundamental research on how frameworks, so-called Bayesian inference and decision networks, that are capable of (a) representing knowledge about a problem of interest (e.g. its current state, its believed underlying working mechanisms or how it might be influenced by a scientist’s or a client’s actions), and (b) reasoning logically with that knowledge may be employed as a more general means to assist scientists in deciding what courses of action they ought to adopt in the presence of uncertainty, with a particular emphasis on optimising the expected desirability of the outcomes that are associated with the available courses of action.

 

Progress in handwriting analysis : towards shape quantification of characters and probabilistic assessment of their inferential value

Responsable : Franco Taroni

The discipline of forensic handwriting examination attracts considerable attention due to its uncertain status under new admissibility standards, notably in the United States of America. It has been highlighted that forensic scientists need to assess continually their practice, scrutinize underlying rationales and ways to evaluate and present findings. Nowadays, systematically compared to the new gold standard of forensic identification, namely DNA profiling, handwriting examination suffers from a lack of empirical research. Courts seem less and less at ease with expert opinions based on purely subjective belief.
With regards to this context, the primary aim of this research proposal is to contribute to the body of (i) operational and useable methods and techniques for the analysis of handwriting, notably quantification of shape characters, and (ii) rigorous probabilistic inferential procedures for coherent assessment of handwriting evidence. Research in this area is fundamental and has the potential of securing the position of the discipline of forensic handwriting examination as one whose results can be trusted by the public and the judiciary.

Recognition of Patterns in Forensic Case Data : The Use of Chemical/Physical Signatures of Illicit Drug Seizures in an Intelligence Perspective

Responsable : Olivier Ribaux

Projet de recherche en collaboration avec le prof Mikhail Kanevski de l'Institut de Géomatique et de l'Analyse du Risque (IGAR) de l'UNIL et Pierre Esseiva de l'IPS. Ce fonds finance le travail d'Anne-Laure Terrettaz.

Le but de cette recherche pluridisciplinaire est de déterminer et caractériser le potentiel réel des données forensiques issues de cas, dans une optique de renseignement. Une part importante de ce projet se concentre sur l'exploitation d'un jeu de données constitué de la composition chimiques et physiques des échantillons de stupéfiants ainsi que de la date et du lieu de saisies effectuées dans une région particulière de la Suisse.

Le pré-traitement des données sera effectué dans le but de créer des bases de données spatiales et temporelles. Les données seront classifiées selon des critères temporels et leur distribution spatiale. Des tests statistiques (corrélations multivariées, regressions, modèle de distribution) serviront à explorer la structure des données disponibles. Sur cette base et grâce aux recherches précédentes, la structure de la base de données sera restructurée en sélectionnant les variables pertinentes. Cela inclus la pureté ainsi que la proportion des alcaloïdes majeurs de la cocaïne et de l'héroïne. Le but de cette exploration est de mettre en évidence des patterns capables de produire des hypothèses pertinentes concernant l'évolution du trafic de stupéfiants.

Les méthodes traditionnelles de pattern recognition et d'analyses multivariées (représentation graphique des variables, hierarchical clustering et analyse en composante principale) seront appliquées sur cette base de données restructurée afin de détecter des patterns géographiques et temporels.

Dans la phase suivante, des méthodes avancées de pattern recognition basées sur des techniques intelligentes d'analyses de données, dont les réseaux de neurones artificiels et les Support Vector Machines seront développés et adaptés.

Un intérêt particulier sera porté sur l'analyse des séries temporelles afin de déterminer la structure temporelle et la prédictibilité. L'analyse comprend des outils de modélisation stochastique et chaotique non linéaire utilisés pour caractériser les comportements complexes du phénomène dans le temps.

Les résultats obtenus seront systématiquement interprétés et confrontés aux autres aspects de l'investigation criminelle. L'intégration de l'information dans le contexte de l'enquête a pour but d'évaluer les nouveaux outils de production de renseignements décrits. Les résultats de cette approche devraient renforcer la connaissance de l'évolution spatiale et temporelle des réseaux, du type et de la nature des liens entre les saisies ainsi que de l'origine des stupéfiants.

Les résultats de cette étude mèneront à l'élaboration d'une méthodologie complète pour l'utilisation de données forensiques issues de cas par les Systèmes d'Information Géographique et les aspects temporels, afin de représenter et de comprendre le trafic de stupéfiants. Ce approche de orientée renseignement permettra aux autorités de lutte contre le trafic de stupéfiant de concentrer leurs efforts sur des périodes de temps et des endroits précis.

Interpretation of forensic evidence - probabilistic foundation of handwriting identification

Responsable : Franco Taroni

Projet de recherche en collaboration avec le Dr M. Schmittbühl de l'Université de Strasbourg (France) et la Dr.sse S. Bozza de l'Université de Venise (Italie). Cette bourse finance le travail de doctorat de M. R. Marquis.

One forensic discipline, handwriting identification, has attracted considerable attention due to its uncertain status under new admissibility standards on how scientific evidence is evaluated. In fact, handwriting experts have been challenged in a numerous of judicial cases. The fact that handwriting expertise has been questioned and continues to be, only goes to highlight scientist's need to continually assess his/her work, the rational behind the work and how to present the findings from the technical work in the written report and in the testimony in front of a Court of Justice. The scientist need to have the scientific basis for the interpretation of his findings and he needs to have some statistical theories to justify and rationalise his/her conclusions.

Objective statistical data are rare, not to say inexistant in the field of handwriting, even if data will provide a more scientific and statistical backing for expert's conclusions. Difficulties still arise in the collection of data, the choice of the relevant technical parameters to analyse and the methodology to extract such an information. The scope of this proposal for a Ph.D research is to reduce these difficulties obtaining relevant statistical data which allow the forensic scientist to contribute to the answer the following questions which represent the premises of handwriting:

Each person's handwriting is different.
No one person will write exactly the same way twice.
Is it possible, given handwriting's samples of several individuals, to establish whether or not an individual was the author of a questioned handwriting by comparing in details its constituents features with adequate comparable specimens known to have been written by that individual?

Such a debate represents a point high on the priority list of considerations of the European Network of Forensic Science Institutes (ENFSI) which underlines that the continued consideration of this 'statistical' point will help to enhance development and credibility within the handwriting field. Furthermore, a beneficial multi-disciplinary co-operation at an international level (Swiss, France and Italy) is the co-operation between forensic handwriting examiners and statisticians. These fields have common features and co-operation between them may produce results that none of the disciplines could have produced separately.

Analysis of recurrent complications in the evaluation of Low Copy Number (LCN) DNA profiling results in forensic science : a graphical probability approach

Responsable : Franco Taroni

 

Projet de recherche en collaboration avec le Prof. S. Bozza de l'Université de Venise (Italie) et le Dr Alex Biedermann de l'Office Fédéral de la Police (Suisse).

 

Increasingly often, only very small amounts of DNA are extracted from swabs collected in forensic contexts. Although well established DNA profiling methods may be applied to such kind of samples (known in the context as Low Copy Number (LCN) samples), the resulting profiles may show substantial morphological changes. This requires particular careful interpretation. Some formative probabilistic evaluative procedures are currently available, but their conceptual underpinnings rely on a series of assumptions and their practical implementation requires considerable technical expertise. This project intends to contribute to this area of research through the study of a graphical probabilistic representation method known as Bayesian networks, that is a reference method for handling uncertainty in expert knowledge. The proposed research aims to rely on Bayesian networks for developing prototype expert system components for evaluating forensic LCN DNA profiling results. The resulting models will be specified numerically using relevant data collections and analyzed in close collaboration with experts working in the field (forensic geneticists).

 

Development of a new analytical method for the profiling of ecstasy

Responsable: Pierre Esseiva

In view of the ever-growing drug problem around the world, there is an increasing need to provide strategic intelligence on sources and trafficking routes of drugs. Chemical characterisation and impurity profiling are valuable scientific tools used to support traditional police investigations (observations, interrogations, taping).

Impurity profiling is the analysis of the various impurities in clandestinely manufactures drugs, by type and quantity, rising to a chemical 'signature'. For drugs manufactured from natural sources such as heroin and cocaine, impurities relate to the origin of the drugs. Impurities in synthetic drugs such as ecstasy and amphetamine arise from the manufacturing process. Signatures can help to establish chemical links between samples and material from different seizures. They can help to classify into groups related samples. Consequently, and most useful for law enforcement authorities, specific links for instance between suppliers and users can be established, drug distribution patterns/networks can be built up and the source, including the geographic origin of drug samples may be identified. Chemical profiling requires specific, highly sensitive and reproducible methods. Nowadays the current methods used for this purpose are not satisfactory, and most important up to now does not exist a profiling method unique to different government forensic laboratories.

The general aim of the present research is to develop an analytic method for ecstasy profiling with a forensic approach, that explores the analytic procedure in order to obtain a data set, which can be classified and interpreted to help the drug intelligence work. Firstly the efforts will be concentrate on the development of an innovative and simple analytical procedure to profile seized ecstasy tablets. This procedure will be based on the Solid-Phase Microextraction (SPME) technique for the extraction of volatile impurities from samples in combination with gas chromatography/mass spectrometry analysis (GC/MS). SPME is a rapid and easily automated method that can eliminate totally the use of organic solvents, which are normally used in other extraction techniques, like liquid-liquid extraction (LLE). The collected data, will be compared and integrate with the result obtained with other techniques, in order to validate the method, and to achieve a sufficient number of information. In a second stage the data set will be interpreted and evaluated in order to analyse the chemical "history" of the different seizures. The data interpretation and evaluation based on chemometric methods will be considered too. The optimised procedure will be applied to illicit ecstasy samples seized in the states of Geneva, Jura, Neuchâtel, Ticino, Vaud and Zürich. To date a national database ("XTC Drug Intelligence" database) is available which resumes optical, physical and chemical (quantity of active substance) observations on ecstasy samples seized. The results obtained from this study will be used to complement the described database and help the Swiss police authorities in their drug intelligence work.

 

Source correlations of gasoline traces using Gas Chromatography - Isotope Ratio Mass Spectrometry

Responsable : Olivier Delémont

L'essence est le produit inflammable qui est le plus fréquemment détecté dans les prélèvements réalisés sur les lieux d'incendies volontaires; elle est couramment utilisée, d'une part par les pyromanes et les incendiaires afin d'allumer un incendie, d'autre part pour la confection de bouteilles incendiaires employées dans des actes de vandalisme. Les investigateurs d'incendies, les enquêteurs et les instances judiciaires effectuent de plus en plus de demandes quant à la possibilité d'établir un lien entre les traces d'un produit inflammable détectées dans un prélèvement et une source potentielle. La chromatographie en phase gazeuse couplée à la spectrométrie de masse (GC-MS), qui est la technique généralement utilisée pour effectuer ces analyses, permet uniquement de réaliser ce type de liens avec des échantillons liquides, et non avec des traces extraites de résidus calcinés.

 

Le but principal de cette recherche est d'individualiser la source d'essence par la chromatographie en phase gazeuse couplée à la spectrométrie de masse à rapport isotopique (GC-IRMS) appliquée à l'analyse des échantillons de débris d'incendies. Cette technique qui mesure les rapports d'isotopes stables de chaque composé constituant un mélange complexe a déjà permis d'inférer la source de produits dans d'autres domaines, en particulier l'environnement et la géochimie organique.

 

En sciences forensiques, les données qui peuvent être obtenues avec cette technique devront être validées afin de les employer à des fins de lutte contre les actes criminels et le vandalisme. Premièrement, la technique sera optimisée pour être appliquée aux analyses d'échantillons d'essence. Ensuite, les données collectées seront comparées et intégrées aux résultats obtenus avec une autre technique (GC-MS), afin de valider la méthode. Dans une troisième étape, les données seront interprétées et évaluées afin de déterminer le niveau de discrimination qu'il est possible d'atteindre pour des échantillons d'origine différente mais de même composition. L'interprétation et l'évaluation des données se feront en recourant à des méthodes chimiométriques. Finalement, la procédure optimisée sera appliquée à l'analyse d'échantillons de débris d'incendies et de bouteilles incendiaires dans le but d'inférer la source de l'essence qu'ils contiennent.

 

D'une part, les résultats obtenus à partir de cette étude fourniront des données fondamentales sur la technique de la GC-IRMS tels que le fractionnement isotopique d'échantillons liquides altérés (en particulier par évaporation et en les brûlant à différents taux) et l'influence des techniques d'échantillonnage sur la composition isotopique. D'autre part, ils pourront être utilisés pour établir un lien entre des échantillons de débris d'incendies et une source potentielle; ceci représentera un nouvel outil pour l'investigation des incendies volontaires, des déprédations et par extension de certains actes terroristes.

 

Analysis of recurrent complications in the evaluation of Low Copy Number (LCN) DNA profiling results in forensic science: a graphical probability approach

 

Responsable : Franco Taroni

Projet de recherche en collaboration avec le Prof. S. Bozza de l'Université de Venise (Italie) et le Dr Alex Biedermann de l'Office Fédéral de la Police (Suisse).

 

Increasingly often, only very small amounts of DNA are extracted from swabs collected in forensic contexts. Although well established DNA profiling methods may be applied to such kind of samples (known in the context as Low Copy Number (LCN) samples), the resulting profiles may show substantial morphological changes. This requires particular careful interpretation. Some formative probabilistic evaluative procedures are currently available, but their conceptual underpinnings rely on a series of assumptions and their practical implementation requires considerable technical expertise. This project intends to contribute to this area of research through the study of a graphical probabilistic representation method known as Bayesian networks, that is a reference method for handling uncertainty in expert knowledge. The proposed research aims to rely on Bayesian networks for developing prototype expert system components for evaluating forensic LCN DNA profiling results. The resulting models will be specified numerically using relevant data collections and analyzed in close collaboration with experts working in the field (forensic geneticists).

Bayesian networks and the analysis of combinations of items of evidence

Responsable : Franco Taroni

Projet de recherche en collaboration avec le Prof. C. Aitken de l'Université de Edinburgh (Ecosse), le Prof. S. Bozza de l'Université de Venise (Italie) et M. Alex Biedermann de l'Office Fédéral de la Police (Suisse).

The current state-of-the art in forensic science, notably in the evaluation of scientific evidence, does not allow scientists to face adequately problems caused by complexity due to, for instance, the combination of distinct items of evidence, even if the latter aspect represents a recurrent task of routine work in practice. Methods of formal reasoning have been proposed to assist forensic scientists to understand all of the dependencies which may exist between different aspects of evidence and to deal with the formal analysis of decision making. Among them are graphical methods, e.g. Bayesian networks, which were found to provide valuable aid for representing relationships between characteristics of interest in situations of uncertainty, unpredictability or imprecision. Recently, several researchers have begun to converge on a common set of issues surrounding the representation of forensic inference problems which are structured with Bayesian networks, a widely applicable formalism for a compact representation of uncertain relationships among parameters in a domain (in this case, forensic science). The Bayesian network formalism incorporates Bayesian probability theory as an integral part which guarantees that uncertainty in knowledge and inferences based upon evidence are handled in a logically rigorous manner. Bayesian networks can assist the task of developing and specifying relevant equations used to evaluate and interpret scientific evidence as arithmetic can be made invisible to the user and almost completely be automated. Most importantly, the intellectually difficult task of organizing and arraying complex sets of evidence to exhibit their dependencies and independencies can be made explicit and intuitive. Bayesian networks are thus considered as a method for discovering valid, novel and potentially useful patterns in data whereas uncertainty is handled in a mathematically rigorous, but simple and logical, way. In addition, their efficient graphical representation allows for an evaluation of all possible stories related to a scenario. In summary, the use of graphical models - Bayesian networks in particular - has some key advantages that could be described as follows:

1) the ability to structure inferential processes, permitting the consideration of problems in a logical and sequential fashion;

2) the requirement to evaluate all potential scenarios;

3) the possibility to calculate the effect of knowing the truth of one proposition or piece of evidence on the plausibility of others;

4) the communication of the processes involved in the inferential problems to others in a succinct manner, illustrating the assumptions made;

5) the ability to focus the discussion on probability and underlying assumptions.

 

Examples of the general use of Bayesian networks in forensic science have already been presented in several papers, but literature does not focus on the inferential problems caused by the combination of items of evidence and on the reliability of data. As always when modelling real world domains (e.g., forensic science), the results obtained from a model rely on the adequacy of the modelling assumptions as well as on the reliability of the data used. Therefore, forensic scientists might tend to maintain a critical attitude towards recommendations provided by expert systems, in particular with regards to input data.

The proposal would like to approach such issues explicitly and focus on the use of several means of explanatory tools (such as sensitivity and qualitative analyses as well as conflict analysis) proposed for analysing decision problems involving the combination of items of evidence.

The aim of the grant proposal is to study the fundamental principles of the logic of combining items of scientific evidence in forensic science.

 

Statistical foundations and ethical considerations involved by partial DNA profiles and familial searching using the Swiss National DNA Database

Responsable : Franco Taroni

Projet de recherche en collaboration avec le Prof. O. Ribaux de l'Université de Lausanne (Suisse) et le Dr V. Castella de l'Université de Lausanne (Suisse).

In recent years, there has been a shift towards intelligence-led strategies in order to increase safety within our democratic society. The detection of threats through the automatic recognition of people within a crowd, the storage and search of patterns within huge quantities of data pertaining to various forms of transactions (account transactions, mobile phones conversations, etc.) or computerised transcription of conversations are declared objectives. Those developments question how modern technologies can be used to protect institutions, people and goods, together with safeguarding human fundamental rights and civil liberties. It is argued that a deeper understanding of intelligence processes through an intensive modelling program is needed in order to better control the implementation of surveillance and investigative technologies. This fundamental research pertains to an original use of DNA databases in this perspective: partial DNA and familial searching.

The launching, in 1995, of the United Kingdom National DNA database, followed closely by the United States of America and a few years later (2000) by Switzerland, is considered a major breakthrough for the criminal justice system. It is viewed as a milestone, because DNA's role has become twofold: it establishes/excludes links between suspects and crime, and generates names of suspects. In traditional criminal investigation, uncertainties mostly pervade through the use of implicit common sense knowledge. There is a need to prioritise information sources from which it is possible to use solid statistical models that make the treatment of uncertainties explicit. In this perspective, there are four areas of fundamental research regarding the use of DNA databases: the interpretation of DNA using (1) partial profiles and (2) familial searching, (3) use of the DNA database for intelligence and (4) ethical issues related to DNA databases.

The aim of the present project is to study both scientific and ethical issues surrounding the use of partial profiles and familial searching in conjunction with DNA national databases for intelligence purposes. The two main objectives of our study are the processes involved by the DNA intelligence database and the establishment of statistical bases for familial searching and partial profiles. New data will be published in order to show the potential of these techniques. Taking both scientific and social data into account will allow a scientific view on these innovative ways to use DNA databases for intelligence. Such a fundamental research will greatly benefit the criminal justice system trough well formalised methods that make uncertainties explicit.

 

Probabilistic graphical models and forensic science

Responsable : Franco Taroni

Projet de recherche en collaboration avec le Prof. C. Champod de l'Université de Lausanne (ESC) et le Prof. J. Smith de l'Université de Warwick (Angleterre). Cette bourse finance le travail de doctorat de Mme M. Bernard.

Intensive research during the last three decades by researchers from several academic fields shows that many complex structure and decision problems can actually be formulated and solved by machines using expert systems. Experts systems can be broadly defined as a computer system that simulates human experts in a given area of specialisation.

Forensic science and judicial literature has pointed out the utility of methods that deal with formal analysis of decision making. Notably, it has been underlined that complex frameworks of circumstances - situations involving many variables - require a logical assistance. Methods of formal reasoning have been proposed to assist the forensic scientist to understand all of the dependencies which may exist among different aspects of the evidence. Bayesian Networks (BNs) provide a valuable aid for representing relationships among characteristics in situations of uncertainty. They assist the user not only in describing a complex problem and communicating information about its structure but also in calculating the effect of knowing the truth of one proposition or one piece of evidence on the plausibility of others.

The proposed research programme will aim at addressing problems in representation, inference, knowledge engineering and explanation within the decision-theoretic framework in forensic science.
This proposal addresses the problem of rational assessment of scientific evidence associated with decision-making processes that are characterised by a high degree of uncertainty with respect to the input data, the process models and connections between such models. We would like to focus on forensic science, with particular emphasis on the problem of combination of evidence (the recovery of two or more stains/marks on a scene of crime) and developed the idea of harmonious and dissonant evidence through the use of the (Bayesian) likelihood ratio to make clear the potential enhancement of a second (or additional) evidence.

From a practical point of view, we propose to deal with these complex issues using probabilistic expert systems, BNs. This because, in practice, we have often encountered the following problems: (a) a lack of logical decision framework, (b) a lack of data analysis and evaluation framework (in fact, even the collected data are not properly analysed and/or visualised, since it is generally not known how, and particularly why to use it), (c) a lack of modelling framework (in fact few models exist to simulate the effect of combining evidence, although it is generally accepted by forensic practitioners that additional evidence will clearly strength the force of the link between (for example) a suspect and a crime scene), and (d) a lack of methodology for assessing the value of additional information.

Nouvelle méthodes d'analyse des peintures automobiles par pyrolyse GC/MS. Utilisation de méthodes chémométriques et de reconnaissance de forme pour le traitement des résultats

Responsable : Geneviève Massonnet

Les avancements technologiques qui ont eu lieu dans le domaine des peintures automobiles, nécessitent, dans le cadre des examens forensiques, le développement de méthodes analytiques plus discriminantes pour les différentes mesures et analyses effectuées sur les échantillons de peinture automobile. En effet, les mises en peinture modernes et futures ne pourront plus être différenciées efficacement sur la base des informations fournies par la spectrométrie infrarouge à transformée de Fourier (FTIR). Les différences chimiques indécelables par cette technique doivent, grâce à l'utilisation de techniques plus performantes, pouvoir être mises en évidence lors de l'analyse de systèmes de peintures très similaires mais différents.

L'objectif principal de cette recherche est le développement d'une procédure analytique standardisée pour l'analyse des échantillons de peinture automobile par l'application de techniques analytiques couplées, à savoir, la pyrolyse / chromatographie en phase gazeuse couplée à un spectromètre de masse (Py-GC-MS). La mise en place d'une telle procédure est nécessaire afin d'accroître les possibilités de discriminer les échantillons de peinture automobile rencontrés en sciences forensiques.

Progress in handwriting analysis: towards shape quantification of characters and probabilistic assessment of their evidential value

Responsable : Franco Taroni

The discipline of forensic handwriting examination attracts considerable attention due to its uncertain status under new admissibility standards, notably in the United States of America. It has been highlighted that forensic scientists need to assess continually their practice, scrutinize underlying rationales and ways to evaluate and present findings. Nowadays, systematically compared to the new gold standard of forensic identification, namely DNA profiling, handwriting examination suffers from a lack of empirical research. Courts seem less and less at ease with expert opinions based on purely subjective belief.

With regards to this context, the primary aim of this research proposal is to contribute to the body of (i) operational and useable methods and techniques for the analysis of handwriting, notably quantification of shape characters, and (ii) rigorous probabilistic inferential procedures for coherent assessment of handwriting evidence. Research in this area is fundamental and has the potential of securing the position of the discipline of forensic handwriting examination as one whose results can be trusted by the public and the judiciary.

 

Development of latent fingerprint detection techniques based on molecular recognition: Use of Cyclodextrin/Ligand and Antibody/Antigen systems

Responsable : Christophe Champod

During the investigation of crime, various elements of forensic nature may help to reconstruct the events and/or potentially associate designated person(s) with the investigated acts. Among these elements, latent fingerprints (or fingermarks) constitute one of the most useful and researched evidence, due to their uniqueness, inalterability and the today's availability of automatic fingerprint recognition systems. A fingermark of sufficient quality possesses all the qualities for personal identification in forensic context. When an object or a surface is touched with the inner surfaces of unprotected hands (or feet), a small amount of secretions (mainly composed of lipids, amino acids, organic molecules, ...) is left on the surface of this object and leave, as a rubber stamp, a latent residual image made of ridges and furrows.

There are three different types of fingerprints that can be discovered: visible, plastic, and latent. The last category is of particular interest here as they are not visible at first examination and require or their visualisation the deployment of dedicated detection techniques. Currently, several different methods exist and are described in Margot & Lennard (1994). The principle in common to all these techniques is the differentiation of search for an optimal detection of the secretion drawings as opposed to the background (= substrate). The existing methods take advantage from optical behaviours, physical or chemical reactions. The visualisation of the end product can take advantage of fluorescence techniques, e.g., ninhydrin analogs (Alaoui & Menzel, 1996), DFO, rare-earth complexes (Allred & Menzel, 1997), or colour-based contrast, e.g., physical developer, cyanoacrylate fuming (Burns et al., 1998), metallic multi-deposition (Schnetz & Margot, 2001). The suitable techniques to apply in a given case is selected as a function of the surface to be treated (porous, semi-porous or non-porous), the environmental constrains, the molecules or elements of interest, e.g., amino acids or lipids (Wilkinson, 1999) and the ability to put the reagents in sequence (Margot & Lennard, 1994).. However, today, there is not a single procedure that would reveal fingermarks in every situation. The large number of different surfaces to be studied (paper, metal, cardboard, wood, clothing, ...), as well as the age and the conditions of conservation of the prints, imply a specific approach dedicated to each situation.

Most of the existing chemical and physical techniques are not free from difficulties and have their own limitations. It is still difficult or even impossible to reveal the latent fingermarks in some cases (such as the human skin). Among the different drawbacks, the background fluorescence constitutes one of the main difficulties encountered with conventional methods, as it decreases the contrast between the print and the support. Furthermore, some techniques, which are successfully developed for laboratory use, are not operational in the field: the toxicity of some reactants or the destruction of the substrates are strong limiting factors here. The aim of our project is to develop a new efficient method for the detection of prints on porous and non-porous surfaces that can take advantage of detection in a luminescence mode (favoured for its increased sensitivity) in a non-toxic environment. We are particularly keen at developing an efficient, portative, non-toxic, and non-destructive detection technique that can challenge the existing processes. Our strategy is based on the use of fluorescent probes combined with molecular recognition as main detecting method, due to its high sensitivity and selectivity. The challenge being to be able to tag the fingerprint with these molecules, e.g., with reactants that specifically interacts with the secretion components or with a metallic layer previously deposited on the print that allows covalent binding of organosulfur molecules. At this purpose, we will firstly explore a way that has not yet been considered, i.e., the use of cyclodextrins as chelating agents, in regard to their successful application in many different domains (food, textiles, chemistry, ...). Secondly, we will focus our work on the improvement of existing biochemical methods, i.e., by introducing a spacer molecule between an antibody/antigen complex and the support of interest in order to higher the resulting fluorescence. These two approaches are described hereafter.

Recherche:
 dans ce site:
   
   
 Go
 
rss/atom
Batochime - 1015 Lausanne  - Suisse  -  Tél. +41 21 69246 00  -  Fax +41 21 69246 05
Swiss University