Research

The topics that this research covers all involves small molecules and nanomaterials, and is currently directed at metabolite and small compound pathways, nanotoxicology, and life sciences in general. Research is applied to the fields of drug discovery, QSAR/QSPR, among others. Past topics include crystallography and polymorph prediction.

The research I am personally interested in is about reducing the error introduced in (chemical or molecular) data analysis by the analysis itself, a field coined Molecular Chemometrics. This involves study of the errors introduced by improper handling of chemical knowledge, improper representation of the problem, and the statistical analysis method used. The solution my group is using to reduce that error involves explicit markup of knowledge using semantic technologies, development of new representation methods for molecular information, and the use of statistics to find, visualize, and validate new patterns.

Practically, however, this also means that to prove my point, I need others to adopt such methods too. This has resulted in that I am involved very much in Applied Cheminformatics too. This research field is about getting sound cheminformatics used in practice, and raising awareness about the problems in Molecular Chemometrics among people in other research fields, such as bioinformatics, metabolomics, QSAR, etc.

Each part of my research is exemplified by one or two key or recent papers.

Semantic Technologies

Technologies studied here include markup languages, like Chemical Markup Language, and the family of semantic approaches around the Resource Description Framework. These methods are being studied for use in cheminformatics and exchange of molecular data. This work has resulted in the development of CMLRSS, QSAR-ML, the Blue Obelisk Descriptor Ontology (BODO), and the CHEMINF ontology.

Molecular Representation

This branch of the research focuses on the presentation of molecular data, such that statistical methods can extract the most information from the data or generate the best prediction models. This research resulted in the development of a new descriptor for molecular crystal structures, the debunking of NMR spectra for some modeling approaches, and the development of applied tools, like the Chemistry Development Kit.

Statistics

Statistical methods help us find, understand and visualize complex patterns in our molecular data. This research focuses on improving the expert validation of our statistical models, by linking the models to external data. The latter brings us back to the semantic technologies to do that accurately. This work resulted in the application of supervised Self-Organizing Maps to classification problems with multiple end points.

Applied Cheminformatics

This part of the research is about getting the above methods used in other scientfic fields. This involves the development of tools that (unfortunately) hide much of the research from the above three fields, so that they can be easily used in other research fields. This research, particularly, is well accepted by the scientific community. Solutions here include the Chemistry Development Kit, Jmol, Bioclipse, and Oscar, which all make more fundamental research in Molecular Chemometrics more accessible.