The aim of bioinformatics is to apply computational approaches and information technology to facilitate the organisation and analysis of biological information. One of the most exciting aspects of bioinformatics is that it is by definition an interdisciplinary field of science, combining molecular biology, information technology, and mathematics into a single discipline. I initially trained in mathematics and computer science, but since the beginning of my work in bioinformatics at the EMBL in Heidelberg and afterwards at the IGBMC in Strasbourg, I have been lucky enough to be able to work in close day-to-day contact with expert molecular biologists. My interest has focused on the analysis of protein sequences, not as a series of individual letters, but because of what they represent; proteins with complex three-dimensional structures, biological functions, and interactions with other proteins.

One of the cornerstones of modern bioinformatics is the comparison or alignment of protein sequences. Pairwise comparisons are used to search the sequence databases for homologous proteins that share a common evolutionary history and often have similar 3D structures and biological functions. With the aid of multiple sequence alignments of the various members of a complete protein family, scientists are able to study the sequence patterns conserved through evolution and the ancestral relationships between different organisms. A close collaboration between biologists and computer scientists is one of the main reasons for the success of our multiple sequence alignment programs, ClustalW and ClustalX and our more recent developments, in particular DbClustal and NorMD. Only expert biologists working in the field can evaluate the results produced by automatic methods in terms of their biological accuracy and significance. Often experimental evidence is required to confirm the predictions made by computer programs, otherwise they remain only unproved hypotheses. Thus, the association of bioinformatics groups with structural biologists and biochemists is essential. But mathematics and information technology can contribute different, complementary skills, such as rigorous mathematical theories and computer engineering techniques for program design and development. Although this is sometimes considered to be of secondary importance by theoretical mathematicians, computer programs will not be used by most biologists if they are not easy to install, robust, and user-friendly. Another crucial aspect of the development of a successful program is the use of a suitable test system to validate the results. We developed the BAliBASE benchmark alignment database specifically for the evaluation of multiple alignment methods. It has been a crucial factor in the development and validation of our new methods and has become a standard reference for others working in the field.

Recent advances in genome sequencing technologies and gene expression analysis using microarrays have led to an explosive growth in the amount of biological information publicly available. There is now an urgent requirement for the creation and management of information databases and automatic tools to analyse and interpret the various types of data. In particular, the multiple sequence alignment problem has been the subject of renewed interest from computer scientists and mathematicians, and a number of different algorithms have been exploited in the search for more accurate alignments, including iterative refinement techniques, genetic algorithms, and Hidden Markov Models. Nevertheless, it is generally acknowledged that a human expert is still required to validate and correct the automatic alignments in more difficult cases. No single algorithm exists today that can cope with the highly complex proteins detected by today’s database search programs, and it is clear that an accurate, biologically meaningful alignment can no longer be constructed from the primary sequence data alone. In order to fully understand the functions and molecular interactions of a particular gene, it will be essential to assemble, validate, and classify such diverse information as cellular location, degradation and modification, three-dimensional structures, mutations, and their associated illnesses. Much of this data is itself predicted by computerised methods and is therefore inherently unreliable. Validation of this information will be of primary importance to the successful development of new sequence-analysis tools in the years to come. Current research in our laboratory is therefore moving towards a co-operative, knowledge-based approach bringing together pertinent information and complementary algorithmic techniques into a single, integrated system.