Developing Agents for Bioinformatics Applications: a Preliminary Design

摘要

In many fields of science, data is accumulating much faster than our ability to convert it into meaningful knowledge. This is perhaps nowhere more true than in the biological sciences where the Human Genome Project and related activities have flooded our databases with molecular data. Our modeling tools are woefully inadequate for the task of integrating all that information into the rest of biology, preventing scientists to effectively take advantage of these data in drawing meaningful biological inferences. Thus, one of the major challenges faced by computer scientists and biologists together is the enhancement of information technology suitable for modeling a diversity of biological relationships and processes, leading to a greater understanding from the influx of data. Instead of allowing the direct expression of high-level concepts natural to a scientific discipline, current software development techniques require mastery of computer science and access to very low level aspects of software development in order to construct significantly complex applications. Even in places where attempts to introduce domain-specific concepts have been made—e.g., design of database formats—scientists are hampered in their efforts by complex issues of interop-eration. As a result, currently only biologists with strong quantitative skills and high com- puter literacy can realistically be expected to undertake the task of transforming the massive amounts of available data into real knowledge. Very few scientists (domain experts) have such computing skills; even if they do, their skills are better utilized in dealing with high-level scientific models than low-level programming issues. To enable scientists to effectively use computers, we need a well-developed methodology, that allows a domain expert (e.g., a biologist) to solve a problem on a computer by developing and programming solutions at the same level of abstraction they are used to think and reason, thus moving the task of programming from software professionals to the domain experts, the end-users of information technology. This approach to software engineering is commonly referred to as Domain Specific Languages and it has been advocated by many researchers over the years.