PubMed2002
In silico biology has gathered momentum as, worldwide, scientists have united in a common quest to sequence, store and analyse complete genomes. This year, a pivotal achievement of this cooperative endeavour was realised in the release of a public draft of the human genome, and with it the promises to improve our understanding of diverse aspects of biology and to yield a healthier future with safe personalized medicines. Key to these goals will be the need to elucidate and characterise the genes and gene products encoded not just in the human genome, but in many genomes. These tasks are underpinned by the concepts and processes of genome and gene/protein evolution, regulation of gene expression, mechanisms of protein folding, the manifestation of protein function, and so on, all of which must be understood in the context of complex, dynamic biological systems. Our use of computers to model such concepts and systems must be placed in the context of the current limits of our understanding of them:- it is important to recognise, for example, that we don't have a common understanding either of what constitutes a gene or a protein function; we can't invariably say that a particular sequence or fold has arisen via divergent or convergent evolution; and we don't fully understand the rules of protein folding. Accepting what we can't do in silico is essential in appreciating what we can do. Without this understanding, it is easy to be misled, as notions of what particular computational approaches can achieve are sometimes rather optimistic. There are valuable lessons to be learned here from the field of Artificial Intelligence, principal among which is the realisation that capturing and representing complex knowledge is time consuming, expensive and hard. Thus, we argue here that if bioinformatics is to tackle biological complexity in earnest, it would be wise to absorb the experience distilled from decades of artificial intelligence research, and to approach the road ahead with caution, rigour and pragmatism.