Sir -- It was encouraging to read the News article on the possibility that the US National Institutes of Health might invest more in computing infrastructure (Nature 398, 93-94; 1999). But, echoing Larry Smarr's comment, one wonders whether 'big iron' supercomputers are really the way to go.

Many biological calculations increasingly emphasize other aspects of computer hardware than fast central processing units for floating-point calculations. Scientists working on large-scale genome comparisons need lots of storage (on disk and in memory) and high-speed networking perhaps more than 'mega-flops'.

And those building biological databases are more concerned with the development of common data formats and easy-to-use devices for the bulk entry of laboratory data. One could even argue that these new genome- and database-oriented calculations represent a different paradigm of scientific computing from physically oriented simulations, in need of a distinct but still extensive computational infrastructure.

Mark Gerstein
Molecular Biophysics & Biochemistry Department, Yale University, Bass 432A, 266 Whitney Avenue, New Haven, Connecticut 06520-8114, USA