© 1999 Macmillan Magazines Ltd

helped to organize a meeting at ArgonneNational Laboratory which triggered sub-sequent meetings at NIH and elsewherethat resulted in the PSI.

All three agencies support the ProteinData Bank, which is managed by theResearch Consortium for Structural Bio-informatics to provide access to the morethan 10,000 known protein structures. Theaddition of 10,000 new structures throughthe PSI will provide the informationneeded to define the several thousand keyprotein folds that, in turn, should enableclassification of proteins into functionalcategories. The contributions of each ofthese agencies (and those of several non-federal organizations) will be needed toenable the NIGMS initiative to succeed.Marvin Cassman*, Ari Patrinos†*Director, National Institute of General MedicalSciences, National Institutes of Health, Bethesda,Maryland 20892-6200, USA†Associate Director for Biological andEnvironmental Research, Department of Energy,19901 Germantown Road, Germantown,Maryland 20874-1290, USA

frequently than they would by movingthrough the entire web.

Mike Gardner2 has rightly suggested thatwe need science-oriented search engineswith sets of scientific metadata, as metadataare the key to better searching. Togetherwith that proposal, an approach similar topeer-reviewing of scientific publicationscould be applied for categorizing andevaluating web pages based on theircontent, quality and subject-specificity. Itwould be feasible to use algorithms andrule-based expert systems to check content-richness, subject-specificity and freshness-based context-relevance ranking forretrieved results.

Millions of dollars need to be invested indeveloping search engines. This investmentcould be cost-effective if it resulted in analmost zero noise-to-signal ratio and pre-cise but comprehensive subject-relevanthits. The development work should be donein the academic sector, but the completedsearch engines would have commercialpotential, and would generate more revenuethan general search engines because of theirsubject-specificity.

Development of subject-specific searchengines would satisfy the growing demandfor the latest, precise, value-added, noise-free hits with a high level of subject rele-vance. Many such search engines togetherwould be able to index most of the World-Wide Web.Vishwas Chavan Centre for Cellular and Molecular Biology,Uppal Road, Hyderabad 500 007, India1. Lawrence, S. & Giles, C. L. Nature 400, 107–109 (1999).

2. Gardner, M. Nature 401, 111 (1999).

English, which is surely an attempt to reachreaders abroad.

It is interesting to learn that 40 per centof researchers submit papers only to Japan-ese publications, while 34.1 per cent alsosubmit to international journals. Does thatmean that 25.9 per cent only publish inoverseas journals? If true, this would showthat there are many internationally mindedJapanese researchers.

You do a good job of pointing out weak-nesses and important goals for Japaneseresearch. But your omission of comparativenumbers for elsewhere in the world doesnot allow readers to see the problem in itstrue proportions. It could even be the casethat your articles inadvertently promoteprejudices about Japanese science, whichwould be unfortunate at a time when manyJapanese research teams are striving toattract foreign scientists. Philippe BuhlmannDepartment of Chemistry, The University of Tokyo,Hongo 7-3-1 Bunkyo-ku, Tokyo 113-0033, Japan

correspondence

458 NATURE | VOL 402 | 2 DECEMBER 1999 | www.nature.com

Search is on for bettersearch enginesSir — Steve Lawrence and C. Lee Gilesstated in their Commentary that most ofthe popular search engines index onlyabout 7–16 per cent of the World-WideWeb1. This is alarming, as many scientificweb pages containing important data maynever be discovered. As the web grows it isgoing to become increasingly difficult forgeneral search engines to give compre-hensive coverage. The answer to theproblem could be the development ofsubject-specific search engines able tocover most of the contents within thatsubject.

Most currently available subject-specific lists and indexes are maintained byhumans. Many of them are merely collec-tions of web addresses and lack context-based relevance ranking and retrieval ofresults in multivariate combinations. Whatis needed are search engines that could tra-verse through pages at the last level in asubject-specific website. They would beable to do this as the numbers of such siteswould be within manageable limits.Crawlers or robots traversing through sucha subject-specific web subset could build upa comprehensive and complete bank of key-words. In turn, such keyword-mountedcrawlers would efficiently and more fre-quently screen the last-level page of the site.

Subject-specific search engines wouldbe able to maintain the freshness of the hits,as the crawlers would check a manageablenumber of specific web pages more

Japan builds bridges torest of the worldSir — Your Opinion and News articleabout research in Japan raise an importantand timely issue (Nature 401, 309 & 314;1999). A focus on higher quality ratherthan quantity of scientific output, andmore emphasis on cooperation withresearchers abroad, are indeed importantgoals for Japanese science. But unfortu-nately the articles did not compare theJapanese situation with that elsewhere.

You report, for example, that 40 per centof researchers submit papers only toJapanese journals, which is said to reflectinsularity. But how can the reader drawsuch a conclusion without knowingcomparable numbers for other countries? Iwould not be surprised if 40 per cent ofresearchers in the United States, the UnitedKingdom or Germany published only innational journals. And you did notmention that a considerable number ofJapanese journals publish articles only in

Galileo had accuratevision of the MoonSir — Martin Kemp writes that “the detailsof Galileo’s cratered Moons are difficult toalign precisely with actual features”1. Iwould like to draw readers’ attention to thework of Ewen Whitaker2, a selenographerat the University of Arizona. Whitakerreviewed earlier efforts at identifying thefeatures drawn by Galileo, and took specialnote of the important contributions ofGuglielmo Righini, Owen Gingerich andStillman Drake. He provided side-by-sidecomparisons of Galileo’s drawings withmodern photographs taken at the samelunar phases, and these provide strikingsupport for his contention that Galileo didindeed observe, and record, veryaccurately.

Whitaker had examined the copperplateengravings of the first edition of Galileo’sSiderius Nuncius and seven manuscriptimages, and he commented on “previouslyunnoticed differences between the manu-script and printed versions of Siderius Nun-cius... In the original edition these engrav-ings present a reasonably well-executedappearance, but subsequent editions utilizewoodcuts, and the quality deteriorates veryrapidly to the point where they are virtuallyunrecognizable as Moon images. Some ofthe disparaging remarks made about thedrawings undoubtedly stem from examina-tions of these cruder images”.M. W. FriedlanderDepartment of Physics, Washington University,St Louis, Missouri 63130, USA

1. Kemp, M. Nature 401, 116 (1999).

2. Whitaker, E. J. Hist. Astron. 9, 155–169 (1978).