In addition to maintaining the GenBank® nucleic acid sequence database, the National Center for Biotechnology Information (NCBI) provides analysis and retrieval resources for the data in GenBank and other biological data made available through the NCBI Web site. NCBI resources include Entrez, the Entrez Programming Utilities, MyNCBI, PubMed, PubMed Central (PMC), Entrez Gene, the NCBI Taxonomy Browser, BLAST, BLAST Link (BLink), Primer-BLAST, COBALT, Electronic PCR, OrfFinder, Splign, ProSplign, RefSeq, UniGene, HomoloGene, ProtEST, dbMHC, dbSNP, dbVar, Epigenomics, Cancer Chromosomes, Entrez Genomes and related tools, the Map Viewer, Model Maker, Evidence Viewer, Trace Archive, Sequence Read Archive, Retroviral Genotyping Tools, HIV-1/Human Protein Interaction Database, Gene Expression Omnibus (GEO), Entrez Probe, GENSAT, Online Mendelian Inheritance in Man (OMIM), Online Mendelian Inheritance in Animals (OMIA), the Molecular Modeling Database (MMDB), the Conserved Domain Database (CDD), the Conserved Domain Architecture Retrieval Tool (CDART), IBIS, Biosystems, Peptidome, OMSSA, Protein Clusters and the PubChem suite of small molecule databases. Augmenting many of the Web applications are custom implementations of the BLAST program optimized to search specialized data sets. All of these resources can be accessed through the NCBI home page at www.ncbi.nlm.nih.gov.

With the development oflarge data banks ofprotein and nucleic acid sequences, the need for efficient methods of searching such banks for sequences similar to a given sequence has become evident. We present an algorithm for the global comparison ofsequences based on matching k-tuples ofsequence elements for a fixed k. The method results in substantial reduction in the time required to search a data bank when compared with prior techniques of similarity analysis, with minimal loss in sensitivity. The algorithm has also been adapted, in a separateimplementation, to produce rigorous sequence alignments. Currently, using the DEC KL-10 system, we can compare all sequences in the entire Protein Data Bank ofthe National Biomedical Research Foundation with a 350-residue query sequence in less than 3 min and carry out a similar analysis with a 500-base query sequence against all eukaryotic sequences in the Los Alamos Nucleic Acid Data Base in less than 2 min.As the number of protein molecules and nucleic acid fragments for which the sequences have been determined has grown into the thousands (the total number of nucleotides so analyzed is now more than one million), it The Algorithm. We shall here describe how two sequences, Si and S2, of lengths N1 and N2, respectively, are to be compared. As motivation, it is useful to think in terms of the dot matrix comparison of Si and S2 (11) in which the beginnings of both sequences are placed to the upper left of the matrix and one sequence is positioned horizontally and the other is positioned vertically. The diagonals running downward from left to right in the dot matrix illustrate the degree of similarity that would be found by a simple sliding comparison with the different possible choices of alignment register. Frequently, one can look at the dot matrix comparison and immediately see certain diagonals that appear to have a number of points above background and, therefore, indicate a level of similarity for the two sequences in certain regions. It is generally true that these significant diagonals are still clearly visible when the dot matrix is filtered to only show matches of length k or greater, where k is a small positive integer. For this reas'on, our attention will be directed to such k-tuples.The first step in the algorithm is the location ofall the k-tuple matches between the sequences Si and S2. In precise terms, a k-tuple match consists of two k-tuples-S1(i),Si(i + 1), . . .,Sl(i + k -1) from Si and S2(j),S2(j + 1), . . .,S2x (1 + k -1) from S2-that are identical. If there are p elements in the alphabet from which the sequences are made, then there are pk possible different k-tuples. To locate all k-tuple matches, we follow a method described by Dumas and Ninio (12). We have chosen a simple method (there are many possible) of converting any k-tuple into an integer between 1 and pk.

In addition to maintaining the GenBank® nucleic acid sequence database, the National Center for Biotechnology Information (NCBI) provides analysis and retrieval resources for the data in GenBank and other biological data made available through the NCBI Website. NCBI resources include Entrez, the Entrez Programming Utilities, MyNCBI, PubMed, PubMed Central (PMC), Gene, the NCBI Taxonomy Browser, BLAST, BLAST Link (BLink), Primer-BLAST, COBALT, Splign, RefSeq, UniGene, HomoloGene, ProtEST, dbMHC, dbSNP, dbVar, Epigenomics, Genome and related tools, the Map Viewer, Model Maker, Evidence Viewer, Trace Archive, Sequence Read Archive, BioProject, BioSample, Retroviral Genotyping Tools, HIV-1/Human Protein Interaction Database, Gene Expression Omnibus (GEO), Probe, Online Mendelian Inheritance in Animals (OMIA), the Molecular Modeling Database (MMDB), the Conserved Domain Database (CDD), the Conserved Domain Architecture Retrieval Tool (CDART), Biosystems, Protein Clusters and the PubChem suite of small molecule databases. Augmenting many of the Web applications are custom implementations of the BLAST program optimized to search specialized data sets. All of these resources can be accessed through the NCBI home page at www.ncbi.nlm.nih.gov.