Elucidation of the structure of DNA by Watson and Crick [Nature 171 (1953) 737–738] has led to many crucial molecular experiments, including studies on DNA replication, transcription, physical mapping, and most recently to serious attempts directed toward the sequencing of large genomes [Watson, Science 248 (1990) 44–49]. I am totally convinced of the great importance of the Human Genome Project, and toward achieving this goal I strongly favor ‘topdown’ approaches consisting of the physical mapping and preparation of contiguous 50–100-kb fragments directly from the genome, followed by their automated sequencing based on the rapid assembly of primers by hexamer ligation together with primer walking. Our ‘top-down’ procedure totally avoids conventional cloning, subcloning and random sequencing, which are the elements of the present ‘bottom-up’ procedures. Fragments of 50–100kb are prepared in sufficient quantities either by in vitro excision with rare-cutting restriction systems (including Achilles' heel cleavage [AC] or the RecA-AC procedures of Koob et al. [Nucleic Acids Res. 20 (1992) 5831–5836]) or by in vivo excision and amplification using the yeast FRT/F1p system or the phage λ, att/Int system. Such fragments, when derived directly from the Escherichia coli genome, are arranged in consecutive order, so that 50 specially constructed strains of E. coli would supply 50 end-to-end arranged approx. 100-kb fragments, which will cover the entire approx. 5-Mb E. coli genome. For the 150-Mb Drosophila melanogaster genome, 1500 of such consecutive 100-kb fragments (supplied by 1500 strains) are required to cover the entire genome. The fragments will be sequenced by the SPEL-6 method involving hexamer ligation [Szybalski, Gene 90 (1990) 177–178; Fresenius J. Anal. Chem. 4 (1992) 343] and primer walking. The 18-mer primers are synthesized in only a few minutes from three contiguous hexamers annealed to the DNA strand to be sequenced when using an over 100-fold excess of hexamers and T4 DNA ligase at room temperature, preferably in the presence of the single-strand-binding (SSB) protein of E. coli. These 18-nt primers are immediately extended by the DNA polymerase, Sequenase 2.0, in the dideoxy sequencing reaction. Very high quality sequencing ladders are obtained for single-stranded DNA or denatured double-stranded approx. 50-kb fragments, as exemplified by phage λ DNA. When automated and used in conjunction with fluorescent dyes and ultrathin gels, the method should permit the sequencing of 500 nucleotides (nt) per 30 min i.e., 1 kb/h and 100 kb in less than a week per one sequencing channel. Automation has to include direct gel readout of over 500 nt, analysis of the terminal 50 nt, computerized selection and robotic assembly of 18-mers from three hexamers followed by their template-dependent ligation, sequencing reactions, instantaneous deproteinization, gel loading, electrophoresis, and again a gel readout followed by the next cycle. With 50 channels and all approx. 100-kb genomic fragments available (see above), one could project that automated sequencing of the entire E. coli genome should take about one week. Sequencing of both strands and larger genomes would require proportionally more time or more of the automated sequencing machines. There is little doubt in my mind that automated ‘top-down’ approaches are the key to the efficient and rapid sequencing of large genomes.