In the course of evolution, mutations do not affect both strands of genomic DNA equally. This imbalance mainly results from asymmetric DNA mutation and repair processes associated with replication and transcription. In prokaryotes, prevalence of G over C and T over A is frequently observed in the leading strand. The sign of the resulting TA and GC skews changes abruptly when crossing replication-origin and termination sites, producing characteristic step-like transitions. In mammals, transcription-coupled skews have been detected, but so far, no bias has been associated with replication. Here, analysis of intergenic and transcribed regions flanking experimentally identified human replication origins and the corresponding mouse and dog homologous regions demonstrates the existence of compositional strand asymmetries associated with replication. Multiscale analysis of human genome skew profiles reveals numerous transitions that allow us to identify a set of 1,000 putative replication initiation zones. Around these putative origins, the skew profile displays a characteristic jagged pattern also observed in mouse and dog genomes. We therefore propose that in mammalian cells, replication termination sites are randomly distributed between adjacent origins. Taken together, these analyses constitute a step toward genome-wide studies of replication mechanisms.replication termination ͉ wavelet transform ͉ compositional bias ͉ skewness C omprehensive knowledge of genome evolution relies on understanding mutational processes that shape DNA sequences. Nucleotide substitutions do not occur at similar rates, and, in particular, owing to strand asymmetries of the DNA mutation and repair processes, they can affect each of the two DNA strands differently. Asymmetries of substitution rates coupled to transcription have been observed in prokaryotes (1-3) and in eukaryotes (4-6). Strand asymmetries (i.e., G C and T A) associated with the polarity of replication have been found in bacterial, mitochondrial, and viral genomes, where they have been used to detect replication origins (7-9). In most cases, the leading replicating strand presents an excess of G over C and of T over A. Along one DNA strand, the sign of this bias changes abruptly at the replication origin and at the terminus. In eukaryotes, the situation is unclear. Several studies failed to show compositional biases related to replication, and analyses of nucleotide substitutions in the region of the -globin replication origin did not support the existence of mutational bias between the leading and the lagging strands (8, 10, 11). In contrast, strand asymmetries associated with replication were observed in the subtelomeric regions of Saccharomyces cerevisiae chromosomes, supporting the existence of replication-coupled asymmetric mutational pressure in this organism (12). Here, we present analyses of strand asymmetries flanking experimentally determined human replication origins, as well as the corresponding mouse and dog homologous regions. Our results demonstrate the exis...