In evaluating the effectiveness of this methodology, the nature of PaSS must first be considered, as it plays the most important part in the method. As PaSS is calculated on the basis of the coordinates of spiddos (see Equation 1 in Materials and methods), the nature of spiddos must be thoroughly investigated. If two genomic DNAs contain common sequence regions that can be amplified by random PCR using the same primer, the resultant DNAs will usually generate similar spiddos (by definition, the spiddos obtained by TGGE represent the crucial points of a genome profile, points at which the temperature corresponds to the beginning of a prominent structural transition in DNA [15]). As shown schematically in Figure 3, two corresponding spiddos derived from two closely related species can be connected by a displacement vector, which consists of two independent elements of mobility (μ) and temperature (θ). The differences in each element (Δμ and Δθ) can be related to the differences between two sequences as shown in Figure 3. The displacement in the ordinate is caused by the difference in length between the two DNAs and is caused by deletion or insertion, whereas that in the abscissa is mainly caused by point mutation (although insertion/deletion can also contribute).

As the extent of these changes is roughly proportional to the evolutionary time since the species diverged, we can expect that the summation of the displacement of each spiddo is approximately proportional to the time since divergence, and thus to the genome-to-genome distance. By using a sufficient number of spiddos, we can obtain statistically reliable results. Empirically, we know that 8-10 spiddos, which can be obtained from a single genome profile, can be significant. However, since the more spiddos the better the result, we tentatively made it a rule to adopt four genome profiles (≈ 32-40 spiddos) - that is, four random-PCR products - as a current standard of initial species identification. Therefore, PaSS has the theoretical and empirical basis to be used as a measure of similarity between genomes, although the extent of its effectiveness remains to be shown experimentally as data accumulates. We have introduced a measure of distance, d', obtained from PaSS as formulated in Equation 2 (see Materials and methods), for the sake of convenience [15].

We call d' a genome sub-distance because it is based not on the whole but a part of the genome sequence. Thus, we introduce (true) genome distance, d, as in Equation 3 (see Materials and methods). Genome distance must have a close relationship with genetic distance, as defined by Nei and others [17,18,19], although there is a difference in the definition. The genetic distance based on sequences is basically the Hamming distance (the number of different letters at each corresponding position of two sequences of letters that are optimally aligned) between two nucleotide (or amino-acid) sequences. In aligning sequences arbitrariness is introduced, depending on the algorithm and parameters used [20]. Another constraint on genetic distance is that it is usually obtained from a limited number of genes, although that is also the case for genome distance. As genome distance is easier to obtain in practice using our method, it should be easier to obtain a lot of data on it compared with genetic distance. On the basis of d (in practice d'), we can construct phylogenetic trees and genome sequence space (an imaginary spherical space in which all the genomes (individuals) can be uniquely located in a finite manner based on the distance between genomes, providing clusters of species (K.N., unpublished observations)).

Although the applicability and effectiveness of genome distance for such purposes needs to be further investigated, it is obvious that an organism that has near-zero genome distance from a certain standard species, as an average over four or more genome sub-distances obtained from as many genome profiles, can be easily assigned to that same species with a high level of confidence. We are not claiming, however, to be able to give the correct taxonomical name to any species using this method. The greater the number of strains registered in the database, the more easily will a species be assigned. Basically, no special efforts, except expanding the database and using sophisticated algorithms, are necessary to raise the proportion of correct assignments. This is the self-developing nature of the database. Therefore, this methodology has two potential great advantages for tentative species identification: first, expertise is not always necessary; and second, database building can be carried out in a self-developing manner (that is, by acquiring more and more accurate data on species) with no waste of information.

Although genome profiling is the basic technology for our current purpose, provisional species identification based on genotype can be fulfilled only by using computer-aided database technology, which is most effectively constructed in the Internet environment. As this methodology is based on use by a large number of scientists, the protocol must be designed to be reproducible and easy to carry out. The processes have been deliberately designed with this in view, and are presented on our website [21].

Genome profiling consists of two basic technologies: random PCR and TGGE, which have been well established [22,23,24]. However, if it is to be used for the purpose of general and universal applications, well-defined standardization is absolutely required to obtain significant results. We have carried out such standardization for genome profiling. The main topics included in the protocol are: preparation of genome DNAs; the set of primers used for random PCR and the internal reference DNAs used for TGGE; experimental conditions for random PCR; and the experimental conditions for TGGE. The protocol also includes the related procedures (extraction of spiddos, calculation of PaSS, and others).

Although the genome profile is a kind of reduction of information contained in the whole genome sequence, it is still too complicated to deal with as it is. Thus, a second reduction is carried out by extracting feature points (spiddos) from the genome profiles. Double-stranded DNAs are known to melt in an intrinsically determined manner, depending on their sequence, when heated gradually [27]. All the intermediate states of DNA have their own structure and mobility in gel. Spiddos correspond to the structural transition points appearing in band patterns (Figure 4). Currently, there are four kinds of spiddos: initial melting point (P_ini); minimum mobility point (P_min); isomobility point (P_iso); and the end melting point (P_end). Empirically, P_ini is the most reproducible. Therefore, P_ini is recommended for working spiddos wherever possible. Further details are given in the standard protocol on our website [21].

A set of spiddos (around ten), assigned to a genome profile on a computer display, is processed to calculate the normalized mobility and temperature of each point. A measure of similarity of two genomes - the PaSS - is introduced as follows.

of each spiddo (1 to n)is its position vector and is a function of temperature and mobility (that is, || = P (T, m)). The superscripts 1 and 2 in parentheses in Equation (1) represent genomes 1 and 2, respectively. PaSS will be unity for a complete match in two sets of spiddos. In general, 0 ≤ PaSS ≤ 1. Genome distance and genome sub-distance (d') are derived from PaSS as follows:

d' = (1 - PaSS)/PaSS (2)

Where d'(i) is the ith genome sub-distance obtained with the ith primer used for random PCR.

The overall process of obtaining an on-web genome profile is shown in Figure 5. There are two steps in this methodology: the local phase and the database phase. In the local phase, genome profiling is carried out for the organism of interest, following the standard protocol presented on our website [21] and outlined in the previous sections. After obtaining a genome profile, the database is accessed and the database phase is begun as a client. The database site requires the client to input an image of the genome profile, to assign spiddos on the genome profile (Figure 4), and to fill in relevant data on the online form. The site will search the database for species with the most similar pattern of spiddos by calculating the PaSS [15].