Introduction

The Savant interface is centered around three HTML pages: one which allows the user to specify a structure of interest, one which shows what data is available for that structure and similar structures, and one which shows the relevant references. These three pages are described below.

Savant

Select your languge orientation preference. The default is "International", i.e., journal articles in all languages are displayed.

Select the number of structures to be displayed. The default is 10, which works well in most cases. If the structure of interest is of a very common or very unusual chemical class, it might be useful to increase this number. If you are only interested in an exact match, set this to "1".

Enter a SMILES for a structure of interest chemical in the space provided and click on the button marked "Submit". You may copy an entry from somewhere else and paste it into the entry field (the specific method depends on the HTML browser being used), or use a previously-configured editor, if available.

The machine might quietly crunch for a while at this point (seconds to a minute, depending on the server hardware and load).

Savant -- search results

A "Query Synopsis" is presented at the top of the page immediately after the page header. This synopsis includes an icon-sized depiction of the query structure and a statment of the current preferences.

A list of similar structures for which desired references are available is presented next, as a graphical index. Each entry consists of an icon-sized depiction, a measure of similarity, and a textual summary of how many references are available for the structure. If appropriate synthetic reference(s) are available for the query structure itself, it will appear first, with a similarity metric of 1.0.

The list is sorted by similarity, with 1.0 meaning "completely similar" (typically, "identical") and 0.0 meaning "completely dissimilar". The similarity metric used is the bitwise Tanimoto coefficient (see the Daylight Theory Manual for more information). To a rough approximation, you can think of this metric as "the fraction of all possible substructures in both structures being compared which are in common". Values of less than about 0.65 are not very meaningful.

If you are connected to the Spresi database (rather than Spresipreps), the search might take a bit longer and you might see index summaries such as "0.96 similar, 14/181 journal articles, 1/41 patents", indicating that one (or more) synthetic patents were also found.

At the very bottom of the page, you may see a line indicating how many of the most similar structures had suitable references. This only appears if some of the most similar structures were eliminated from the list for various reasons (i.e., those that have no synthetic references in a desired language).

The icon-sized depictions are "hot": click on the depiction to show the corresponding references.

Savant -- synthetic references

For patents, the patent owner, class, number, and country are also included. The patent country is a two-letter abbrevation. Most are obvious, but some can be confusing unless you remember that many of these countries don't exist any more, e.g., SU stands for "Soviet Union" DD for "Deutsche Demokratische Republik" (East Germany), DB for "Deutsche Bundesrepublik" (West Germany), etc.