Introduction
This WWW-based interactive chemoinformatics data service determines sets of non-overlapping structural fragments from specific fragment collections which can be assembled to form a complete, user-defined molecule. Such fragment sets are, for example, useful for the prediction of physico-chemical properties by group contribution methods.

A complete description of the program can be found in the PDF Manual.

The fragment set used for the analysis is not fixed. Enterprising users can modify the fragment definitions which are stored in the text box on the lower left side of the input area before submitting their structure data. Two default schemes (UNIFAC and Sedlbauer-Majer) are predefined.

Input of Structures
The simplest way to compute the fragment set of a structure is to use the Java editor (JME, courtesy of P. Ertl/Novartis) in the panel to the left. Simply draw a structure and press the <Analyze> button. The results will be displayed in this area.

Alternatively, it is possible to upload a structure file. Use the file selector below the editor to select a structure file from your local disk. Its format is automatically detected. Note that we support only standard structure exchange formats, not binary editor-specific file formats. Useful formats include, for example, MDL SD-files and SMILES files. Files can have multiple records, although a maximum of 1000 records are processed at a time to avoid server overload.

Structure Analysis
The algorithm behind this service will attempt to find sets of non-overlapping fragments which completely cover all non-hydrogen atoms of the submitted structures.

It can both happen that no solution is found (for example, because an exotic combination of structural features and/or elements is encountered which is not covered by the fragments in the collection) or that multiple solutions of comparable quality are generated. Duplicate solutions with the same selection of fragments are suppressed. Every found combination of fragments is output only once.

For reasons of performance and general usefulness, the algorithm will not generate an exhaustive list of solutions. Rather, it will preferably use larger fragments. If a part of the structure can be covered by a large fragment and also by a combination of smaller, less specific fragments, only solutions with the larger, more specific fragment are presented. Fragments which contain rings have a higher relative precedence than non-ring fragments. The algorithm uses a couple of heuristics to limit the effects of combinatorial explosion for larger structures.

Usage Conditions
This service can be used free of charge by anybody. However, we log the submitted structures and reserve the right to use them for analysis of algorithm performance and potential problems, and to publish results, including the exact formula of processed structures and their origin. Please do not submit any confidential structures.

Users accessing the service by means of robotic scripts will be denied future access, unless prior approval was obtained from the maintainer.

In-house versions of this software are available for licensing.