CMBI - LHASA: Perception and Functional Groups

Centre for Molecular and Biomolecular Informatics
Radboud University Nijmegen

LHASA Perception and Functional Groups

The first stage in the processing of a target structure as drawn by the user is the perception. The initial structural information available to LHASA is just a connection table with `simple' bond orders (single, double, and triple) and with stereochemistry indicated by dashed and/or wedged bonds. LHASA analyzes this table to detect rings, aromaticity, tautomerism, absolute stereochemistry, symmetry, and functional groups.

Some structural features prompt LHASA to reject a target structure: obviously valency violations make a structure invalid, but also the presence of both defined and undefined stereocenters is not allowed. Impossible structures such as a cyclopropyne or a trans-bridged small ring are not accepted either. Since these illegal structures are awkward to diagnose using traditional programming languages, a small portion of the knowledge base is devoted to their identification. The next step in perception is the detection of features that, while not impossible, are deemed too unstable, e.g., primary enamines, gem-diols, and potentially aromatic rings (an example is cyclohexadienone which would enolize to phenol). Unstable features also lead to rejection of the structure because they would not survive any reactions.

When a structure has passed all the tests for impossible or unstable features, the functional groups are scrutinized further. It is important to note that LHASA recognizes only 78 different functional groups. For each of these groups, a pattern has been defined to describe the group; anything not matching one of the patterns is not recognized as a functional group. Examples of groups presently unrecognized by LHASA are thioketone, acylsilane, ketene, and sulfonamide. The 78 functional groups are divided into three categories called core, masked, and reactive groups. Core functionality includes all common functional groups such as ketone, olefin, amine, and acid. Most transforms in the knowledge base are keyed by core groups. In contrast, masked groups are less common groups which are often `disguised' versions of core groups. Examples are ketals, epoxides, oximes, and TMS ethers. The category of reactive groups consists of functional groups that are likely to interfere with almost any reaction, e.g., acid halide, nitroso, and isocyanate. A small section of the knowledge base is devoted to removing such groups. When LHASA detects a reactive group in the target, these transforms are automatically applied before any strategy is selected. The precursor(s), now containing core functionality, can be re-submitted to the program.

Since few transforms have non-core functional groups in their retrons, it is usually advisable to let LHASA convert masked groups into core groups. At least this offers a reasonable alternative target molecule. In fact, LHASA asks the user to apply unmasking transforms whenever masked functionality is detected. This policy can be thought of as a strongly suggested strategy. Unmasking transforms are simply goal transforms which are flagged as unmasking. Thus, even unrecognized functionality can be unmasked since it is possible to write the unmasking transform as a pattern transform, independent of functional group perception.

Why is recognizing functional groups so important to LHASA? Firstly, functional groups are naturally essential in the retron descriptions of one-group, two-group, FGI, and FGR transforms. Secondly, functional groups are used to assess the effects of reaction conditions on portions of the molecule that are not supposed to react (vide infra).