Subgoal generation

A companion paper ) describes the subgoal-generating and synthesis-discovery subsystems of SYNCHEM in sufficient detail to enable the interested reader to make a critical evaluation of this work. We briefly synopsize these programs here to clarify the results to follow. 

Sridharan, N. S., Gelemter, H., Hart, A. J. Subgoal generation and synthesis search procedures in SYNCHEM, a computer program for the discovery of synthesis routes for organic molecules. In preparation. 

This capability is particularly useful during a run. Subgoal compounds generated as clauses and incorporated in the problem solving tree can be displayed on the screen in graphical format for user inspection. A complex data structure in the form of a doubly linked list is used by XTSYN to convert the molecule representations from one form to another (11). 

While having too many "chemical restrictions," the reaction rules have no "structural restrictions." In Fig. 6, which shows the first retro-synthetic steps SYNLMA considered for cocaine synthesis, we see that four Bredt s rule violations, enamines 27a,b and 28a.b were accepted as subgoals. While discussing Fig. 6, it should be noted that structures 22, 23, 24, and 26 are not allowed when constraints are on. Structures 23, 24, 27, and 28 are typical of current SYNLMA output. When it finds a reaction rule, it applies the rule exhaustively. Structures 29 and 30 are not synthetically demodulated and represent wasted CPU time. Finally, one second generation structure, 31, is shown because it represents an interesting variation of an N-oxide ene cycloaddition reaction that has been used to synthesize tropanol (16), the basic cocaine ring system. 

The next N-test checks for secondary alcohols, with no result (hence negative) indicated. Note that a call for subroutine 41 was not necessary here, since the desired information was discovered on the extended attribute table, where it had been placed in an earlier step. Since tertiary alcohols cannot be oxidized to the ketone, they do not affect this reaction, and no test is made for them. We remark that it is because this reaction will go in the presence of tertiary alcohols or diols (which can be protected with a cyclic ether, as specified by the seventh N-test in the list) that a simple A-test will not serve to exclude generating subgoals with primary or secondary alcohols. 

---